Preface 1 What is Data Mining? 1.1 Big Data 1.1.1 The Data Warehouse 1.1.2 Timelines 1.2 Types of Data-Mining Problems 1.3 The Pedigree of Data Mining 1.3.1 Databases 1.3.2 Statistics 1.3.3 Machine Learning 1.4 Is Big Better? 1.4.1 Strong Statistical Evaluation 1.4.2 More Intensive Search 1.4.3 More Controlled Experiments 1.4.4 Is Big Necessary 1.5 The Tasks of Predictive Data Mining 1.5.1 Data Preparation 1.5.2 Data Reduction 1.5.3 Data Modeling and Prediction 1.5.4 Case and Solution Analyses 1.6 Data Mining: Art or Science 1.7 An Overview of the Book 1.8 Bibliographic and Historical Remarks
2 Statistical Evaluation for Big Data 2.1 The Idealized Model 2.1.1 Classical Statistical Comparison and Evaluation 2.2 It's Big but Is It Biased 2.2.1 Objective Versus Survey Data 2.2.2 Significance and Predictive Value 18.104.22.168 Too Many Comparisons? 2.3 Classical Types of Statistical Prediction 2.3.1 Predicting True-or-False: Classification 22.214.171.124 Error Rates 2.3.2 Forecasting Numbers: Regression 126.96.36.199 Distance Measures 2.4 Measuring Predictive Performance 2.4.1 Independent Testing 188.8.131.52 Random Training and Testing 184.108.40.206 How Accurate Is the Error Estimate? 220.127.116.11 Comparing Results for Error Measures 18.104.22.168 Ideal or Real-World Sampling? 22.214.171.124 Training and Testing from Different Time Periods 2.5 Too Much Searching and Testing? 2.6 Why Are Errors Made? 2.7 Bibliographic and Historical Remarks
3 Preparing the Data 3.1 A Standard Form 3.1.1 Standard Measurements 3.1.2 Goals 3.2 Data Transformations 3.2.1 Normalizations 3.2.2 Data Smoothing 3.2.3 Differences and Ratios 3.3 Missing Data 3.4 Time-Dependent Data 3.4.1 Time Series 3.4.2 Composing Features from Time Series 126.96.36.199 Current Values 188.8.131.52 Moving Averages 184.108.40.206 Trends 220.127.116.11 Seasonal Adjustments 3.5 Hybrid Time-Dependent Applications 3.5.1 Multivariate Time Series 3.5.2 Classification and Time Series 3.5.3 Standard Cases and Time-Series Attributes 3.6 Text Mining 3.7 Bibliographic and Historical Remarks
4 Data Reduction 4.1 Selecting the Best Features 4.2 Feature Selection from Means and Variances 4.2.1 Independent Features 4.2.2 Distance-Based Optimal Feature Selection 4.2.3 Heuristic Feature Selection 4.3 Principal Constraints 4.4 Feature Selection by Decision Trees 4.5 How Many Measured Values 4.5.1 Reducing and Smoothing Values 18.104.22.168 Rounding 22.214.171.124 K-Means Clustering 126.96.36.199 Class Entropy 4.6 How Many Cases? 4.6.1 A Single Sample 4.6.2 Incremental Samples 4.6.3 Average Samples 4.6.4 Specialized Case-Reduction Techniques 188.8.131.52 Sequential Sampling over Time 184.108.40.206 Strategic Sampling of Key Events 220.127.116.11 Adjusting Prevalence 4.7 Bibliographic and Historical Remarks
5 Looking for Solutions 5.1 Overview 5.2 Math Solutions 5.2.1 Linear Scoring 5.2.2 Nonlinear Scoring: Neural Nets 5.2.3 Advanced Statistical Methods 5.3 Distance Solutions 5.4 Logic Solutions 5.4.1 Decision Trees 5.4.2 Decision Rules 5.5 What Do the Answers Mean? 5.5.1 Is It Safe to Edit Solutions? 5.6 Which Solution is Preferable? 5.7 Combining Different Answers 5.7.1 Multiple Prediction Methods 5.7.2 Multiple Samples 5.8 Bibliographic and Historical Remarks
6 What's Best for Data Reduction and Mining? 6.1 Let's Analyze Some Real Data 6.2 The Experimental Methods 6.3 The Empirical Results 6.3.1 Significance Testing 6.4 So What Did We Learn? 6.4.1 Feature Selection 6.4.2 Value Reduction 6.4.3 Subsampling or All Cases 6.5 Graphical Trend Analysis 6.5.1 Incremental Case Analysis 6.5.2 Incremental Complexity Analysis 6.6 Maximum Data Reduction 6.7 Are There Winners and Losers in Performance? 6.8 Getting the Best Results 6.9 Bibliogaphic and Historical Remarks
7 Art or Science? Case Studies in Data Mining 7.1 Why These Case Studies? 7.2 A Summary of Tasks for Predictive Data Mining 7.2.1 A Checklist for Data Preparation 7.2.2 A Checklist for Data Reduction 7.2.3 A Checklist for Data Modeling and Prediction 7.2.4 A Checklist for Case and Solution Analyses 7.3 The Case Studies 7.3.1 Transaction Processing 7.3.2 Text Mining 7.3.3 Outcomes Analysis 7.3.4 Process Control 7.3.5 Marketing and User Profiling 7.3.6 Exploratory Analysis 7.4 Looking Ahead 7.5 Bibliographic and Historical Remarks
Appendix: Data-Miner Software Kit
The potential business advantages of data mining are well documented in publications for executives and managers. However, developers implementing major data-mining systems need concrete information about the underlying technical principles—and their practical manifestations—in order to either integrate commercially available tools or write data-mining programs from scratch. This book is the first technical guide to provide a complete, generalized roadmap for developing data-mining applications, together with advice on performing these large-scale, open-ended analyses for real-world data warehouses.
Note: If you already own Predictive Data Mining: A Practical Guide, please see ISBN 1-55860-477-4 to order the accompanying software. To order the book/software package, please see ISBN 1-55860-478-2.
- Focuses on the preparation and organization of data and the development of an overall strategy for data mining.
- Reviews sophisticated prediction methods that search for patterns in big data.
- Describes how to accurately estimate future performance of proposed solutions.
- Illustrates the data-mining process and its potential pitfalls through real-life case studies.
- No. of pages:
- © Morgan Kaufmann 1997
- 8th December 1997
- Morgan Kaufmann
- eBook ISBN:
- Paperback ISBN:
"I enjoy reading PREDICTIVE DATA MINING. It presents an excellent perspective on the theory and practice of data mining. It can help educate statisticians to build alliances between statisticians and data miners." --Emanuel Parzen, Distinguished Professor of Statistics, Texas A&M University
Sholom M. Weiss is a professor of computer science at Rutgers University and the author of dozens of research papers on data mining and knowledge-based systems. He is a fellow of the American Association for Artificial Intelligence, serves on numerous editorial boards of scientific journals, and has consulted widely on the commercial application of advanced data mining techniques. He is the author, with Casimir Kulikowski, of Computer Systems That Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning, and Expert Systems, which is also available from Morgan Kaufmann Publishers.
Nitin Indurkhya is on the faculty at the Basser Department of Computer Science, University of Sydney, Australia. He has published extensively on Data Mining and Machine Learning and has considerable experience with industrial data-mining applications in Australia, Japan and the USA.