Data Mining for Business Intelligence: Concepts, Techniques, and Applications in Microsoft Office Excel^® with XLMiner^®

by Galit Shmueli (Univ. of Maryland, College Park ); Nitin R. Patel (Massachusetts Institute of Technology ); Peter C. Bruce (statistics.com )

• ISBN13: 9780470084854
• ISBN10: 0470084855
• Format: Hardcover
• Copyright: 2006-12-01
• Publisher: Wiley-Interscience

Summary

In today's world, businesses are becoming more capable of accessing their ideal consumers, and an understanding of data mining contributes to this success. Data Mining for Business Intelligence, which was developed from a course taught at the Massachusetts Institute of Technology's Sloan School of Management, and the University of Maryland's Smith School of Business, uses real data and actual cases to illustrate the applicability of data mining intelligence to the development of successful business models. Featuring XLMiner, the Microsoft Office Excel add-in, this book allows readers to follow along and implement algorithms at their own speed, with a minimal learning curve. In addition, students and practitioners of data mining techniques are presented with hands-on, business-oriented applications. An abundant amount of exercises and examples are provided to motivate learning and understanding. Data Mining for Business Intelligence: * Provides both a theoretical and practical understanding of the key methods of classification, prediction, reduction, exploration, and affinity analysis * Features a business decision-making context for these key methods * Illustrates the application and interpretation of these methods using real business cases and data This book helps readers understand the beneficial relationship that can be established between data mining and smart business practices, and is an excellent learning tool for creating valuable strategies and making wiser business decisions.

Author Biography

GALIT SHMUELI, PHD, is Assistant Professor of Statistics in the Decision and Information Technologies Department of the Robert H. Smith School of Business at the University of Maryland.

NITIN R. PATEL, PHD, is Chairman, Founder, and Chief Technology Officer of Cambridge-based Cytel Incorporated and a Visiting Professor in the Engineering Systems Division at the Massachusetts Institute of Technology.

PETER C. BRUCE is President and owner of statistics.com, the leading provider of professional development courses in statistics.

Table of Contents

Foreword	xiii
Preface	xv
Acknowledgments	xvii
Introduction	1	(8)
What Is Data Mining?	1	(1)
Where Is Data Mining Used?	2	(1)
The Origins of Data Mining	2	(1)
The Rapid Growth of Data Mining	3	(1)
Why Are There So Many Different Methods?	4	(1)
Terminology and Notation	4	(2)
Road Maps to This Book	6	(3)
Overview of the Data Mining Process	9	(26)
Introduction	9	(1)
Core Ideas in Data Mining	9	(2)
Supervised and Unsupervised Learning	11	(1)
The Steps in Data Mining	11	(2)
Preliminary Steps	13	(8)
Building a Model: Example with Linear Regression	21	(6)
Using Excel for Data Mining	27	(8)
Problems	31	(4)
Data Exploration and Dimension Reduction	35	(18)
Introduction	35	(1)
Practical Considerations	35	(2)
Example 1: House Prices in Boston	36	(1)
Data Summaries	37	(1)
Data Visualization	38	(2)
Correlation Analysis	40	(1)
Reducing the Number of Categories in Categorical Variables	41	(1)
Principal Components Analysis	41	(12)
Example 2: Breakfast Cereals	42	(3)
Principal Components	45	(1)
Normalizing the Data	46	(3)
Using Principal Components for Classification and Prediction	49	(2)
Problems	51	(2)
Evaluating Classification and Predictive Performance	53	(22)
Introduction	53	(1)
Judging Classification Performance	53	(19)
Accuracy Measures	53	(3)
Cutoff for Classification	56	(4)
Performance in Unequal Importance of Classes	60	(1)
Asymmetric Misclassification Costs	61	(5)
Oversampling and Asymmetric Costs	66	(6)
Classification Using a Triage Strategy	72	(1)
Evaluating Predictive Performance	72	(3)
Problems	74	(1)
Multiple Linear Regression	75	(16)
Introduction	75	(1)
Explanatory vs. Predictive Modeling	76	(1)
Estimating the Regression Equation and Prediction	76	(5)
Example: Predicting the Price of Used Toyota Corolla Automobiles	77	(4)
Variable Selection in Linear Regression	81	(10)
Reducing the Number of Predictors	81	(1)
How to Reduce the Number of Predictors	82	(4)
Problems	86	(5)
Three Simple Classification Methods	91	(20)
Introduction	91	(1)
Example 1: Predicting Fraudulent Financial Reporting	91	(1)
Example 2: Predicting Delayed Flights	92	(1)
The Naive Rule	92	(1)
Naive Bayes	93	(10)
Conditional Probabilities and Pivot Tables	94	(1)
A Practical Difficulty	94	(1)
A Solution: Naive Bayes	95	(5)
Advantages and Shortcomings of the naive Bayes Classifier	100	(3)
k-Nearest Neighbors	103	(8)
Example 3: Riding Mowers	104	(1)
Choosing k	105	(1)
k-NN for a Quantitative Response	106	(1)
Advantages and Shortcomings of k-NN Algorithms	106	(2)
Problems	108	(3)
Classification and Regression Trees	111	(26)
Introduction	111	(2)
Classification Trees	113	(1)
Recursive Partitioning	113	(1)
Example 1: Riding Mowers	113	(7)
Measures of Impurity	115	(5)
Evaluating the Performance of a Classification Tree	120	(1)
Example 2: Acceptance of Personal Loan	120	(1)
Avoiding Overfitting	121	(9)
Stopping Tree Growth: CHAID	121	(4)
Pruning the Tree	125	(5)
Classification Rules from Trees	130	(1)
Regression Trees	130	(2)
Prediction	130	(1)
Measuring Impurity	131	(1)
Evaluating Performance	132	(1)
Advantages, Weaknesses, and Extensions	132	(5)
Problems	134	(3)
Logistic Regression	137	(30)
Introduction	137	(1)
The Logistic Regression Model	138	(8)
Example: Acceptance of Personal Loan	139	(2)
Model with a Single Predictor	141	(2)
Estimating the Logistic Model from Data: Computing Parameter Estimates	143	(1)
Interpreting Results in Terms of Odds	144	(2)
Why Linear Regression Is Inappropriate for a Categorical Response	146	(2)
Evaluating Classification Performance	148	(2)
Variable Selection	148	(2)
Evaluating Goodness of Fit	150	(3)
Example of Complete Analysis: Predicting Delayed Flights	153	(7)
Data Preprocessing	154	(1)
Model Fitting and Estimation	155	(1)
Model Interpretation	155	(1)
Model Performance	155	(2)
Goodness of fit	157	(1)
Variable Selection	158	(2)
Logistic Regression for More Than Two Classes	160	(7)
Ordinal Classes	160	(1)
Nominal Classes	161	(2)
Problems	163	(4)
Neural Nets	167	(20)
Introduction	167	(1)
Concept and Structure of a Neural Network	168	(1)
Fitting a Network to Data	168	(13)
Example 1: Tiny Dataset	169	(1)
Computing Output of Nodes	170	(2)
Preprocessing the Data	172	(1)
Training the Model	172	(4)
Example 2: Classifying Accident Severity	176	(1)
Avoiding overfitting	177	(4)
Using the Output for Prediction and Classification	181	(1)
Required User Input	181	(1)
Exploring the Relationship Between Predictors and Response	182	(1)
Advantages and Weaknesses of Neural Networks	182	(5)
Problems	184	(3)
Discriminant Analysis	187	(16)
Introduction	187	(1)
Example 1: Riding Mowers	187	(1)
Example 2: Personal Loan Acceptance	188	(1)
Distance of an Observation from a Class	188	(3)
Fisher's Linear Classification Functions	191	(3)
Classification Performance of Discriminant Analysis	194	(1)
Prior Probabilities	195	(1)
Unequal Misclassification Costs	195	(1)
Classifying More Than Two Classes	196	(1)
Example 3: Medical Dispatch to Accident Scenes	196	(1)
Advantages and Weaknesses	197	(6)
Problems	200	(3)
Association Rules	203	(16)
Introduction	203	(1)
Discovering Association Rules in Transaction Databases	203	(1)
Example 1: Synthetic Data on Purchases of Phone Faceplates	204	(1)
Generating Candidate Rules	204	(2)
The Apriori Algorithm	205	(1)
Selecting Strong Rules	206	(6)
Support and Confidence	206	(1)
Lift Ratio	207	(1)
Data Format	207	(2)
The Process of Rule Selection	209	(1)
Interpreting the Results	210	(1)
Statistical Significance of Rules	211	(1)
Example 2: Rules for Similar Book Purchases	212	(1)
Summary	212	(7)
Problems	215	(4)
Cluster Analysis	219	(22)
Introduction	219	(1)
Example: Public Utilities	220	(2)
Measuring Distance Between Two Records	222	(5)
Euclidean Distance	223	(1)
Normalizing Numerical Measurements	223	(1)
Other Distance Measures for Numerical Data	223	(3)
Distance Measures for Categorical Data	226	(1)
Distance Measures for Mixed Data	226	(1)
Measuring Distance Between Two Clusters	227	(1)
Hierarchical (Agglomerative) Clustering	228	(5)
Minimum Distance (Single Linkage)	229	(1)
Maximum Distance (Complete Linkage)	229	(1)
Group Average (Average Linkage)	230	(1)
Dendrograms: Displaying Clustering Process and Results	230	(1)
Validating Clusters	231	(1)
Limitations of Hierarchical Clustering	232	(1)
Nonhierarchical Clustering: The k-Means Algorithm	233	(8)
Initial Partition into k Clusters	234	(3)
Problems	237	(4)
Cases	241	(30)
Charles Book Club	241	(9)
German Credit	250	(4)
Tayko Software Cataloger	254	(4)
Segmenting Consumers of Bath Soap	258	(4)
Direct-Mail Fundraising	262	(3)
Catalog Cross-Selling	265	(2)
Predicting Bankruptcy	267	(4)
References	271	(2)
Index	273