Visit https://ebookgate.

com to download the full version and

explore more ebooks

Data Mining A Tutorial Based Primer 2nd Edition

Richard J. Roiger

_____ Click the link below to download _____

https://ebookgate.com/product/data-mining-a-tutorial-
based-primer-2nd-edition-richard-j-roiger/

Explore and download more ebooks at ebookgate.com

Here are some recommended products that might interest you.
You can download now and explore!

Making Sense of Data A Practical Guide to Exploratory Data

Analysis and Data Mining 1st Edition Glenn J. Myatt

https://ebookgate.com/product/making-sense-of-data-a-practical-guide-
to-exploratory-data-analysis-and-data-mining-1st-edition-glenn-j-
myatt/
ebookgate.com

Data Mining Using SAS Enterprise Miner A Case Study

Approach 2nd Edition

https://ebookgate.com/product/data-mining-using-sas-enterprise-miner-
a-case-study-approach-2nd-edition/

ebookgate.com

Clustering for Data Mining A Data Recovery Approach 1st

Edition Boris Mirkin

https://ebookgate.com/product/clustering-for-data-mining-a-data-
recovery-approach-1st-edition-boris-mirkin/

ebookgate.com

Data Mining and Analysis Fundamental Concepts and

Algorithms Mohammed J. Zaki

https://ebookgate.com/product/data-mining-and-analysis-fundamental-
concepts-and-algorithms-mohammed-j-zaki/

ebookgate.com
Encyclopedia of Data Warehousing and Mining 2nd Edition
John Wang

https://ebookgate.com/product/encyclopedia-of-data-warehousing-and-
mining-2nd-edition-john-wang/

ebookgate.com

Data Mining 3rd ed Edition Jiawei Han

https://ebookgate.com/product/data-mining-3rd-ed-edition-jiawei-han/

ebookgate.com

Statistical Data Mining Using SAS Applications Second

Edition Chapman Hall CRC Data Mining and Knowledge
Discovery Series George Fernandez
https://ebookgate.com/product/statistical-data-mining-using-sas-
applications-second-edition-chapman-hall-crc-data-mining-and-
knowledge-discovery-series-george-fernandez/
ebookgate.com

Handbook of Statistics 24 Data Mining and Data

Visualization C.R. Rao

https://ebookgate.com/product/handbook-of-statistics-24-data-mining-
and-data-visualization-c-r-rao/

ebookgate.com

R Data Mining Blueprints 1st edition Edition Mishra

https://ebookgate.com/product/r-data-mining-blueprints-1st-edition-
edition-mishra/

ebookgate.com
DATA MINING
A Tutorial-Based Primer
SECOND EDITION
 Chapman & Hall/CRC
Data Mining and Knowledge Discovery Series

SERIES EDITOR
Vipin Kumar
University of Minnesota
Department of Computer Science and Engineering
Minneapolis, Minnesota, U.S.A.

AIMS AND SCOPE

This series aims to capture new developments and applications in data mining and knowledge
discovery, while summarizing the computational tools and techniques useful in data analysis. This
series encourages the integration of mathematical, statistical, and computational methods and
techniques through the publication of a broad range of textbooks, reference works, and hand-
books. The inclusion of concrete examples and applications is highly encouraged. The scope of the
series includes, but is not limited to, titles in the areas of data mining and knowledge discovery
methods and applications, modeling, algorithms, theory and foundations, data and knowledge
visualization, data mining systems and tools, and privacy and security issues.

PUBLISHED TITLES
ACCELERATING DISCOVERY: MINING UNSTRUCTURED INFORMATION FOR
HYPOTHESIS GENERATION
Scott Spangler
ADVANCES IN MACHINE LEARNING AND DATA MINING FOR ASTRONOMY
Michael J. Way, Jeffrey D. Scargle, Kamal M. Ali, and Ashok N. Srivastava
BIOLOGICAL DATA MINING
Jake Y. Chen and Stefano Lonardi
COMPUTATIONAL BUSINESS ANALYTICS
Subrata Das
COMPUTATIONAL INTELLIGENT DATA ANALYSIS FOR SUSTAINABLE
DEVELOPMENT
Ting Yu, Nitesh V. Chawla, and Simeon Simoff
COMPUTATIONAL METHODS OF FEATURE SELECTION
Huan Liu and Hiroshi Motoda
CONSTRAINED CLUSTERING: ADVANCES IN ALGORITHMS, THEORY,
AND APPLICATIONS
Sugato Basu, Ian Davidson, and Kiri L. Wagstaff
CONTRAST DATA MINING: CONCEPTS, ALGORITHMS, AND APPLICATIONS
Guozhu Dong and James Bailey
DATA CLASSIFICATION: ALGORITHMS AND APPLICATIONS
Charu C. Aggarwal
DATA CLUSTERING: ALGORITHMS AND APPLICATIONS
Charu C. Aggarwal and Chandan K. Reddy
DATA CLUSTERING IN C++: AN OBJECT-ORIENTED APPROACH
Guojun Gan
DATA MINING: A TUTORIAL-BASED PRIMER, SECOND EDITION
Richard J. Roiger
DATA MINING FOR DESIGN AND MARKETING
Yukio Ohsawa and Katsutoshi Yada
DATA MINING WITH R: LEARNING WITH CASE STUDIES, SECOND EDITION
Luís Torgo
EVENT MINING: ALGORITHMS AND APPLICATIONS
Tao Li
FOUNDATIONS OF PREDICTIVE ANALYTICS
James Wu and Stephen Coggeshall
GEOGRAPHIC DATA MINING AND KNOWLEDGE DISCOVERY,
SECOND EDITION
Harvey J. Miller and Jiawei Han
GRAPH-BASED SOCIAL MEDIA ANALYSIS
Ioannis Pitas
HANDBOOK OF EDUCATIONAL DATA MINING
Cristóbal Romero, Sebastian Ventura, Mykola Pechenizkiy, and Ryan S.J.d. Baker
HEALTHCARE DATA ANALYTICS
Chandan K. Reddy and Charu C. Aggarwal
INFORMATION DISCOVERY ON ELECTRONIC HEALTH RECORDS
Vagelis Hristidis
INTELLIGENT TECHNOLOGIES FOR WEB APPLICATIONS
Priti Srinivas Sajja and Rajendra Akerkar
INTRODUCTION TO PRIVACY-PRESERVING DATA PUBLISHING: CONCEPTS
AND TECHNIQUES
Benjamin C. M. Fung, Ke Wang, Ada Wai-Chee Fu, and Philip S. Yu
KNOWLEDGE DISCOVERY FOR COUNTERTERRORISM AND
LAW ENFORCEMENT
David Skillicorn
KNOWLEDGE DISCOVERY FROM DATA STREAMS
João Gama
MACHINE LEARNING AND KNOWLEDGE DISCOVERY FOR
ENGINEERING SYSTEMS HEALTH MANAGEMENT
Ashok N. Srivastava and Jiawei Han
MINING SOFTWARE SPECIFICATIONS: METHODOLOGIES AND APPLICATIONS
David Lo, Siau-Cheng Khoo, Jiawei Han, and Chao Liu
MULTIMEDIA DATA MINING: A SYSTEMATIC INTRODUCTION TO
CONCEPTS AND THEORY
Zhongfei Zhang and Ruofei Zhang
MUSIC DATA MINING
Tao Li, Mitsunori Ogihara, and George Tzanetakis
NEXT GENERATION OF DATA MINING
Hillol Kargupta, Jiawei Han, Philip S. Yu, Rajeev Motwani, and Vipin Kumar
RAPIDMINER: DATA MINING USE CASES AND BUSINESS ANALYTICS
APPLICATIONS
Markus Hofmann and Ralf Klinkenberg
RELATIONAL DATA CLUSTERING: MODELS, ALGORITHMS,
AND APPLICATIONS
Bo Long, Zhongfei Zhang, and Philip S. Yu
SERVICE-ORIENTED DISTRIBUTED KNOWLEDGE DISCOVERY
Domenico Talia and Paolo Trunfio
SPECTRAL FEATURE SELECTION FOR DATA MINING
Zheng Alan Zhao and Huan Liu
STATISTICAL DATA MINING USING SAS APPLICATIONS, SECOND EDITION
George Fernandez
SUPPORT VECTOR MACHINES: OPTIMIZATION BASED THEORY,
ALGORITHMS, AND EXTENSIONS
Naiyang Deng, Yingjie Tian, and Chunhua Zhang
TEMPORAL DATA MINING
Theophano Mitsa
TEXT MINING: CLASSIFICATION, CLUSTERING, AND APPLICATIONS
Ashok N. Srivastava and Mehran Sahami
TEXT MINING AND VISUALIZATION: CASE STUDIES USING OPEN-SOURCE
TOOLS
Markus Hofmann and Andrew Chisholm
THE TOP TEN ALGORITHMS IN DATA MINING
Xindong Wu and Vipin Kumar
UNDERSTANDING COMPLEX DATASETS: DATA MINING WITH MATRIX
DECOMPOSITIONS
David Skillicorn
DATA MINING
A Tutorial-Based Primer
SECOND EDITION

Richard J. Roiger
This book was previously published by Pearson Education, Inc.

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2017 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

Version Date: 20161025

International Standard Book Number-13: 978-1-4987-6397-4 (Pack - Book and Ebook)

This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-
ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-
lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-
ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Contents

List of Figures, xvii

List of Tables, xxix
Preface, xxxi
Acknowledgments, xxxix
Author, xli

SECTION I Data Mining Fundamentals

CHAPTER 1 ◾ Data Mining: A First View 3

CHAPTER OBJECTIVES 3
1.1 DATA SCIENCE, ANALYTICS, MINING, AND KNOWLEDGE
DISCOVERY IN DATABASES 4
1.1.1 Data Science and Analytics 4
1.1.2 Data Mining 5
1.1.3 Data Science versus Knowledge Discovery in Databases 5
1.2 WHAT CAN COMPUTERS LEARN? 6
1.2.1 Three Concept Views 6
1.2.1.1 The Classical View 6
1.2.1.2 The Probabilistic View 7
1.2.1.3 The Exemplar View 7
1.2.2 Supervised Learning 8
1.2.3 Supervised Learning: A Decision Tree Example 9
1.2.4 Unsupervised Clustering 11
1.3 IS DATA MINING APPROPRIATE FOR MY PROBLEM? 14
1.3.1 Data Mining or Data Query? 14
1.3.2 Data Mining versus Data Query: An Example 15
1.4 DATA MINING OR KNOWLEDGE ENGINEERING? 16

vii
viii ◾ Contents

1.5 A NEAREST NEIGHBOR APPROACH 18

1.6 A PROCESS MODEL FOR DATA MINING 19
1.6.1 Acquiring Data 20
1.6.1.1 The Data Warehouse 20
1.6.1.2 Relational Databases and Flat Files 21
1.6.1.3 Distributed Data Access 21
1.6.2 Data Preprocessing 21
1.6.3 Mining the Data 23
1.6.4 Interpreting the Results 23
1.6.5 Result Application 24
1.7 DATA MINING, BIG DATA, AND CLOUD COMPUTING 24
1.7.1 Hadoop 24
1.7.2 Cloud Computing 24
1.8 DATA MINING ETHICS 25
1.9 INTRINSIC VALUE AND CUSTOMER CHURN 26
1.10 CHAPTER SUMMARY 27
1.11 KEY TERMS 28

CHAPTER 2 ◾ Data Mining: A Closer Look 33

CHAPTER OBJECTIVES 33
2.1 DATA MINING STRATEGIES 34
2.1.1 Classification 34
2.1.2 Estimation 35
2.1.3 Prediction 36
2.1.4 Unsupervised Clustering 39
2.1.5 Market Basket Analysis 40
2.2 SUPERVISED DATA MINING TECHNIQUES 41
2.2.1 The Credit Card Promotion Database 41
2.2.2 Rule-Based Techniques 42
2.2.3 Neural Networks 44
2.2.4 Statistical Regression 46
2.3 ASSOCIATION RULES 47
2.4 CLUSTERING TECHNIQUES 48
2.5 EVALUATING PERFORMANCE 49
2.5.1 Evaluating Supervised Learner Models 50
2.5.2 Two-Class Error Analysis 52
 Contents ◾ ix

2.5.3 Evaluating Numeric Output 53

2.5.4 Comparing Models by Measuring Lift 53
2.5.5 Unsupervised Model Evaluation 55
2.6 CHAPTER SUMMARY 56
2.7 KEY TERMS 57

CHAPTER 3 ◾ Basic Data Mining Techniques 63

CHAPTER OBJECTIVES 63
3.1 DECISION TREES 64
3.1.1 An Algorithm for Building Decision Trees 64
3.1.2 Decision Trees for the Credit Card Promotion Database 70
3.1.3 Decision Tree Rules 73
3.1.4 Other Methods for Building Decision Trees 73
3.1.5 General Considerations 74
3.2 A BASIC COVERING RULE ALGORITHM 74
3.3 GENERATING ASSOCIATION RULES 80
3.3.1 Confidence and Support 80
3.3.2 Mining Association Rules: An Example 82
3.3.3 General Considerations 84
3.4 THE K-MEANS ALGORITHM 85
3.4.1 An Example Using K-means 86
3.4.2 General Considerations 89
3.5 GENETIC LEARNING 90
3.5.1 Genetic Algorithms and Supervised Learning 91
3.5.2 General Considerations 95
3.6 CHOOSING A DATA MINING TECHNIQUE 95
3.7 CHAPTER SUMMARY 97
3.8 KEY TERMS 98

SECTION II Tools for Knowledge Discovery

CHAPTER 4 ◾ Weka—An Environment for Knowledge Discovery 105

CHAPTER OBJECTIVES 105
4.1 GETTING STARTED WITH WEKA 106
4.2 BUILDING DECISION TREES 109
4.3 GENERATING PRODUCTION RULES WITH PART 117
x ◾ Contents

4.4 ATTRIBUTE SELECTION AND NEAREST NEIGHBOR CLASSIFICATION 122

4.5 ASSOCIATION RULES 127
4.6 COST/BENEFIT ANALYSIS, (OPTIONAL) 131
4.7 UNSUPERVISED CLUSTERING WITH THE K-MEANS ALGORITHM 137
4.8 CHAPTER SUMMARY 141

CHAPTER 5 ◾ Knowledge Discovery with RapidMiner 145

CHAPTER OBJECTIVES 145
5.1 GETTING STARTED WITH RAPIDMINER 146
5.1.1 Installing RapidMiner 146
5.1.2 Navigating the Interface 146
5.1.3 A First Process Model 149
5.1.4 A Decision Tree for the Credit Card Promotion Database 156
5.1.5 Breakpoints 158
5.2 BUILDING DECISION TREES 159
5.2.1 Scenario 1: Using a Training and Test Set 160
5.2.2 Scenario 2: Adding a Subprocess 165
5.2.3 Scenario 3: Creating, Saving, and Applying the Final Model 167
5.2.3.1 Saving a Model to an Output File 167
5.2.3.2 Reading and Applying a Model 168
5.2.4 Scenario 4: Using Cross-Validation 168
5.3 GENERATING RULES 173
5.3.1 Scenario 1: Tree to Rules 173
5.3.2 Scenario 2: Rule Induction 176
5.3.3 Scenario 3: Subgroup Discovery 178
5.4 ASSOCIATION RULE LEARNING 181
5.4.1 Association Rules for the Credit Card Promotion Database 182
5.4.2 The Market Basket Analysis Template 183
5.5 UNSUPERVISED CLUSTERING WITH K-MEANS 187
5.6 ATTRIBUTE SELECTION AND NEAREST NEIGHBOR CLASSIFICATION 191
5.7 CHAPTER SUMMARY 194

CHAPTER 6 ◾ The Knowledge Discovery Process 199

CHAPTER OBJECTIVES 199
6.1 A PROCESS MODEL FOR KNOWLEDGE DISCOVERY 199
6.2 GOAL IDENTIFICATION 201
 Contents ◾ xi

6.3 CREATING A TARGET DATA SET 202

6.4 DATA PREPROCESSING 203
6.4.1 Noisy Data 203
6.4.1.1 Locating Duplicate Records 204
6.4.1.2 Locating Incorrect Attribute Values 204
6.4.1.3 Data Smoothing 204
6.4.1.4 Detecting Outliers 205
6.4.2 Missing Data 207
6.5 DATA TRANSFORMATION 208
6.5.1 Data Normalization 208
6.5.2 Data Type Conversion 209
6.5.3 Attribute and Instance Selection 209
6.5.3.1 Wrapper and Filtering Techniques 210
6.5.3.2 More Attribute Selection Techniques 211
6.5.3.3 Genetic Learning for Attribute Selection 211
6.5.3.4 Creating Attributes 212
6.5.3.5 Instance Selection 213
6.6 DATA MINING 214
6.7 INTERPRETATION AND EVALUATION 214
6.8 TAKING ACTION 215
6.9 THE CRISP-DM PROCESS MODEL 215
6.10 CHAPTER SUMMARY 216
6.11 KEY TERMS 216

CHAPTER 7 ◾ Formal Evaluation Techniques 221

CHAPTER OBJECTIVES 221
7.1 WHAT SHOULD BE EVALUATED? 222
7.2 TOOLS FOR EVALUATION 223
7.2.1 Single-Valued Summary Statistics 224
7.2.2 The Normal Distribution 225
7.2.3 Normal Distributions and Sample Means 226
7.2.4 A Classical Model for Hypothesis Testing 228
7.3 COMPUTING TEST SET CONFIDENCE INTERVALS 230
7.4 COMPARING SUPERVISED LEARNER MODELS 232
7.4.1 Comparing the Performance of Two Models 233
7.4.2 Comparing the Performance of Two or More Models 234
xii ◾ Contents

7.5 UNSUPERVISED EVALUATION TECHNIQUES 235

7.5.1 Unsupervised Clustering for Supervised Evaluation 235
7.5.2 Supervised Evaluation for Unsupervised Clustering 235
7.5.3 Additional Methods for Evaluating an Unsupervised Clustering 236
7.6 EVALUATING SUPERVISED MODELS WITH NUMERIC OUTPUT 236
7.7 COMPARING MODELS WITH RAPIDMINER 238
7.8 ATTRIBUTE EVALUATION FOR MIXED DATA TYPES 241
7.9 PARETO LIFT CHARTS 244
7.10 CHAPTER SUMMARY 247
7.11 KEY TERMS 248

SECTION III Building Neural Networks

CHAPTER 8 ◾ Neural Networks 253

CHAPTER OBJECTIVES 253
8.1 FEED-FORWARD NEURAL NETWORKS 254
8.1.1 Neural Network Input Format 254
8.1.2 Neural Network Output Format 255
8.1.3 The Sigmoid Evaluation Function 256
8.2 NEURAL NETWORK TRAINING: A CONCEPTUAL VIEW 258
8.2.1 Supervised Learning with Feed-Forward Networks 258
8.2.1.1 Training a Neural Network: Backpropagation Learning 258
8.2.1.2 Training a Neural Network: Genetic Learning 259
8.2.2 Unsupervised Clustering with Self-Organizing Maps 259
8.3 NEURAL NETWORK EXPLANATION 260
8.4 GENERAL CONSIDERATIONS 262
8.5 NEURAL NETWORK TRAINING: A DETAILED VIEW 263
8.5.1 The Backpropagation Algorithm: An Example 263
8.5.2 Kohonen Self-Organizing Maps: An Example 266
8.6 CHAPTER SUMMARY 268
8.7 KEY TERMS 269

CHAPTER 9 ◾ Building Neural Networks with Weka 271

CHAPTER OBJECTIVES 271
9.1 DATA SETS FOR BACKPROPAGATION LEARNING 272
9.1.1 The Exclusive-OR Function 272
9.1.2 The Satellite Image Data Set 273
 Contents ◾ xiii

9.2 MODELING THE EXCLUSIVE-OR FUNCTION: NUMERIC OUTPUT 274

9.3 MODELING THE EXCLUSIVE-OR FUNCTION: CATEGORICAL OUTPUT 280
9.4 MINING SATELLITE IMAGE DATA 282
9.5 UNSUPERVISED NEURAL NET CLUSTERING 287
9.6 CHAPTER SUMMARY 288
9.7 KEY TERMS 289

CHAPTER 10 ◾ Building Neural Networks with RapidMiner 293

CHAPTER OBJECTIVES 293
10.1 MODELING THE EXCLUSIVE-OR FUNCTION 294
10.2 MINING SATELLITE IMAGE DATA 301
10.3 PREDICTING CUSTOMER CHURN 306
10.4 RAPIDMINER’S SELF-ORGANIZING MAP OPERATOR 311
10.5 CHAPTER SUMMARY 313

SECTION IV Advanced Data Mining Techniques

CHAPTER 11 ◾ Supervised Statistical Techniques 317

CHAPTER OBJECTIVES 317
11.1 NAÏVE BAYES CLASSIFIER 317
11.1.1 Naïve Bayes Classifier: An Example 318
11.1.2 Zero-Valued Attribute Counts 321
11.1.3 Missing Data 321
11.1.4 Numeric Data 322
11.1.5 Implementations of the Naïve Bayes Classifier 324
11.1.6 General Considerations 324
11.2 SUPPORT VECTOR MACHINES 324
11.2.1 Linearly Separable Classes 332
11.2.2 The Nonlinear Case 336
11.2.3 General Considerations 337
11.2.4 Implementations of Support Vector Machines 340
11.3 LINEAR REGRESSION ANALYSIS 340
11.3.1 Simple Linear Regression 344
11.3.2 Multiple Linear Regression 344
11.3.2.1 Linear Regression—Weka 344
11.3.2.2 Linear Regression—RapidMiner 345
xiv ◾ Contents

11.4 REGRESSION TREES 349

11.5 LOGISTIC REGRESSION 350
11.5.1 Transforming the Linear Regression Model 350
11.5.2 The Logistic Regression Model 351
11.6 CHAPTER SUMMARY 352
11.7 KEY TERMS 352

CHAPTER 12 ◾ Unsupervised Clustering Techniques 357

CHAPTER OBJECTIVES 357
12.1 AGGLOMERATIVE CLUSTERING 358
12.1.1 Agglomerative Clustering: An Example 358
12.1.2 General Considerations 360
12.2 CONCEPTUAL CLUSTERING 360
12.2.1 Measuring Category Utility 361
12.2.2 Conceptual Clustering: An Example 362
12.2.3 General Considerations 364
12.3 EXPECTATION MAXIMIZATION 364
12.3.1 Implementations of the EM Algorithm 365
12.3.2 General Considerations 365
12.4 GENETIC ALGORITHMS AND UNSUPERVISED CLUSTERING 371
12.5 CHAPTER SUMMARY 374
12.6 KEY TERMS 374

CHAPTER 13 ◾ Specialized Techniques 377

CHAPTER OBJECTIVES 377
13.1 TIME-SERIES ANALYSIS 377
13.1.1 Stock Market Analytics 378
13.1.2 Time-Series Analysis—An Example 379
13.1.2.1 Creating the Target Data Set—Numeric Output 380
13.1.2.2 Data Preprocessing and Transformation 380
13.1.2.3 Creating the Target Data Set—Categorical Output 382
13.1.2.4 Mining the Data—RapidMiner 382
13.1.2.5 Mining the Data—Weka 387
13.1.2.6 Interpretation, Evaluation, and Action 390
13.1.3 General Considerations 390
 Contents ◾ xv

13.2 MINING THE WEB 391

13.2.1 Web-Based Mining: General Issues 391
13.2.1.1 Identifying the Goal 391
13.2.2 Preparing the Data 392
13.2.2.1 Mining the Data 393
13.2.2.2 Interpreting and Evaluating Results 393
13.2.2.3 Taking Action 394
13.2.3 Data Mining for Website Evaluation 395
13.2.4 Data Mining for Personalization 395
13.2.5 Data Mining for Website Adaptation 396
13.2.6 PageRank and Link Analysis 396
13.2.7 Operators for Web-Based Mining 398
13.3 MINING TEXTUAL DATA 398
13.3.1 Analyzing Customer Reviews 399
13.4 TECHNIQUES FOR LARGE-SIZED, IMBALANCED,
AND STREAMING DATA 404
13.4.1 Large-Sized Data 404
13.4.2 Dealing with Imbalanced Data 405
13.4.2.1 Methods for Addressing Rarity 406
13.4.2.2 Receiver Operating Characteristics Curves 406
13.4.3 Methods for Streaming Data 412
13.5 ENSEMBLE TECHNIQUES FOR IMPROVING PERFORMANCE 413
13.5.1 Bagging 413
13.5.2 Boosting 414
13.5.3 AdaBoost—An Example 414
13.6 CHAPTER SUMMARY 417
13.7 KEY TERMS 418

CHAPTER 14 ◾ The Data Warehouse 423

CHAPTER OBJECTIVES 423
14.1 OPERATIONAL DATABASES 424
14.1.1 Data Modeling and Normalization 424
14.1.2 The Relational Model 425
14.2 DATA WAREHOUSE DESIGN 426
14.2.1 Entering Data into the Warehouse 427
xvi ◾ Contents

14.2.2 Structuring the Data Warehouse: The Star Schema 429

14.2.2.1 The Multidimensionality of the Star Schema 430
14.2.2.2 Additional Relational Schemas 431
14.2.3 Decision Support: Analyzing the Warehouse Data 432
14.3 ONLINE ANALYTICAL PROCESSING 434
14.3.1 OLAP: An Example 435
14.3.2 General Considerations 438
14.4 EXCEL PIVOT TABLES FOR DATA ANALYTICS 438
14.5 CHAPTER SUMMARY 445
14.6 KEY TERMS 446

APPENDIX A—SOFTWARE AND DATA SETS FOR DATA MINING, 451

APPENDIX B—STATISTICS FOR PERFORMANCE EVALUATION, 455

BIBLIOGRAPHY, 461

INDEX, 465
List of Figures

Figure 1.1 A decision tree for the data in Table 1.1. 10

Figure 1.2 Data mining versus expert systems. 17
Figure 1.3 A process model for data mining. 20
Figure 1.4 A perfect positive correlation (r = 1). 22
Figure 1.5 A perfect negative correlation (r = −1). 23
Figure 1.6 Intrinsic versus actual customer value. 26
Figure 2.1 A hierarchy of data mining strategies. 34
Figure 2.2 A fully connected multilayer neural network. 45
Figure 2.3 An unsupervised clustering of the credit card database. 49
Figure 2.4 Targeted versus mass mailing. 54
Figure 3.1 A partial decision tree with root node = income range. 68
Figure 3.2 A partial decision tree with root node = credit card insurance. 69
Figure 3.3 A partial decision tree with root node = age. 70
Figure 3.4 A three-node decision tree for the credit card database. 71
Figure 3.5 A two-node decision tree for the credit card database. 71
Figure 3.6 Domain statistics for the credit card promotion database. 76
Figure 3.7 Class statistics for life insurance promotion = yes. 77
Figure 3.8 Class statistics for life insurance promotion = no. 78
Figure 3.9 Statistics for life insurance promotion = yes after removing five instances. 79
Figure 3.10 A coordinate mapping of the data in Table 3.6. 86
Figure 3.11 A K-means clustering of the data in Table 3.6 (K = 2). 89
Figure 3.12 Supervised genetic learning. 91

xvii
xviii ◾ List of Figures

Figure 3.13 A crossover operation. 94

Figure 4.1 Weka GUI Chooser. 107
Figure 4.2 Explorer four graphical user interfaces (GUI’s). 107
Figure 4.3 Weka install folder. 108
Figure 4.4 Sample data sets. 108
Figure 4.5 Instances of the contact-lenses file. 109
Figure 4.6 Loading the contact-lenses data set. 110
Figure 4.7 Navigating the Explorer interface. 111
Figure 4.8 A partial list of attribute filters. 112
Figure 4.9 Command line call for J48. 112
Figure 4.10 Parameter setting options for J48. 113
Figure 4.11 Decision tree for the contact-lenses data set. 113
Figure 4.12 Weka’s tree visualizer. 114
Figure 4.13 Decision tree output for the contact-lenses data set. 115
Figure 4.14 Classifier output options. 116
Figure 4.15 Actual and predicted output. 116
Figure 4.16 Customer churn data. 118
Figure 4.17 A decision list for customer churn data. 118
Figure 4.18 Customer churn output generated by PART. 119
Figure 4.19 Loading the customer churn instances of unknown outcome. 120
Figure 4.20 Predicting customers likely to churn. 121
Figure 4.21 Nearest neighbor output for the spam data set. 123
Figure 4.22 Weka’s attribute selection filter. 124
Figure 4.23 Options for the attribute selection filter. 124
Figure 4.24 Parameter settings for ranker. 125
Figure 4.25 Most predictive attributes for the spam data set. 126
Figure 4.26 IBk output after removing the 10 least predictive attributes. 126
Figure 4.27 Association rules for the contact-lenses data set. 128
Figure 4.28 Parameters for the Apriori algorithm. 129
 List of Figures ◾ xix

Figure 4.29 The supermarket data set. 130

Figure 4.30 Instances of the supermarket data set. 130
Figure 4.31 Ten association rules for the supermarket data set. 131
Figure 4.32 A J48 classification of the credit card screening data set. 132
Figure 4.33 Invoking a cost/benefit analysis. 133
Figure 4.34 Cost/benefit output for the credit card screening data set. 133
Figure 4.35 Cost/benefit analysis set to match J48 classifier output. 134
Figure 4.36 Invoking a cost/benefit analysis. 135
Figure 4.37 Minimizing total cost. 135
Figure 4.38 Cutoff scores for credit card application acceptance. 136
Figure 4.39 Classes to clusters evaluation for simpleKmeans. 137
Figure 4.40 Include standard deviation values for simpleKmeans. 138
Figure 4.41 Classes to clusters output. 139
Figure 4.42 Partial list of attribute values for the K-means clustering in Figure 4.41. 140
Figure 4.43 Additional attribute values for the SimpleKMeans clustering in Figure 4.41. 140
Figure 5.1 An introduction to RapidMiner. 147
Figure 5.2 Creating a new blank process. 147
Figure 5.3 A new blank process with helpful pointers. 148
Figure 5.4 Creating and saving a process. 150
Figure 5.5 Importing the credit card promotion database. 150
Figure 5.6 Selecting the cells to import. 151
Figure 5.7 A list of allowable data types. 152
Figure 5.8 Changing the role of Life Ins Promo. 152
Figure 5.9 Storing a file in the data folder. 153
Figure 5.10 The credit card promotion database. 153
Figure 5.11 A successful file import. 154
Figure 5.12 Connecting the credit card promotion database to an output port. 154
Figure 5.13 Summary statistics for the credit card promotion database. 155
Figure 5.14 A bar graph for income range. 155
xx ◾ List of Figures

Figure 5.15 A scatterplot comparing age and life insurance promotion. 156
Figure 5.16 A decision tree process model. 157
Figure 5.17 A decision tree for the credit card promotion database. 158
Figure 5.18 A decision tree in descriptive form. 158
Figure 5.19 A list of operator options. 159
Figure 5.20 Customer churn—A training and test set scenario. 160
Figure 5.21 Removing instances of unknown outcome from the churn data set. 161
Figure 5.22 Partitioning the customer churn data. 162
Figure 5.23 The customer churn data set. 163
Figure 5.24 Filter Examples has removed all instances of unknown outcome. 163
Figure 5.25 A decision tree for the customer churn data set. 164
Figure 5.26 Output of the Apply Model operator. 164
Figure 5.27 A performance vector for the customer churn data set. 165
Figure 5.28 Adding a subprocess to the main process window. 166
Figure 5.29 A subprocess for data preprocessing. 167
Figure 5.30 Creating and saving a decision tree model. 168
Figure 5.31 Reading and applying a saved model. 169
Figure 5.32 An Excel file stores model predictions. 169
Figure 5.33 Testing a model using cross-validation. 170
Figure 5.34 A subprocess to read and filter customer churn data. 171
Figure 5.35 Nested subprocesses for cross-validation. 171
Figure 5.36 Performance vector for a decision tree tested using cross-validation. 172
Figure 5.37 Subprocess for the Tree to Rules operator. 174
Figure 5.38 Building a model with the Tree to Rules operator. 174
Figure 5.39 Rules generated by the Tree to Rules operator. 175
Figure 5.40 Performance vector for the customer churn data set. 175
Figure 5.41 A process design for rule induction. 176
Figure 5.42 Adding the Discretize by Binning operator. 177
Figure 5.43 Covering rules for customer churn data. 177
 List of Figures ◾ xxi

Figure 5.44 Performance vector for the covering rules of Figure 5.43. 178
Figure 5.45 Process design for subgroup discovery. 179
Figure 5.46 Subprocess design for subgroup discovery. 179
Figure 5.47 Rules generated by the Subgroup Discovery operator. 180
Figure 5.48 Ten rules identifying likely churn candidates. 181
Figure 5.49 Generating association rules for the credit card promotion database. 182
Figure 5.50 Preparing data for association rule generation. 183
Figure 5.51 Interface for listing association rules. 184
Figure 5.52 Association rules for the credit card promotion database. 184
Figure 5.53 Market basket analysis template. 185
Figure 5.54 The pivot operator rotates the example set. 186
Figure 5.55 Association rules for the Market Basket Analysis template. 186
Figure 5.56 Process design for clustering gamma-ray burst data. 188
Figure 5.57 A partial clustering of gamma-ray burst data. 189
Figure 5.58 Three clusters of gamma-ray burst data. 189
Figure 5.59 Decision tree illustrating a gamma-ray burst clustering. 190
Figure 5.60 A descriptive form of a decision tree showing a clustering
of gamma-ray burst data. 190
Figure 5.61 Benchmark performance for nearest neighbor classification. 192
Figure 5.62 Main process design for nearest neighbor classification. 192
Figure 5.63 Subprocess for nearest neighbor classification. 193
Figure 5.64 Forward selection subprocess for nearest neighbor classification. 193
Figure 5.65 Performance vector when forward selection is used for choosing
attributes. 194
Figure 5.66 Unsupervised clustering for attribute evaluation. 197
Figure 6.1 A seven-step KDD process model. 200
Figure 6.2 The Acme credit card database. 203
Figure 6.3 A process model for detecting outliers. 205
Figure 6.4 Two outlier instances from the diabetes patient data set. 206
xxii ◾ List of Figures

Figure 6.5 Ten outlier instances from the diabetes patient data set. 207
Figure 7.1 Components for supervised learning. 222
Figure 7.2 A normal distribution. 225
Figure 7.3 Random samples from a population of 10 elements. 226
Figure 7.4 A process model for comparing three competing models. 239
Figure 7.5 Subprocess for comparing three competing models. 240
Figure 7.6 Cross-validation test for a decision tree with maximum depth = 5. 240
Figure 7.7 A matrix of t-test scores. 241
Figure 7.8 ANOVA comparing three competing models. 241
Figure 7.9 ANOVA operators for comparing nominal and numeric attributes. 242
Figure 7.10 The grouped ANOVA operator comparing class and maximum heart
rate. 243
Figure 7.11 The ANOVA matrix operator for the cardiology patient data set. 243
Figure 7.12 A process model for creating a lift chart. 244
Figure 7.13 Preprocessing the customer churn data set. 245
Figure 7.14 Output of the Apply Model operator for the customer churn data set. 245
Figure 7.15 Performance vector for customer churn. 246
Figure 7.16 A Pareto lift chart for customer churn. 247
Figure 8.1 A fully connected feed-forward neural network. 254
Figure 8.2 The sigmoid evaluation function. 257
Figure 8.3 A 3 × 3 Kohonen network with two input-layer nodes. 260
Figure 8.4 Connections for two output-layer nodes. 266
Figure 9.1 Graph of the XOR function. 272
Figure 9.2 XOR training data. 273
Figure 9.3 Satellite image data. 274
Figure 9.4 Weka four graphical user interfaces (GUIs) for XOR training. 275
Figure 9.5 Backpropagation learning parameters. 276
Figure 9.6 Architecture for the XOR function. 278
Figure 9.7 XOR training output. 278
 List of Figures ◾ xxiii

Figure 9.8 Network architecture with associated connection weights. 279

Figure 9.9 XOR network architecture without a hidden layer. 280
Figure 9.10 Confusion matrix for XOR without a hidden layer. 281
Figure 9.11 XOR with hidden layer and categorical output. 281
Figure 9.12 XOR confusion matrix and categorical output. 282
Figure 9.13 Satellite image data network architecture. 284
Figure 9.14 Confusion matrix for satellite image data. 284
Figure 9.15 Updated class assignment for instances 78 through 94 of the satellite
image data set. 286
Figure 9.16 Initial classification for pixel instances 78 through 94 of the satellite
image data set. 286
Figure 9.17 Parameter settings for Weka’s SelfOrganizingMap. 287
Figure 9.18 Applying Weka’s SelfOrganizingMap to the diabetes data set. 288
Figure 10.1 Statistics for the XOR function. 295
Figure 10.2 Main process for learning the XOR function. 295
Figure 10.3 Default settings for the hidden layers parameter. 296
Figure 10.4 A single hidden layer of five nodes. 297
Figure 10.5 Performance parameters for neural network learning. 298
Figure 10.6 Neural network architecture for the XOR function. 298
Figure 10.7 Hidden-to-output layer connection weights. 299
Figure 10.8 Prediction confidence values for the XOR function. 299
Figure 10.9 Performance vector for the XOR function. 300
Figure 10.10 Absolute error value for the XOR function. 300
Figure 10.11 Attribute declarations for the satellite image data set. 302
Figure 10.12 Main process for mining the satellite image data set. 302
Figure 10.13 Subprocesses for satellite image data set. 303
Figure 10.14 Network architecture for the satellite image data set. 304
Figure 10.15 Performance vector for the satellite image data set. 304
Figure 10.16 Removing correlated attributes from the satellite image data set. 305
xxiv ◾ List of Figures

Figure 10.17 Green and red have been removed from the satellite image data set. 305
Figure 10.18 Correlation matrix for the satellite image data set. 306
Figure 10.19 Neural network model for predicting customer churn. 307
Figure 10.20 Preprocessing the customer churn data. 308
Figure 10.21 Cross-validation subprocess for customer churn. 308
Figure 10.22 Performance vector for customer churn. 309
Figure 10.23 Process for creating and saving a neural network model. 309
Figure 10.24 Process model for reading and applying a neural network model. 310
Figure 10.25 Neural network output for predicting customer churn. 310
Figure 10.26 SOM process model for the cardiology patient data set. 312
Figure 10.27 Clustered instances of the cardiology patient data set. 312
Figure 11.1 RapidMiner’s naïve Bayes operator. 325
Figure 11.2 Subprocess for applying naïve Bayes to customer churn data. 326
Figure 11.3 Naïve Bayes Distribution Table for customer churn data. 326
Figure 11.4 Naïve Bayes performance vector for customer churn data. 327
Figure 11.5 Life insurance promotion by gender. 328
Figure 11.6 Naïve Bayes model with output attribute = LifeInsPromo. 329
Figure 11.7 Predictions for the life insurance promotion. 329
Figure 11.8 Hyperplanes separating the circle and star classes. 330
Figure 11.9 Hyperplanes passing through their respective support vectors. 331
Figure 11.10 Maximal margin hyperplane separating the star and circle classes. 335
Figure 11.11 Loading the nine instances of Figure 11.8 into the Explorer. 338
Figure 11.12 Invoking SMO model. 339
Figure 11.13 Disabling data normalization/standardization. 339
Figure 11.14 The SMO-created MMH for the data shown in Figure 11.8. 340
Figure 11.15 Applying mySVM to the cardiology patient data set. 341
Figure 11.16 Normalized cardiology patient data. 342
Figure 11.17 Equation of the MMH for the cardiology patient data set. 342
Figure 11.18 Actual and predicted output for the cardiology patient data. 343
 List of Figures ◾ xxv

Figure 11.19 Performance vector for the cardiology patient data. 343
Figure 11.20 A linear regression model for the instances of Figure 11.8. 345
Figure 11.21 Main process window for applying RapidMiner’s linear regression
operator to the gamma-ray burst data set. 346
Figure 11.22 Subprocess windows for the Gamma Ray burst experiment. 346
Figure 11.23 Linear regression—actual and predicted output for the gamma-ray
burst data set. 347
Figure 11.24 Summary statistics and the linear regression equation for the
gamma-ray burst data set. 347
Figure 11.25 Scatterplot diagram showing the relationship between t90 and t50. 348
Figure 11.26 Performance vector resulting from the application of linear
regression to the gamma-ray burst data set. 348
Figure 11.27 A generic model tree. 349
Figure 11.28 The logistic regression equation. 351
Figure 12.1 A Cobweb-created hierarchy. 363
Figure 12.2 Applying EM to the gamma-ray burst data set. 366
Figure 12.3 Removing correlated attributes from the gamma-ray burst data set. 367
Figure 12.4 An EM clustering of the gamma-ray burst data set. 367
Figure 12.5 Summary statistics for an EM clustering of the gamma-ray burst data set. 368
Figure 12.6 Decision tree representing a clustering of the gamma-ray burst data set. 368
Figure 12.7 The decision tree of Figure 12.6 in descriptive form. 369
Figure 12.8 Classes of the sensor data set. 370
Figure 12.9 Generic object editor allows us to specify the number of clusters. 370
Figure 12.10 Classes to clusters summary statistics. 371
Figure 12.11 Unsupervised genetic clustering. 372
Figure 13.1 A process model for extracting historical market data. 380
Figure 13.2 Historical data for XIV. 381
Figure 13.3 Time-series data with numeric output. 382
Figure 13.4 Time-series data with categorical output. 383
Figure 13.5 Time-series data for processing with RapidMiner. 383
xxvi ◾ List of Figures

Figure 13.6 A 3-month price chart for XIV. 384

Figure 13.7 A process model for time-series analysis with categorical output. 385
Figure 13.8 Predictions and confidence scores for time-series analysis. 385
Figure 13.9 Performance vector—time-series analysis for XIV. 386
Figure 13.10 Predicting the next-day closing price of XIV. 386
Figure 13.11 Time-series data formatted for Weka—categorical output. 387
Figure 13.12 Time-series data formatted for Weka—numeric output. 388
Figure 13.13 Time-series analysis with categorical output. 388
Figure 13.14 Time-series analysis with numeric output. 389
Figure 13.15 Cluster analysis using time-series data. 389
Figure 13.16 A generic Web usage data mining model. 392
Figure 13.17 Creating usage profiles from session data. 395
Figure 13.18 Hypertext link recommendations from usage profiles. 396
Figure 13.19 A page link structure. 397
Figure 13.20 A main process model for mining textual data. 401
Figure 13.21 A template to enter folder names used for textual data. 401
Figure 13.22 Class and folder names containing textual data. 402
Figure 13.23 Subprocess for tokenizing and stemming textual data. 402
Figure 13.24 A tokenized positive evaluation. 403
Figure 13.25 A tokenized and stemmed positive evaluation. 403
Figure 13.26 Textual data reduced to two dimensions. 403
Figure 13.27 Rules defining the three product evaluation classes. 404
Figure 13.28 An ROC graph for four competing models. 407
Figure 13.29 The PART algorithm applied to the spam data set. 408
Figure 13.30 An ROC curve created by applying PART to the spam data set. 409
Figure 13.31 Locating the true- and false-positive rate position in the ROC curve. 410
Figure 13.32 Confidence scores for predicted values with the spam data set. 410
Figure 13.33 Sorted confidence scores for the spam data set. 411
Figure 13.34 A main process for testing the AdaBoost operator. 415
 List of Figures ◾ xxvii

Figure 13.35 Subprocess using a decision tree without AdaBoost. 415

Figure 13.36 Subprocess using AdaBoost, which builds several decision trees. 416
Figure 13.37 T-test results for testing AdaBoost. 416
Figure 13.38 Results of the ANOVA with the AdaBoost operator. 417
Figure 14.1 A simple entity-relationship diagram. 424
Figure 14.2 A data warehouse process model. 428
Figure 14.3 A star schema for credit card purchases. 429
Figure 14.4 Dimensions of the fact table shown in Figure 14.3. 431
Figure 14.5 A constellation schema for credit card purchases and promotions. 433
Figure 14.6 A multidimensional cube for credit card purchases. 435
Figure 14.7 A concept hierarchy for location. 436
Figure 14.8 Rolling up from months to quarters. 437
Figure 14.9 Creating a pivot table. 439
Figure 14.10 A blank pivot table. 440
Figure 14.11 A comparison of credit card insurance and income range. 440
Figure 14.12 A chart comparing credit card insurance and income range. 441
Figure 14.13 A credit card promotion cube. 442
Figure 14.14 A pivot table for the cube shown in Figure 14.13. 442
Figure 14.15 Pivot table position corresponding to the highlighted cell
in Figure 14.13. 443
Figure 14.16 Drilling down into the cell highlighted in Figure 14.15. 444
Figure 14.17 Highlighting female customers with a slice operation. 444
Figure 14.18 A second approach for highlighting female customers. 445
Figure A.1 A successful installation. 452
Figure A.2 Locating and installing a package. 452
Figure A.3 List of installed packages. 452
List of Tables

Table 1.1 Hypothetical training data for disease diagnosis 9

Table 1.2 Data instances with an unknown classification 10
Table 1.3 Acme Investors Incorporated 12
Table 2.1 Cardiology patient data 37
Table 2.2 Typical and atypical instances from the cardiology domain 38
Table 2.3 Credit card promotion database 41
Table 2.4 Neural network training: actual and computed output 46
Table 2.5 A three-class confusion matrix 51
Table 2.6 A simple confusion matrix 52
Table 2.7 Two confusion matrices each showing a 10% error rate 52
Table 2.8 Two confusion matrices: no model and an ideal model 54
Table 2.9 Two confusion matrices for alternative models with lift equal to 2.25 55
Table 3.1 The credit card promotion database 67
Table 3.2 Training data instances following the path in Figure 3.4 to credit card
insurance = no 72
Table 3.3 A subset of the credit card promotion database 82
Table 3.4 Single-item sets 83
Table 3.5 Two-item sets 83
Table 3.6 K-means input values 86
Table 3.7 Several applications of the K-means algorithm (K = 2) 89
Table 3.8 An initial population for supervised genetic learning 92
Table 3.9 Training data for genetic learning 93

xxix
xxx ◾ List of Tables

Table 3.10 A second-generation population 94

Table 6.1 Initial population for genetic attribute selection 212
Table 7.1 A confusion matrix for the null hypothesis 229
Table 7.2 Absolute and squared error (output attribute = life insurance promotion) 237
Table 8.1 Initial weight values for the neural network shown in Figure 8.1 257
Table 8.2 A population of weight elements for the network in Figure 8.1 259
Table 9.1 Exclusive-OR function 272
Table 11.1 Data for Bayes classifier 318
Table 11.2 Counts and probabilities for attribute gender 318
Table 11.3 Addition of attribute age to the Bayes classifier data set 323
Table 12.1 Five instances from the credit card promotion database 358
Table 12.2 Agglomerative clustering: first iteration 358
Table 12.3 Agglomerative clustering: second iteration 359
Table 12.4 Data for conceptual clustering 364
Table 12.5 Instances for unsupervised genetic learning 373
Table 12.6 A first-generation population for unsupervised clustering 373
Table 13.1 Daily closing price for XIV, January 4, 2016 to January 8, 2016 381
Table 13.2 Daily closing price for XIV January 5, 2016 to January 11, 2016 381
Table 14.1 Relational table for vehicle-type 425
Table 14.2 Relational table for customer 426
Table 14.3 Join of Tables 14.1 and 14.2 426
Preface

Data mining is the process of finding interesting patterns in data. The objective of data
mining is to use discovered patterns to help explain current behavior or to predict future
outcomes. Several aspects of the data mining process can be studied. These include

• Data gathering and storage

• Data selection and preparation
• Model building and testing
• Interpreting and validating results
• Model application

A single book cannot concentrate on all areas of the data mining process. Although
we furnish some detail about all aspects of data mining and knowledge discovery, our
primary focus is centered on model building and testing, as well as on interpreting and
validating results.

OBJECTIVES
I wrote the text to facilitate the following student learning goals:

• Understand what data mining is and how data mining can be employed to solve real
problems
• Recognize whether a data mining solution is a feasible alternative for a specific
problem
• Step through the knowledge discovery process and write a report about the results of
a data mining session
• Know how to apply data mining software tools to solve real problems
• Apply basic statistical and nonstatistical techniques to evaluate the results of a data
mining session

xxxi
xxxii ◾ Preface

• Recognize several data mining strategies and know when each strategy is appropriate
• Develop a comprehensive understanding of how several data mining techniques
build models to solve problems
• Develop a general awareness about the structure of a data warehouse and how a data
warehouse can be used
• Understand what online analytical processing (OLAP) is and how it can be applied
to analyze data

UPDATED CONTENT AND SOFTWARE CHANGES

The most obvious difference between the first and second editions of the text is the change
in data mining software. Here is a short list of the major changes seen with the second
edition:

• In Chapter 4, I introduce the Waikato Environment for Knowledge Analysis (Weka),

an easy-to-use, publicly available tool for data mining. Weka contains a wealth of
preprocessing and data mining techniques, graphical features, and visualization
capabilities.

• Chapter 5 is all about data mining using RapidMiner Studio, a powerful open-source
and code-free version of RapidMiner’s commercial product. RapidMiner uses a
drag and drop workflow paradigm for building models to solve complex problems.
RapidMiner’s intuitive user interface, visualization capabilities, and assortment of
operators for preprocessing and mining data are second to none.

• This edition covers what are considered to be the top 10 data mining algorithms
(Wu and Kumar, 2009). Nine of the algorithms are used in one or more tutorials.

• Tutorials have been added for attribute selection, dealing with imbalanced data, out-
lier analysis, time-series analysis, and mining textual data.

• Over 90% of the tutorials are presented using both Weka and RapidMiner. This
allows readers maximum flexibility for their hands-on data mining experience.

• Selected new topics include

• A brief introduction to big data and data analytics

• Receiver operating characteristic (ROC) curves

• Methods for handling large-sized, streaming, and imbalanced data

• Extended coverage of textual data mining

• Added techniques for attribute and outlier analysis

 Preface ◾ xxxiii

DATA SETS FOR DATA MINING

All data sets used for tutorials, illustrations, and end-of-chapter exercises are described in
the text. The data sets come from several areas including business, health and medicine, and
science. The data sets together with screenshots in PowerPoint and PDF format showing
what you will see as you work through the tutorials can be downloaded from two locations:

• The CRC website: https://www.crcpress.com/Data-Mining-A-Tutorial-Based-Primer

-Second-Edition/Roiger/p/book/9781498763974, under the Downloads tab
• https://krypton.mnsu.edu/~sa7379bt

INTENDED AUDIENCE
I developed most of the material for this book while teaching a one-semester data mining
course open to students majoring or minoring in business or computer science. In writing
this text, I directed my attention toward four groups of individuals:

• Educators in the areas of decision science, computer science, information systems,

and information technology who wish to teach a unit, workshop, or entire course on
data mining and knowledge discovery
• Students who want to learn about data mining and desire hands-on experience with
a data mining tool
• Business professionals who need to understand how data mining can be applied to
help solve their business problems
• Applied researchers who wish to add data mining methods to their problem-solving
and analytics tool kit

CHAPTER FEATURES
I take the approach that model building is both an art and a science best understood from
the perspective of learning by doing. My view is supported by several features found within
the pages of the text. The following is a partial list of these features.

• Simple, detailed examples. I remove much of the mystery surrounding data mining
by presenting simple, detailed examples of how the various data mining techniques
build their models. Because of its tutorial nature, the text is appropriate as a self-study
guide as well as a college-level textbook for a course about data mining and knowl-
edge discovery.
• Overall tutorial style. All examples in Chapters 4, 5, 9, and 10 are tutorials. Selected
sections in Chapters 6, 7, 11, 12, 13, and 14 offer easy-to-follow, step-by-step tutorials
xxxiv ◾ Preface

for performing data analytics. All selected section tutorials are highlighted for easy
differentiation from regular text.

• Data sets for data mining. A variety of data sets from business, medicine, and science
are ready for data mining.

• Key term definitions. Each chapter introduces several key terms. A list of definitions
for these terms is provided at the end of each chapter.

• End-of-chapter exercises. The end-of-chapter exercises reinforce the techniques

and concepts found within each chapter. The exercises are grouped into one of
three categories—review questions, data mining questions, and computational
questions.

• Review questions ask basic questions about the concepts and content found
within each chapter. The questions are designed to help determine if the reader
understands the major points conveyed in each chapter.

• Data mining questions require the reader to use one or several data mining tools
to perform data mining sessions.

• Computational questions have a mathematical flavor in that they require the

reader to perform one or several calculations. Many of the computational ques-
tions are appropriate for challenging the more advanced student.

CHAPTER CONTENT
The ordering of the chapters and the division of the book into separate parts is based
on several years of experience in teaching courses on data mining. Section I introduces
material that is fundamental to understanding the data mining process. The presenta-
tion is informal and easy to follow. Basic data mining concepts, strategies, and tech-
niques are introduced. Students learn about the types of problems that can be solved
with data mining.
Once the basic concepts are understood, Section II provides the tools for knowledge
discovery with detailed tutorials taking you through the knowledge discovery process.
The fact that data preprocessing is fundamental to successful data mining is empha-
sized. Also, special attention is given to formal data mining evaluation techniques.
Section III is all about neural networks. A conceptual and detailed presentation is offered
for feed-forward networks trained with backpropagation learning and self-organizing
maps for unsupervised clustering. Section III contains several tutorials for neural network
learning with Weka and RapidMiner.
Section IV focuses on several specialized techniques. Topics of current interest such as
time-series analysis, textual data mining, imbalanced and streaming data, as well as Web-
based data mining are described.
 Preface ◾ xxxv

Section I: Data Mining Fundamentals

• Chapter 1 offers an overview of data analytics and all aspects of the data mining pro-
cess. Special emphasis is placed on helping the student determine when data mining
is an appropriate problem-solving strategy.
• Chapter 2 presents a synopsis of several common data mining strategies and tech-
niques. Basic methods for evaluating the outcome of a data mining session are described.
• Chapter 3 details a decision tree algorithm, the Apriori algorithm for producing asso-
ciation rules, a covering rule algorithm, the K-means algorithm for unsupervised
clustering, and supervised genetic learning. Tools are provided to help determine
which data mining techniques should be used to solve specific problems.

Section II: Tools for Knowledge Discovery

• Chapter 4 presents a tutorial introduction to Weka’s Explorer. Several tutorials pro-

vide a hands-on experience using the algorithms presented in the first three chapters.
• Chapter 5 introduces RapidMiner Studio 7, an open-source, code-free version of
RapidMiner’s commercial product. The chapter parallels the tutorials presented in
Chapter 4 for building, testing, saving, and applying models.
• Chapter 6 introduces the knowledge discovery in databases (KDD) process model
as a formal methodology for solving problems with data mining. This chapter offers
tutorials for outlier detection.
• Chapter 7 describes formal statistical and nonstatistical methods for evaluating the
outcome of a data mining session. The chapter illustrates how to create and read
Pareto lift charts and how to apply RapidMiner’s ANOVA, Grouped ANOVA, and
T-Test statistical operators for model evaluation.

Section III: Building Neural Networks

• Chapter 8 presents two popular neural network models. A detailed explanation of

neural network training is offered for the more technically inclined reader.
• Chapter 9 offers tutorials for applying Weka’s MultilayerPerceptron neural network
function for supervised learning and Weka’s SelfOrganizingMap for unsupervised
clustering.
• Chapter 10 presents tutorials on applying RapidMiner’s Neural Network operator for
supervised learning, and Self-Organizing Map operator for unsupervised clustering.
xxxvi ◾ Preface

Section IV: Advanced Data Mining Techniques

• Chapter 11 details several supervised statistical techniques including naive Bayes

classifier, support vector machines, linear regression, logistic regression, regression
trees, and model trees. The chapter contains several examples and tutorials.
• Chapter 12 presents several unsupervised clustering techniques including agglomera-
tive clustering, hierarchical conceptual clustering, and expectation maximization (EM).
Tutorials on using supervised learning for unsupervised cluster evaluation are presented.
• Chapter 13 introduces techniques for performing time-series analysis, Web-based
mining, and textual data mining. Methods for dealing with large-sized, imbalanced,
and streaming data are offered. Bagging and boosting are described as methods for
improving model performance. Tutorials and illustrations for time-series analysis,
textual data mining, and ensemble learning are presented. A detailed example using
receiver operator curves is offered.
• Chapter 14 provides a gentle introduction to data warehouse design and OLAP. A
tutorial on using Excel pivot tables for data analysis is included.

INSTRUCTOR SUPPLEMENTS
The following supplements are provided to help the instructor organize lectures and write
examinations:

• PowerPoint slides. Each figure and table in the text is part of a PowerPoint presenta-
tion. These slides are also offered in PDF format.
• A second set of slides containing the screenshots seen as you work through the
tutorials in Chapters 4 through 14.
• All RapidMiner processes used in the tutorials, demonstrations, and end-of-chapter
exercises are readily available together with simple installation instructions.
• Test questions. Several test questions are provided for each chapter.
• Answers to selected exercises. Answers are given for most of the end-of-chapter
exercises.
• Lesson planner. The lesson planner contains ideas for lecture format and points for
discussion. The planner also provides suggestions for using selected end-of-chapter
exercises in a laboratory setting.

Please note that these supplements are available to qualified instructors only. Contact
your CRC sales representative or get help by visiting https://www.crcpress.com/contactus
to access this material. Supplements will be updated as needed.
 Preface ◾ xxxvii

USING WEKA AND RAPIDMINER

Students are likely to benefit most by developing a working knowledge of both tools. This
is best accomplished by students beginning their data mining experience with Weka’s
Explorer interface. The Explorer is easy to navigate and makes several of the more dif-
ficult preprocessing tasks transparent to the user. Missing data are automatically handled
by most data mining algorithms, and data type conversions are automatic. The format
for model evaluation, be it a training/test set scenario or cross-validation, is implemented
with a simple click of the mouse. This transparency allows the beginning student to
immediately experience the data mining process with a minimum of frustration. The
Explorer is a great starting point but still supports the resources for a complete data min-
ing experience.
Once students become familiar with the data mining process, they are ready to advance
to Chapter 5 and RapidMiner. RapidMiner Studio’s drag-and-drop workflow environ-
ment gives students complete control over their model building experience. Just a few of
RapidMiner’s features include an intuitive user interface, excellent graphics, and over 1500
operators for data visualization and preprocessing, data mining, and result evaluation. An
interface for cloud computing as well as extensions for mining textual data, financial data,
and the Web is supplemented by a large user community to help answer your data mining
questions.
Here are a few suggestions when using both models:

• Cover the following sections to gain enough knowledge to understand the tutorials
presented in later chapters.
• If Weka is your choice, at a minimum, work through Sections 4.1, 4.2, and 4.7 of
Chapter 4.
• If you are focusing on RapidMiner, cover at least Sections 5.1 and 5.2 of Chapter 5.
• Here is a summary of the tutorials given in Chapters 6, 7, 11, 12, 13, and 14.
• Chapter 6: RapidMiner is used to provide a tutorial on outlier analysis.
• Chapter 7: Tutorials are presented using RapidMiner’s T-Test and ANOVA opera-
tors for comparing model performance.
• Chapter 11: Both models are used for tutorials highlighting naive Bayes classifier
and support vector machines.
• Chapter 12: RapidMiner and Weka are used to illustrate unsupervised clustering
with the EM (Expectation Maximization) algorithm.
• Chapter 13: Both RapidMiner and Weka are employed for time-series analysis.
RapidMiner is used for a tutorial on textual data mining. Weka is employed for
a tutorial on ROC curves. RapidMiner is used to give an example of ensemble
learning.
xxxviii ◾ Preface

• Chapter 14: Tutorials are given for creating simple and multidimensional MS
Excel pivot tables.
• Chapter 9 is about neural networks using Weka. Chapter 10 employs RapidMiner
to cover the same material. There are advantages to examining at least some of the
material in both chapters. Weka’s neural network function is able to mine data hav-
ing a numeric output attribute, and RapidMiner’s self-organizing map operator can
perform dimensionality reduction as well as unsupervised clustering.

SUGGESTED COURSE OUTLINES

The text is appropriate for the undergraduate information systems or computer science
student. It can also provide assistance for the graduate student who desires a working
knowledge of data mining and knowledge discovery. The text is designed to be covered in
a single semester.

A Data Mining Course for Information Systems Majors or Minors

Cover Chapters 1, 2, and 3 in detail. However, Section 3.5 of Chapter 3 may be omitted or
lightly covered. Spend enough time on Chapters 4 and 5 for students to feel comfortable
working with the software tools.
If your students lack a course in basic statistics, Sections 7.1 through 7.7 can be lightly
covered. The tutorial material in Sections 7.8 through 7.10 is instructive. Section 8.5 can be
excluded, but cover all of either Chapter 9 or Chapter 10. Cover topics from Chapters 11
through 13 as appropriate for your class. For Chapter 13, all students need some exposure
to time-series analysis as well as Web-based and textual data mining.

A Data Mining Course for Undergraduate Computer Science Majors or Minors

Cover Chapters 1 through 13 in detail. Spend a day or two on the material in Chapter 14 to
provide students with a basic understanding of online analytical processing and data ware-
house design. Spend extra time covering material in Chapter 13. For a more intense course,
the material in Appendix B, “Statistics for Performance Evaluation,” can be covered as part
of the regular course.

A Data Mining Short Course

The undergraduate or graduate student interested in quickly developing a working knowl-
edge of data mining should devote time to Chapters 1, 2, and 3, and Chapter 4 or 5. A
working knowledge of neural networks can be obtained through the study of Chapter 8
(Sections 8.1 through 8.4) and Chapter 9 or Chapter 10.
Acknowledgments

I am indebted to my editor Randi Cohen for the confidence she placed in me and for allow-
ing me the freedom to make critical decisions about the content of the text. I am very grate-
ful to Dr. Mark Polczynski and found his constructive comments to be particularly helpful
during revisions of the manuscript. Finally, I am most deeply indebted to my wife Suzanne
for her extreme patience, helpful comments, and consistent support.

xxxix
Author

Richard J. Roiger, PhD, is a professor emeritus at Minnesota State University, Mankato,

where he taught and performed research in the Computer Information Science Department
for 27 years. Dr. Roiger earned his PhD degree in computer and information sciences at the
University of Minnesota. Dr. Roiger has presented conference papers and written several
journal articles about topics in data mining and knowledge discovery. After retirement,
Dr. Roiger continues to serve as a part-time faculty member teaching courses in data min-
ing, artificial intelligence, and research methods. He is a board member of the Retired
Education Association of Minnesota, where he serves as their financial advisor.

xli
 I
Data Mining Fundamentals

1
Random documents with unrelated
content Scribd suggests to you:
Thanking the good woman behind the counter for her help, I set out
to find Mrs. Fred Frost.
The door of No. 18, Cripps Court, was opened by a wan, haggard-
looking woman, whom the summons had apparently disturbed in the
act of suckling a sickly-looking baby, which she held on one arm,
while the hand of her other arm was fumbling at the unbuttoned
bosom of her dress.
"Good morning," I said, raising my hat. "Can you tell me, please, if
Mrs. Fred Frost is in?"
"No, sir, she's not," she answered civilly; "her baby's dead, and she's
gone to find her husband, who's trying to get a job at the docks."
"Oh! Poor woman! I'm very sorry!" I said, gently. "The fact is, her
name has been given me as one whose husband is out of work, and
I ventured to call to see if she'd allow me to send in some groceries,
and other things by way of being of some small assistance during
the hard time. If you'll allow me, I'll call again."
"She'll be very grateful to you, sir, I'm sure," the woman replied.
"Having the child ill has made it very hard for her just lately."
"Is there anybody else living here with whom things are going
badly? If so, perhaps you'd tell me! I can't do very much, but what I
can do, in the way of sending in some tea and some meat and a few
groceries, I'd be very glad to."
"Well, sir," answered the woman, "there's an old couple in the back
room, living alone with their little grandson (the child's father and
mother are dead). But they've gone out—all three of them—to try
and get a relief ticket somewhere. If you were to ask for them when
you come back to see Mrs. Frost, you'll see for yourself by the very
look of them how things are. The little boy—he's all right. They've
managed it, though I don't know how, between them, 'cept by
starving themselves to give to him, for skin and bone is about all
that's left of the two old people."
"I'm very much obliged to you," I said. "And I shall venture, as I say,
to call in again, perhaps in an hour's time. Good morning."
"Good morning, sir, and thank you," she said quietly.
As I was turning away the sun, which had not before been visible
that morning, suddenly broke out from behind the clouds. Standing,
as she had been, in a dark passage, and partly behind the half-
opened door, she was so much in the shadow that I could not
observe her closely; nor, for the matter of that, had I tried to do so,
being anxious not to seem curious or inquisitive. But as the sun fell
full upon her face, and I marked the hollows in her cheeks, and the
dark rings around her eyes, I stopped suddenly, impulsively.
"Please don't think me impertinent," I said. "But you look far from
strong yourself. I hope—I do hope—your husband isn't out of work,
too."
"Yes, sir; he's been out five weeks now, come Tuesday."
"And have you any children? Again I ask you to forgive me."
"There's no offence, sir," she said quietly, but I saw that she was
trying hard to stay the trembling of her lips.
"Yes, sir, I've five, and—and—there's been no food in the house
since yesterday."
"Yet you never asked help for yourself!" I said, gentle reproach
perhaps in my voice, but wonder and reverence at my heart. "You
are a brave woman, a true woman, and I honour and respect you.
But, for the children's sake, you mustn't refuse, if I ask you to let me
try to be of some little help while the hard time lasts."
She was sobbing piteously now—more, I suspect, because she was
faint and weak and in want of food than for any other reason.
"I'm sure I've—I've—I've tried hard to get some work, and so's Joe."
Then she pulled herself together.
"Will you come in, sir?"
Uncovering, I followed her into the wall-bare room. I say wall-bare
advisedly, for, except for an old box in the corner, every stick of
furniture had, as I discovered, been pawned or sold for food. Yet
here seven of my fellow-creatures, made in the image of God, were
herded together, within the space of a few square feet.
A wan, ragged, and unkempt man was sitting on the upturned box,
his elbows on his knees, his hands thrust in the hair that was bushed
over his ears.
He leapt up morosely, savagely, at my entrance, and muttered
something about "More —— spies!"
But I was not born an Irishman for nothing. Three minutes had not
passed before I had won him to friendliness; five minutes had not
gone by before the youngest child was sitting on my knee, listening,
open-mouthed, to stories about a performing dog.
After a little time I said:
"Now I wonder if Timmy there—he's nine, you said, Mrs. Wright—I
wonder whether he's a good hand at shopping, and if he'd come
with me to get a few things at the butcher's, and the baker's, and
the grocer's, and then help me to bring them back? Do you know,
Timmy, I'm a very, very greedy man, and want a cup of tea badly;
and somehow I've got an idea that your mother, here, is a good
hand at making tea; and when you and I come back, I'm going to
beg her to be so very kind as to make me a cup, and then, while all
the rest of you have a cup too, and something to eat with it, I'll
finish that story of mine about the dog."
But I had miscalculated Timmy's strength. He and I stopped at the
first shop we came to—a grocer's—and borrowed a wicker basket. It
had the word "Margarine" stencilled or painted in big black letters on
one side; and by the same token, as, for weeks to come, I had
occasion to borrow that same basket, and came to be a familiar
figure in the streets, I was known and spoken of in the district as
"Mr. Margarine." Into this basket Timmy and I stacked away tea,
sugar, butter, and other groceries. Then we returned to my friend
the woman in the baker's shop, and added to our store a loaf or two
of bread.
"And now, Timmy," I said, "I daresay this kind lady could find you a
piece of cake and a glass of milk. Meanwhile, I'll run across the road
and interview the butcher."
I had hardly entered the butcher's shop before I heard the sudden
pulling up of a horse and cart in the street, and saw the driver
hastily dismounting.
Timmy, supposing I had meant him to follow at once with the
basket, had taken it up, and must have passed out almost at my
heels. Half-way across the street he had suddenly reeled and fallen,
and now lay white and unconscious.
"What's the matter?" I asked the woman who was kneeling beside
him with his head on her arm.
"Oh, nothing!" she answered, bitterly. "You've got eyes in your head,
haven't you, and can see for yourself? He's fainted for want of food
—that's what's the matter. He's only starving!"
 CHAPTER XXVII.
RE-ENTER THE DUMPLING.
"He's only starving!" the woman had said. But as I stooped to lift the
frail little figure from the ground, as I hurried with it across the
street and into the baker's shop, there to wet the white lips with a
restorative—a prayer that was like the spurting of blood from a
wound, a prayer that shot a pang of actual physical pain to my heart
—so poignant, so terrible was my remorse—surged up unuttered but
not unheard, within me.
"Lord Christ, lover of little children, spare and heal this Thy stricken
little child. Forgive and pardon such as I, who, living our selfish,
easeful lives, have closed our ears and shut our eyes to the suffering
and the misery around us. Help us to repent. Lash us, if need be,
with Thy whip of pain, burn us, if it seem good to Thee, with the
scorching of fire, but awaken us, arouse us, thaw this cruel, frozen
heart within us, that we be forgetful of the sufferings of these our
fellow-creatures no more. Amen."
And the prayer was heard—as every prayer that is uttered in earnest
is heard—on high. The restorative did its work. The cup of warm
milk, which—perhaps with a thought of Him who said, "And
whosoever shall give to drink one of these little ones a cup of cold
water only ... shall in no wise lose his reward"—was held to his lips
by the good woman, put new life into him; and before long Timmy
and I were able to gather together our belongings, and to return to
No. 18, Cripps Court.
There, his mother brewing the tea, and the rest of the household
sitting around discussing the contents of my basket, I made the
happiest meal of my life. The Wright family were equally happy, and
listened to the continuation of my story of the performing dog with
uproarious laughter.
The mere fact that one, not in their own station of life, could, by
coming among them as one of themselves—asking no questions,
laying down no laws, but by talking to the children, taking the little
ones upon his knee, to show his watch or the dog's head on his
walking-stick—so please and delight them, was to me tragic.
The good fellowship, the realisation of the common bond of human
sympathy which—whether they wear corduroy or broadcloth, silk
hats or coster caps, satin gown or cotton frock—should bind men
and women together, seemed to cheer them in spirit as much, if not
more, than their bodies had been cheered by the food.
After a time Mrs. Wright departed, taking the children with her, to
execute some shopping commissions for me, and then Wright and I
settled down for a chat.
It was not long before I learned his reason for fearing, when I had
first entered, that I was what he had called "another of the ——
spies."
"It's this way, sir," he said. "Two ladies called afore you come. The
missis was out, or she wouldn't have let 'em in, but they came in
without bein' asked, and sat down—at least, one on 'em did—on this
box, and the other, she walked all around the place, sniffing like to
herself, and talking aloud about what she called the 'dirt.'
"'So your name's Wright, and you're out of work, are you?' says the
sitting-down one to me.
"'Yes,' I says, short like.
"'And what church do you attend?'
"'I don't attend none,' I says. 'The missis she goes sometimes of a
Sunday evening, and the kids goes to Sunday school.'
"'And wot church does your wife go to—and what Sunday school do
the children attend?' puts in the other lady, having done all the
sniffing round she could.
"'Wesleyan,' I says.
"'H'm!' she says. 'Dissenters! That's bad! Don't you and your wife
know that the Church of England is the Established Church of the
land, and that it's displeasing to Almighty God—not to say anything
about your wife's duty—in letting the children attend a sectarian
place of worship?'
"'No,' I says; 'I don't.'
"'And what wages do you earn when you're in work, Wright?' she
says.
"'Well,' I says, 'when I was a farm labourer I only made thirteen bob
a week, but since I come to town I got about eighteen shillings
when I was in work.'
"'And how much did you save out of that?'
"'Why, nothink,' I said. 'I 'ad a 'ouse then, and the rent was eight
and six a week; and nine and six don't go far to feed and clothe
five.'
"'Nonsense!' she says. 'You should always put by a percentage of
what you earn, no matter how small it is, against a rainy day. The
improvidence, the thriftlessness, of you poor is criminal! I'm afraid
you're an idle, worthless fellow, Wright,' she says, 'and not deserving
of any help until I see that you've reformed your ways.'"
Then, moved perhaps by the memory of his recent privations, he
suddenly seemed to lose all self-control. Springing to his feet, he
shook his clenched fist in the air, as with flaming face and voice
hoarse with frenzy, he shouted:
"Reform my ways? Idle, worthless fellow, am I? Ah! but the days of
grinding us down are gone. The days of them that oppress the poor
are over! It's Labour that's going to lord it now, and make the laws,
and rule the land. The millions on millions who starve and sweat and
labour are no longer to be the toys and tools of the few thousands
that sweat them, so that the idle few may live on the fat of the land.
The millions have found their leader at last—the man to organise
them into armies, and to lay the mines of the revolution that's close
at hand, the man that's——"
"Cease your ranting!" I interrupted sternly, for his voice had risen to
a screech. "Do you want to arouse the neighbourhood? I've heard all
that stuff before, and from the lips of the man who first uttered it—
for you are not speaking your own words. You are only speaking
words that have been put into your mouth. Wrongs you have, no
doubt, but rant and rioting won't right them, and I tell you——"
I was not allowed to finish the sentence. Someone who had no
doubt been listening in the passage outside suddenly sprang in, and
before I could turn—for my back was to the door—had seized me so
as to pin both my arms to my sides.
I did not need to be told the name of my assailant, to see his face,
or to hear his voice. Only one man whom I had ever met had arms
of such gorilla-like length, of such giant strength.
It was the Dumpling!
 CHAPTER XXVIII.
MUTINY AND A MESMERIST.
"You noisy fool, Wright!" he said, still holding me in his iron grip. "I
heard you ranting in the street outside." Then, "Am I hurting you,
Mr. Rissler?" he asked.
Foolishly I opened my mouth to reply, but hardly had I done so
before a handkerchief was thrust between my lips, and stuffed the
next instant half-way down my throat.
"Now, Wright," he said, "you can tie his hands and feet while I hold
him. Do it so that the knot won't give, but carefully, so as not to hurt
him. Come along, my good man! You are very slow! Do as I bid you,
and be quick about it!"
Somewhat to my surprise, and more so, evidently, to the
Dumpling's, Wright refused point-blank.
"I won't," he said, not budging from where he stood. "The
gentleman's been kind to me and mine, and lift a hand against him I
won't."
"Do as I bid you!" was the answer, in a voice that was almost a hiss.
"I won't," was the dogged reply.
The next jiffey I was lifted off my feet by the Dumpling's powerful
arms, and, before I knew what he was about, I was lying upon my
back on the floor, bewildered but unharmed.
Putting a knee upon my chest, and holding me down meanwhile
with both hands, he raised his head and looked Wright straight in
the eyes.
For the space of a few seconds the fellow stood gaping at the
Dumpling as if fascinated. Then his eyes dulled to a fixed, foolish
stare, and finally the lids of them drooped and closed. He was
mesmerised—that fact was evident.
"Open your eyes," the Dumpling commanded.
Wright did so.
"Put your hand in my right-side pocket, and you'll find a strap."
Automatically the other obeyed.
"Strap his ankles securely. No; it is no use your struggling, Rissler,"
he went on, as, getting my feet against the wainscot for leverage, I
put all my strength into an effort to throw him off.
Then, looking over his shoulder, he inquired:
"Have you got that strap, Wright?"
"I have got the strap," the other replied mechanically.
"Then be ready to use it," said the Dumpling. "Slip it around his
ankles when I say 'Now'!"
Still holding me down with both hands, he suddenly straddled his
powerful knees open, as if they had been a pair of nut-crackers, and
the two legs of me—outside which his own legs now lay—the two
halves of a walnut. Then he closed his knees, clipping mine together
within them, as the pair of crackers might clip the shell; and though
I struggled with all my strength, for I was furious at finding myself
proved to be not only as a child in his hands, but as an infant upon
his knees—or between them—I knew, when I heard him say 'Now,'
that I was as neatly hobbled as any horse.
"It's a pity you put us and yourself to all this unnecessary trouble,"
he said philosophically. "Now, then, Wright, look in his pockets for a
handkerchief, and lash his wrists while I hold them together."
As he spoke, the five fingers of his right hand closed on my left
hand, as the talons of a pouncing hawk close upon a field-mouse.
Then, in spite of my futile efforts, he drew my left hand towards my
right, and suddenly spanning the two with the fingers of his single
hand, nonchalantly arranging the set of his collar with the other
meanwhile, he said:
"Now, Wright, just tie his wrists together, will you?" and once again
the mesmerised man did as he was bidden.
Rising, the Dumpling turned from me to Wright. Looking that worthy
full in the eyes, he said, in a low voice:
"You are awake."
"I am awake," repeated Wright, automatically, as his eyes, still fixed
in a stupid stare, turned sluggishly in their sockets, following his
master's movements.
Lifting his dropped left arm slightly, the Dumpling touched, with a
finger tip, the finger tips of the other's loosely open hand. Then he
put his right hand on Wright's shoulder, and with a gentle shake,
said:
"Wake!"
Wright came to himself with a start, and looked foolishly around.
"What did you mean by disobeying me just now?" asked the
Dumpling with sudden fierceness.
"I'm sorry, sir; I never done it before, as you know, and I'm sorry if I
opened my mouth too wide and made a noise. Me and mine has
gone through hard times since you were here last. The money you
left us was finished nigh upon two weeks ago, and not knowing
where you was, and not being able to get any work, we've had
nothing to eat. This gentleman here has been very kind to us, and
though I've never disobeyed you before, sir, I couldn't lift hand
against him, and that's a fact."
The Dumpling seemed genuinely concerned.
"Has it been so bad as that with you, my poor fellow?" he said
gravely. "Humphreys shall answer to me for this. I had to be away,
it's true, but I left plenty of money with him for all requirements,
and I particularly told him that no one, who is in with us, should be
allowed to want.
"I'm glad my friend on the floor has been good to you," he
continued. "He's of the right stuff, and ought to be with us, Wright—
will be with us one day, I hope and believe. You needn't be anxious
about him. I don't mean to do him any harm. On the contrary, I
want to do him a good turn, if he'll let me. My reason for gagging
him was because, when I first came in, I thought you and he were
quarrelling. One doesn't do a man any serious amount of damage by
sticking a piece of rag in his mouth and tying him up for a half-hour.
He, on the other hand, not understanding the situation, and
supposing himself to be in danger, might have called out and
brought the police about our ears. He'd be sorry for it after, when he
knew the facts, so we protected ourselves, and protected him
against himself, by taking the precaution of making sure of him first.
See?
"Well, now, I want to have a few words with the gentleman, and I
can do so better if he and I are alone. You stay outside the door
meanwhile, Wright, and see that no one comes in."
Wright gone, and the door closed and locked, the Dumpling turned
to me.
"Now, Mr. Rissler, to finish our interrupted conversation of the other
night. But first of all about that gag. If I take it off, will you give me
your promise, your parole, not to call or cry out, or do anything to
bring outsiders in? Nod if you agree. Shake your head, if not."
I nodded.
"All right," he said. "But you'll be more comfortable sitting up, with
your back to the wall, than lying on the floor full length. See, I'll
prop you up that way, and now to take off the gag."
While he was doing so I was planning a little surprise for him. Up to
now he had, no doubt, for reasons of his own, been very careful to
conceal his name from me. He had, to be sure, declared himself to
be Napoleon, but Napoleon's present address is not to be found in
any directory available for the purpose.
That my theory of the identity of John Carleton, of No. 5, Taunton
Square, with the man whom I knew as the Dumpling was correct, I
was, in my own mind, entirely persuaded; but I was shy of theories,
and anxious to replace them by actual knowledge. I was curious,
too, to see what effect the fact that his real name was known to me
would have upon him. Would he admit it? Would he deny it? Would
it arouse him to anger or to fear? Means of knowing I had none,
other than by putting the matter to the test; so, no sooner was I
sitting up, with the gag removed, than I looked him straight in the
face and shot my bolt.
"Well, and how are you, Mr. John Carleton?" I said.
 CHAPTER XXIX.
THE MYSTERY OF THE THIRD MAN.
That my knowledge of his identity would come to him as a surprise, I
was tolerably sure; and, in springing it thus suddenly upon him, I
hoped to gather by his face, which I was watching narrowly
meanwhile, whether he were moved by anger or by dismay.
The result was a fiasco, a fizzle, an abject failure. Either he was the
most consummate of actors, or, as is more likely, I the most
consummate of fools; for he looked at me at first with bewilderment,
and then with amusement, and finally burst into a hearty laugh.
"So that's who you think I am, is it?" he chuckled. "I saw by that
speaking countenance of yours that you were hatching an egg (you
are an Irishman, aren't you? so you won't mind my mixing my
metaphors), but I didn't expect so weird a fowl as this."
He took that morning's paper from his pocket, and holding it open
before me, pointed to a paragraph, which ran as follows:
"The freedom of the City of Carlisle will to-day, at noon, be conferred
upon Mr. John Carleton, the eminent inventor, who was born in that
city, and has devoted a large portion of the fortune he amassed in
America to improving the homes of the poor in the place of his birth.
Miss Clara Carleton and her niece, Miss Kate Carleton, have left 5,
Taunton Square, W., for Homburg, where Mr. Carleton will shortly
join them."
"I admit," commented the Dumpling ironically, "that the fact of Mr.
Carleton devoting a portion of his fortune to improving the homes of
the poor, points suspiciously to me. How are you to be sure, Mr.
Rissler, that I didn't send that paragraph myself to the newspaper, in
the expectation of meeting you here, and, by showing it to you, to
put you off the scent? And the portrait! Stop a minute—there is a
portrait. It's in another part of the paper. Ah! here it is."
The portrait, which was inscribed "John Carleton, Esq., the eminent
inventor and philanthropist," was that of a flat-foreheaded, clean-
shaven man, absolutely bald, and so shrunken and fleshless of face
as to seem all skin and bone. The striking resemblance to a skull,
caused by his singular fleshlessness, was heightened by the fact that
he wore huge glasses of great magnifying power, out of which his
hollow eyes loomed cavernously.
"It's like me, isn't it?" said the Dumpling. "But then you are not to be
deceived by anything of that sort, are you? It may dupe the police,
but you will see at a glance that that portrait is only myself, cleverly
got up to look like somebody else. And when you read in to-
morrow's papers an account of the ceremony at Carlisle—and, by
the bye," pulling out his watch, "it's just about taking place now—
you will say to yourself, 'What a thing it is to be a detective! Here's
all this hullaballoo about the conferring of the freedom of the City of
Carlisle upon Mr. John Carleton at noon yesterday; and all the time I
could have told them—if they'd only asked me, which for some
inexplicable reason they didn't—that the real John Carleton was at
18, Cripps Court, Shadwell.'"
Then the bantering tone died out of his voice.
"I've chaffed you a bit about the mistake, Mr. Rissler, but it was a
very natural mistake to fall into," he said. "My coming to No. 5,
Taunton Square, twice in a night, my shadowing it beforehand, my
knowing all about it—no, I can't see, as I say, that you are very
much to blame, after all. It is curious that you should have thought
me to be John Carleton when, if there is one man in all the world
with whom I would not, for all the world can offer, change places,
that one man is John Carleton.
"When I tell you the story of my life, as I hope I shall one day, you
will understand and appreciate—but not till then—my reason for
speaking as I do. But now to talk of other matters. You have thought
over what I said to you the other night?"
"I have," I answered.
"And you are going to throw in your lot with us?"
"No."
"Rissler," he said gravely, "don't compel me to kill you, as kill you I
must under certain circumstances. There is blood on my hands
already, and more blood on my hands there must, of necessity, be
before my work is done; but, of all men in the world, you are the
one man whom I am most anxious to spare. You must see for
yourself that you know too much, that you are too dangerous to be
left at large, except as one of us. I have shown my heart, my hopes,
my dreams, to you, as I have shown them to no other man. Can I
do nothing, say nothing, to influence you? You feel as I do about the
poor, as witness the fact of what you have done for Wright and his
family. What moved you to do that? What brought you here, in
Wright's house, at all?"
He stopped, as if expecting an answer; so, briefly as possible, I told
him of the impression that had been made upon me by his words
about the poor, and that, for the time, at least, I had thrown up my
detective work, in order to devote myself to doing what lay in my
power to alleviate the sufferings of my fellow-creatures.
He was genuinely moved, and when he inquired how I came to
know of Wright, and I told him of my system of making inquiries at a
small shop, in the very poorest district, he put a hand upon my
shoulder, and said excitedly:
"Rissler, you have no choice in this matter. God has called you to the
task, and you may not say Him nay. It is only a question of time.
Two days—only two days ago, you were against us. You announced
yourself as my enemy, as one who was set upon hunting me down.
Now you tell me of your own accord—and I believe you—that you
have abandoned this ignoble work of hunting down a fellow-creature
who, whether his methods be right or wrong in your eyes, is at least
consumed with a passionate desire to spend and to sacrifice
everything he has, life itself, if necessary, to succour and to help the
poor and the oppressed.
"Two days ago, Rissler, as I say, you were against us. To-day you are
against us no more. Two days ago you cared nothing for the
sufferings of the poor, you gave no thought to them. To-day you are
here amongst them, ministering to them with your own hand. If two
days have wrought this change in you, what change may not
another two days work? Another two days may see you working with
us, one of us, leading the Labour hosts in this battle of the Lord.
"Now, listen to me. I'll be frank with you, and tell you that from the
first moment I saw you something within me warned me to beware
of you, and cried out, 'Kill! kill! kill!' That night in the wine cellar, to-
day in this squalid room, I should, had I followed my impulses, have
strangled you without mercy, without remorse, and without a
thought. Why don't I kill you? Why do I spend time which I can't
afford to spend? Why do I run risks which I never ought to run, in
talking to you, in explaining things to you, in trying to persuade you
to join us?
"I will tell you. It is because God has revealed to me that you are
destined to play a great part in the history of this rising. It was by no
chance that you came that night to the opium den. It was by no
chance, it was not entirely by your resourcefulness and skill, that
you escaped with your life. It was no chance which drew you to the
house in Taunton Square, no chance which sent you here to Cripps
Court.
"The part you are to play, God has not yet revealed to me; but I will
tell you what I believe that part to be. The army I command may be
counted by many millions, but leader there is only one—myself. And
the battle—which shall be called Armageddon—the battle which shall
set Labour upon the throne as Lord and Ruler of this land—that
battle approaches, and in that battle I shall fall. If I fall, all falls,
unless God raise up a second in command who shall be the leader of
the people after I am gone. That leader I believe you are marked
out to be. That is why I dare not kill you; that is why I am going to
do the maddest thing a sane man ever did.
"Of my own will I set you free to go from here unharmed. As yet you
are not with us. As yet God has not made known His will to you. As
yet, though I have twice appealed to you to throw in your lot with
us, you have resisted my entreaty. But I am not dismayed. Once
again I shall come to you. Once again I shall appeal to you, and that
third time I shall give you such assured proof of the triumph of our
cause, that after that third time I shall need to appeal to you no
more. The victory will be won. Our cause, the people's cause, God's
cause, you will, on that third appeal, espouse. Of my own accord I
set you free."
As he spoke these last words, he stooped to unfasten my bonds,
and, in doing so, looked me for a moment in the eyes. Once again a
dream-tableau seemed to shape itself before me.
I saw myself—as one might see another person—in some dark place
underground. By my side stood the Dumpling, and far back in the
shadow was another man.
In my dream I could not see the face of the third man. I could see
the Dumpling's face, and upon it was a look of fiendish triumph as
he pointed me to the third man. But I could see my own face, and
on my own face, as it was turned to the face of the third man, was
such a look of incredulous horror as, waking or dreaming, I shall
never see upon human face again.
Then the dream passed. I was free of my bonds, and the Dumpling
was holding open the door.
"Good-bye, Rissler!" he said. "We shall meet again and soon; and
that meeting will mean great things for both of us."
Without a word, without so much as a "Thank you" or a "Good day,"
I passed out, like one who walks in his sleep. I could think of
nothing save the unseen face of the third man in my dream, and of
the incredulous horror which had been upon my own face on
realising who that third man was.
 CHAPTER XXX.
FORTY MILES IN A PERAMBULATOR!
My discomfiture at the Dumpling's derisive repudiation of the
supposition that he was John Carleton was completed next morning,
when John Carleton himself returned to town, and John Carleton in
the flesh I with my own eyes several times saw, as he went in and
out of his house in Taunton Square.
"That man," said I to myself, "should be an object-lesson to you in
the futility of theory building. First, you called yourself a fool for not
having seen that John Carleton was the Dumpling, and the
Dumpling, John Carleton. Now, you have the pleasure of knowing
yourself a double-distilled donkey, for ever having supposed
anything of the sort."
Upon the theory—the fact, as I had thought it to be—that John
Carleton and the Dumpling were one and the same man, rested the
only explanation I had to offer in regard to the letter I had received
from the two Miss Carletons, aunt and niece. That theory being now
entirely exploded, their extraordinary behaviour remained as much,
if not more, of a mystery than ever.
A mystery—so far as I was concerned—I decided that it might
remain. Of detective work and of theory building I had had more
than enough, and so I betook myself that very afternoon to
Shadwell, to renew the investigations which my meeting with the
Dumpling had interrupted.
The first name on my list was that of a tailor's "hand," named
Holmes, a widower who, I was told, had five young children, and
was out of work. He was a consumptive-looking creature, hollow of
cheek, eye, and chest, and with a hacking cough.
"Yes, sir," he said civilly, in reply to my inquiries. "It is quite true that
I am out of work, and that I have children; but I can't take your
help, asking your pardon all the same, sir, for seeming rude and
ungrateful."
"On the contrary," I said, "it is I who have to apologise to you, Mr.
Holmes, for what you might very well think my impertinence in
coming here at all. But I happened to hear, quite by chance, how
beautifully you keep your children; and how nice they always look;
and learning that you were out of work, and being very, very fond of
children (I haven't any myself: I wish I had), I thought there
wouldn't be any harm, at least, in calling, just to see whether there
were any little thing I could do for you, until you're in work again.
I'm a working man, as you are, though I happen to work with a pen,
while you happen to work with a needle. And I'm a poor man, too,
for the matter of that; but just lately I chanced, by a stroke of luck,
to make a pound or two more than usual, and when I have a stroke
of luck I like to share it with someone who has been less lucky—just
as I believe you'd be ready to share your good luck, when it comes,
with me, if I happened to need it. But I respect your independence
and pride, and I ask you again to forgive me for calling."
"It isn't pride, sir," he said; "and, if the children were in absolute
want, I'd take your help and thank God for it. It's this way, sir. This
week we have just enough money left out of my savings to last us—
me and the children—in bread. It has only been bread, and dry
bread, it's true; and if when Monday comes I haven't got work, there
won't even be bread, for my money will be entirely gone. If you
should be this way then, and would look in, and I haven't found
work, I will take your help—putting it the way you do, sir—and thank
God for it. But when I know of hundreds of little children who
haven't had even a piece of bread for days, I can't take——But I
thank you kindly. God bless you, sir. I must go now. I hear one of
the children calling. Good afternoon."
He closed the door in my face—not rudely, but in haste, lest I should
see how shaken he was by emotion; and bowing my head, and with
my own heart rising strangely in my throat, I turned away.
Just for the moment, I did not feel like facing the eyes in the street;
so, as a slight rain was falling, I took shelter in a dark passage
leading to a court, and stood there out of sight of passers, to collect
my thoughts.
It was not long before my attention was attracted by a curious sight.
A gipsy-like, wolf-faced man was wheeling a child's perambulator, in
which, to my astonishment, I saw curled up the figure of a full-
grown woman. I recognised the couple at a glance. Walking once
along the high road from Epping to London, I had seated myself
upon a five-barred gate by the wayside for a quiet smoke. The gate
stood between thick hedge-rows, and, as it was set back a little, the
folk passing along the road could not see me until they were almost
level with the gate. By and by I heard what struck me as a very
pretty altercation between a man and a woman who were
approaching me slowly, but whom as yet I could not see. The man,
as I discovered when they came into sight, was wheeling a
perambulator (the same perambulator, in fact) in which were a
number of ferns and primrose roots that he was carrying to London
to sell. This perambulator the woman was pleading to be allowed to
take a turn at pushing, urging that as the man had been up since
four in the morning to gather the ferns and primroses, and had had
to wheel the perambulator five miles out and five miles back, he
must consequently be very tired. He, protesting that he was not
tired at all, point-blank refused, declaring that, as she had only just
come out of hospital, she must be much more tired than he. And so
the petty quarrel continued, until the pair came opposite to the gate,
and I saw that she was a sickly, blear-eyed, unlovely woman, and he
an unkempt, gipsy-like fellow with lean face and hungry, wolf's eyes.
Well, to cut a long story short, I had contrived to make their
acquaintance, and had found that, underneath their rags and dirt,
beat two honest and unselfish hearts. I had told them to come
always to me if in need of assistance of any sort—an invitation of
which they took advantage only once, and then when their straits
were desperate. On every other occasion I had found them touchily
independent, and though I sometimes bought flowers, bullrushes,
mistletoe, or fern-roots from them for the decoration of my house or
garden, they would not accept a farthing from me in the shape of
charity. If I wished to buy the wares they had for sale, that was
another matter; and even then I have reason to know that I got
more flowers, bullrushes, or fern roots for sixpence than their usual
customers got for a shilling.
For some twelvemonth we continued the best of friends. Then
suddenly their visits ceased, and I set eyes on neither again until I
saw the pair of them at Shadwell—the woman curled up in the
perambulator, and the man pushing it.
"Nash!" I called out, running after them. "Nash, where have you
been all this time? And why haven't you and Mrs. Nash been to see
me?"
"We have been doing pretty much the same as usual, sir," he replied
stiffly; "and thank you for asking." Then touching his ragged cap, he
said brusquely, "Good day, sir," and, pushing the perambulator
before him, passed along.
But I was not thus easily to be shaken off. At first he stood very
much on his dignity, answering my questions, in regard to himself
and his doings, with civil but manifest unwillingness, but at last I
contrived—and then only with difficulty—to discover wherein I had
offended.
On the last occasion, when they had visited me, I had said to him,
as he was passing out: "Well, good-bye, Nash. Mind, if ever you get
into trouble, be sure to come or to send to me, and I'll do my best to
get you out."
By "trouble" I had meant illness, or the inability to scrape together
the small sum they paid as rent for the miserable hovel in which
they lived.
But in George Nash's world "trouble"—so I learned for the first time
—has only one meaning when applied to a man (the word is used in
a different sense in regard to a woman), and that meaning—jail.
"I don't see why you should have thought that of us, sir," Nash said
with quiet dignity. "Poor we may be, but at least we've managed to
keep honest. And the inside of a prison we're never likely to see. We
thank you kindly for what you've done for us, sir, the missis and me,
but if you think as we're that sort, well, sir, we've made a mistake
about you, and you've made a mistake about us, and we wish you
good-day."
Turning doggedly to the perambulator, he touched his hat and
passed on.
"Why, my dear fellow," I said hotly, following him, and taking him by
the hand, "such a thought never entered my head. I'd leave you—
and for the matter of that I have left you or your wife—in my room
alone with every farthing I possess lying about openly, and never
even dream of counting it, or of thinking of it at all.
"Well," I went on, when I had at last persuaded him that he had
done me an injustice, "well, and what on earth is the meaning of
Mrs. Nash being cooped up in this perambulator? She looks very
white and thin. I do hope she isn't ill."
"Yes, sir; she's very ill," was the answer. "Got something wrong
inside her, the doctor said, that'll have to be cut out. I'm taking her
to Reading now."
"To Reading?" I said. "But why to Reading? I can easily arrange to
get her into a good hospital for women here."
"No, sir, thank you kindly. She's set on going to Reading, and
nowhere else. The doctor there (she's been there afore, you know)
don't treat poor folk as some other doctors do. They don't mean not
to be kind, but they speak so sharp, it frightens her. The doctor at
Reading—ah! he is different. She ain't a bit afraid of him. She won't
go anywhere but to Reading. She's set on it, sir, and so am I."
Knowing the man as I did, I could see that it was no use to argue
with him.
"I see," I said. "Quite right, George. I'll come with you as far as
Paddington, if you'll let me. Shadwell Station is some way yet. You
look hot and tired already, and so I'll take a turn at pushing the
pram while you rest. But if I may make a suggestion, I should say
that the best thing to do is to steer for the nearest place where
we're likely to find a four-wheeled cab and let me drive you to
Paddington. How did you propose taking Mrs. Nash there?"
"Same way as I'm taking her to Reading, sir," he said unconcernedly;
"in the pram, of course."
"The pram!" I ejaculated. "My dear Nash, what nonsense! It's forty
miles! You can't wheel a grown woman forty miles in a child's
perambulator."
"Can't I, sir?" he said, smiling with an air of superiority. "I've taken
her there twice before in the perambulator, and by picking up a bit
of work on the way we've managed nicely."
Then he looked at me queerly.
"Mr. Rissler," he said in a low voice, "will you take a word from a
man as you've been a friend to, and as'd like to prove himself a
friend of yours? I can't answer no question, and I didn't ought to say
what I'm going to say. I know as you're the poor man's friend,
though you are one of the gentry. But there's them as don't know it;
and, sir, believe me, there's trouble ahead for the likes o' you—bitter
trouble, bloody trouble. You take my word for it. And this is what I
want to say to you, sir. When the trouble comes, if you should find
yourself among enemies, if you should find yourself in danger o'
your life, as'll happen to many like you afore long, just you throw up
your left arm with your fist closed, and say, 'God and Napoleon and
the Dumpling strike with a granite arm!'"
He looked furtively around him as if afraid of being overheard by
eavesdroppers, and then repeated the sentence, "'God and Napoleon
and the Dumpling strike with a granite arm!' That's the word, sir. Do
you think you can remember it?"
Inwardly amused at the seriousness with which the foolish fellow
was taking the Dumpling's rhodomontade, but hiding my amusement
under a face portentously grave, lest I should give my well-meaning
friend offence, I replied:
"Yes, I can remember it, and I'll be sure to bear your words in mind,
if necessity comes. Thank you very much, Nash."
But to myself I said:
"A grown woman! Wheeled forty miles in a perambulator to undergo
an operation! And for no other reason than that the doctor at
Reading is kind and doesn't speak sharp to the poor! My God!"
 CHAPTER XXXI.
THE GREAT INSURRECTION BEGINS.
Looking back now upon the time of which I am writing, I cannot
altogether acquit myself of criminal negligence for failing to realise—
until it was too late to take action—how insidiously and how
thoroughly the Dumpling was doing his work. Nash's warning—
though I was by no means disposed to take it seriously—had not
been altogether a surprise to me, for I knew already that
inflammatory speeches were being delivered, inflammatory literature
circulated broadcast. To these I attached small importance, having
too much faith in the common sense and in the conservatism of my
fellow-countrymen to believe that the Dumpling could induce them
to take concerted action upon any considerable scale. I have since
learned that secret meetings were held nearly every night; but
instead of one mass meeting, which must inevitably have attracted
the attention of the police, the Dumpling, a prince of organisers, had
arranged for innumerable small gatherings in every part of London.
At each of these meetings some member of the General Council, and
therefore in close touch with the Dumpling himself, would preside,
and in this way their leader's plans were made known, a plan of
campaign laid down, and concerted action arranged in the most
secret yet thorough way. Immense sums of money, so I afterwards
learned, were expended in the purchase and in the secret storage of
arms; and foreign mercenaries and expert marksmen, whose
services the Dumpling had requisitioned, were constantly pouring
into London to place themselves at his orders.
Had I still been engaged in detective work, something of all this
must, I think, have come to my notice; but I am so constituted as to
be able to do one thing only at a time. Whatever pursuit I take up,
into that pursuit I throw myself heart and soul, to the exclusion of
everything else. This temperamental defect—if a defect it be—may
be the secret of some of my many failures; it may be the secret of
my few successes. Concentration of interests generally means
limitation of interests, and whether one be racking the heavens
nightly through a telescope in search of new worlds, or only peering
through a microscope, to isolate bacilli of this or that disease—one is
equally apt to become absent-minded in other matters. So entirely
had I given myself up to studying the problem of the poor, that I had
eyes for nothing else.
It is not, however, my intention further to describe, in these pages,
the harrowing scenes I witnessed while so occupied. Were I a
commissioner, appointed to report to a Committee of Inquiry upon
the condition of the poor, I should, it is true, have painful, revolting,
and even incredible facts to recount. I could give chapter and verse
in proof of inconceivable infamy. I could give instances of men,
women, and even children living under circumstances more
degrading than could be found in any so-called savage race. I
should, in common honesty, be compelled to admit that by many of
these who are most in evidence, as in search of work—work is the
very last thing in the world that they really desire to find. Hymn-
bawling in the streets is the nearest most of them have ever come to
earning their bread by the sweat of their brow; a few hours
processioning and posing as unemployed, the hardest day's work
many of them ever did.
And yet, admitting all this, and speaking as one who has seen
something of the poor, and of their homes, I say, and in all sincerity,
of the very poor as a whole, that I find it hard to express my
admiration, my respect, and my reverence for the unselfishness, the
courage, and the nobility which I have known them to display.
There came a time at last when the strenuousness of the work I was
doing began to tell terribly upon me. It has been said that profound
sympathies are always in association with keen sensibilities, and that
keen sensibilities expose their possessor to a depth of anguish
utterly unintelligible to those who are differently constituted. In my
own case the hopelessness of the struggle in which I was engaged
weighed constantly upon me. Men and women—and, worst of all,
little children—were starving literally by the thousand, and all my
efforts could do no more than bring relief each day to perhaps a
dozen. My money was gone; the health by which I could earn more
money was fast giving way, and what I had accomplished, and could
hope to accomplish, seemed, when compared with what remained to
be done, like the taking of a drop of water from the sea.
One evening, after visiting a case of destitution and misery so
harrowing that it was only by a tremendous effort I was able to
control myself, and to speak cheerfully and hopefully to the
sufferers, I came out into the dark street, and, once alone, to my
unspeakable disgust and dismay, burst into tears.
When a man of strong physique, normal by nature, and in no sense
hysterical, gets into a condition so over-wrought as this, he is on the
verge of a nervous breakdown, as was proved in my own person.
The next morning I was so seriously ill that I was compelled to keep
my bed for some days, after which I was ordered to Brighton.

I returned to my chambers in Adelphi Terrace on Sunday morning, to

find London dull to the very point of stagnation. So few people I do
not ever remember, even on a Sunday, to have seen in the streets.
The Strand was for once justified of its name, for it was like some
sea strand or beach from which the tide has withdrawn, leaving a
long stretch of untenanted sands. Yet that same night the returning
up-gathered tidal wave swept and broke over London like a
devastating sea.
A meeting of the unemployed was, I saw by the papers, to be held
that afternoon in Victoria Park. Instead of the few thousands whom
the public had expected to assemble, the meeting—no doubt by pre-
concerted action—mustered nearly half a million. The organisation of
this unwieldy mob was wonderful.
At five minutes to three the people were sullenly, suspiciously silent.
At three a bugle-call was heard, and suddenly in their midst the red
flag was raised, and at that sight the hounds of insurrection gave
tongue, baying for blood, in one prolonged and awful roar, that
might have been heard a mile away.
Then the Dumpling, who was standing on the pedestal of a drinking
fountain, raised his hand, and again the bugle-call rang out. Another
roar burst forth as, in every part of the open space of the park, poles
were set up, on the top of which were immense black boards, each
with a letter of the alphabet printed prominently upon it in white.
With one accord the multitude broke up to sort itself into huge
companies, according to the letter of the alphabet under which each
had been previously instructed to place himself. Another bugle-call,
and from hundreds of houses, surrounding the park, companies of
men were seen to come forth, carrying weapons for distribution. For
months this secret storage of weapons in private houses had been
going on all over London, with the result that when the outbreak
came, every man knew where to obtain a weapon. In the
neighbourhood of the park itself it was, of course, not possible to
store a sufficient number of rifles to arm more than some few
thousands of men; but those so armed had been more or less drilled
and trained to shoot. They were placed in the van of the Labour
army with the foreign mercenaries—expert riflemen—immediately in
the forefront, ready to resist the first attacking force, whether of the
military or of the police.
Then, at a signal from the Dumpling, the bugle was once again
sounded, and the westward march commenced by a route so
planned that weapons for those still unarmed could be picked up at
the various storage centres on the way.
The rebel army, more than half a million strong, and led by its new
Napoleon, was marching on London. The Revolution of which the
Dumpling had so often boasted had begun.
 CHAPTER XXXII.
BLOODSHED.
When, at the head of his Labour legions, the Dumpling set forth on
that eventful Sunday afternoon to march westwards, he and his
lieutenants—and, by means of his lieutenants, his men—knew
exactly what work lay in front of them. That work done, the order
had gone forth that the multitude was to break up, and each man
was to return quietly to his home, so that next day, at the hour
appointed, the legions might re-assemble, rested and ready for the
fray. It was known that the Dumpling had told off a certain number
of picked and tried men to patrol the streets after the hour of
dismissal, and through the night, and that these patrols would deal
summarily with all who disobeyed orders.
The man with a genius for organisation is rare. The man who has
not only a genius for organisation, but has also a genius for
compelling other people implicitly to accept his orders, and to abide
by his organisation —the man who can, at will, mould mankind in
the lump, as the potter moulds clay—comes only once in a hundred
years. Such a man was the Dumpling. Instead of letting his vast
army straggle invertebrately westwards under his single command,
his system of subdividing it into companies, each under a separate
picked leader, taking orders from himself, worked out with
surprisingly successful results. There was no aimless moving from
place to place. Though East London was practically in the hands of
the rioters, orders had been given that no man was to leave the
ranks, and that London Fields was to be the first halting-place, each
company to march thither by a pre-arranged route. At every open
space where two or more roads met—as, for instance, Hackney
Triangle—and where assailing forces of police or military might be
expected, expert riflemen, sent on slightly in advance of the main
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookgate.com