I Components of the decision-making process 1

1 Business intelligence 3

1.1 Effective and timely decisions 3

1.2 Data, information and knowledge 6

1.3 The role of mathematical models 8

1.4 Business intelligence architectures 9

1.4.1 Cycle of a business intelligence analysis 11

1.4.2 Enabling factors in business intelligence projects 13

1.4.3 Development of a business intelligence system 14

1.5 Ethics and business intelligence 17

1.6 Notes and readings 18

2 Decision support systems 21

2.1 Definition of system 21

2.2 Representation of the decision-making process 23

2.2.1 Rationality and problem solving 24

2.2.2 The decision-making process 25

2.2.3 Types of decisions 29

2.2.4 Approaches to the decision-making process 33

2.3 Evolution of information systems 35

2.4 Definition of decision support system 36

2.5 Development of a decision support system 40

2.6 Notes and readings 43

3 Data warehousing 45

3.1 Definition of data warehouse 45

3.1.1 Data marts 49

3.1.2 Data quality 50

3.2 Data warehouse architecture 51

3.2.1 ETL tools 53

3.2.2 Metadata 54

3.3 Cubes and multidimensional analysis 55

3.3.1 Hierarchies of concepts and OLAP operations 60

3.3.2 Materialization of cubes of data 61

3.4 Notes and readings 62

II Mathematical Models and Methods 63

4 Mathematical models for decision making 65

4.1 Structure of mathematical models 65

4.2 Development of a model 67

4.3 Classes of models 70

4.4 Notes and readings 75

5 Data mining 77

5.1 Definition of data mining 77

5.1.1 Models and methods for data mining 79

5.1.2 Data mining, classical statistics and OLAP 80

5.1.3 Applications of data mining 81

5.2 Representation of input data 82

5.3 Data mining process 84

5.4 Analysis methodologies 90

5.5 Notes and readings 94

6 Data preparation 95

6.1 Data validation 95

6.1.1 Incomplete data 96

6.1.2 Data affected by noise 97

6.2 Data transformation 99

6.2.1 Standardization 99

6.2.2 Feature extraction 100

6.3 Data reduction 100

6.3.1 Sampling 101

6.3.2 Feature selection 102

6.3.3 Principal component analysis 104

6.3.4 Data discretization 109

7 Data exploration 113

7.1 Univariate analysis 113

7.1.1 Graphical analysis of categorical attributes 114

7.1.2 Graphical analysis of numerical attributes 116

7.1.3 Measures of central tendency for numerical attributes 118

7.1.4 Measures of dispersion for numerical attributes 121

7.1.5 Measures of relative location for numerical attributes 126

7.1.6 Identification of outliers for numerical attributes 127

7.1.7 Measures of heterogeneity for categorical attributes 129

7.1.8 Analysis of the empirical density 130

7.1.9 Summary statistics 135

7.2 Bivariate analysis 136

7.2.1 Graphical analysis 136

7.2.2 Measures of correlation for numerical attributes 142

7.2.3 Contingency tables for categorical attributes 145

7.3 Multivariate analysis 147

7.3.1 Graphical analysis 147

7.3.2 Measures of correlation for numerical attributes 149

7.4 Notes and readings 152

8 Regression 153

8.1 Structure of regression models 153

8.2 Simple linear regression 156

8.2.1 Calculating the regression line 158

8.3 Multiple linear regression 161

8.3.1 Calculating the regression coefficients 162

8.3.2 Assumptions on the residuals 163

8.3.3 Treatment of categorical predictive attributes 166

8.3.4 Ridge regression 167

8.3.5 Generalized linear regression 168

8.4 Validation of regression models 168

8.4.1 Normality and independence of the residuals 169

8.4.2 Significance of the coefficients 172

8.4.3 Analysis of variance 174

8.4.4 Coefficient of determination 175

8.4.5 Coefficient of linear correlation 176

8.4.6 Multicollinearity of the independent variables 177

8.4.7 Confidence and prediction limits 178

8.5 Selection of predictive variables 179

8.5.1 Example of development of a regression model 180

8.6 Notes and readings 185

9 Time series 187

9.1 Definition of time series 187

9.1.1 Index numbers 190

9.2 Evaluating time series models 192

9.2.1 Distortion measures 192

9.2.2 Dispersion measures 193

9.2.3 Tracking signal 194

9.3 Analysis of the components of time series 195

9.3.1 Moving average 196

9.3.2 Decomposition of a time series 198

9.4 Exponential smoothing models 203

9.4.1 Simple exponential smoothing 203

9.4.2 Exponential smoothing with trend adjustment 204

9.4.3 Exponential smoothing with trend and seasonality 206

9.4.4 Simple adaptive exponential smoothing 207

9.4.5 Exponential smoothing with damped trend 208

9.4.6 Initial values for exponential smoothing models 209

9.4.7 Removal of trend and seasonality 209

9.5 Autoregressive models 210

9.5.1 Moving average models 212

9.5.2 Autoregressive moving average models 212

9.5.3 Autoregressive integrated moving average models 212

9.5.4 Identification of autoregressive models 213

9.6 Combination of predictive models 216

9.7 The forecasting process 217

9.7.1 Characteristics of the forecasting process 217

9.7.2 Selection of a forecasting method 219

9.8 Notes and readings 219

10 Classification 221

10.1 Classification problems 221

10.1.1 Taxonomy of classification models 224

10.2 Evaluation of classification models 226

10.2.1 Holdout method 228

10.2.2 Repeated random sampling 228

10.2.3 Cross-validation 229

10.2.4 Confusion matrices 230

10.2.5 ROC curve charts 233

10.2.6 Cumulative gain and lift charts 234

10.3 Classification trees 236

10.3.1 Splitting rules 240

10.3.2 Univariate splitting criteria 243

10.3.3 Example of development of a classification tree 246

10.3.4 Stopping criteria and pruning rules 250

10.4 Bayesian methods 251

10.4.1 Naive Bayesian classifiers 252

10.4.2 Example of naive Bayes classifier 253

10.4.3 Bayesian networks 256

10.5 Logistic regression 257

10.6 Neural networks 259

10.6.1 The Rosenblatt perceptron 259

10.6.2 Multi-level feed-forward networks 260

10.7 Support vector machines 262

10.7.1 Structural risk minimization 262

10.7.2 Maximal margin hyperplane for linear separation 266

10.7.3 Nonlinear separation 270

10.8 Notes and readings 275

11 Association rules 277

11.1 Motivation and structure of association rules 277

11.2 Single-dimension association rules 281

11.3 Apriori algorithm 284

11.3.1 Generation of frequent itemsets 284

11.3.2 Generation of strong rules 285

11.4 General association rules 288

11.5 Notes and readings 290

12 Clustering 293

12.1 Clustering methods 293

12.1.1 Taxonomy of clustering methods 294

12.1.2 Affinity measures 296

12.2 Partition methods 302

12.2.1 K-means algorithm 302

12.2.2 K-medoids algorithm 305

12.3 Hierarchical methods 307

12.3.1 Agglomerative hierarchical methods 308

12.3.2 Divisive hierarchical methods 310

12.4 Evaluation of clustering models 312

12.5 Notes and readings 315

III Business Intelligence Applications 317

13 Marketing models 319

13.1 Relational marketing 320

13.1.1 Motivations and objectives 320

13.1.2 An environment for relational marketing analysis 327

13.1.3 Lifetime value 329

13.1.4 The effect of latency in predictive models 332

13.1.5 Acquisition 333

13.1.6 Retention 334

13.1.7 Cross-selling and up-selling 335

13.1.8 Market basket analysis 335

13.1.9 Web mining 336

13.2 Salesforce management 338

13.2.1 Decision processes in salesforce management 339

13.2.2 Models for salesforce management 342

13.2.3 Response functions 343

13.2.4 Sales territory design 346

13.2.5 Calls and product presentations planning 347

13.3 Business case studies 352

13.3.1 Retention in telecommunications 352

13.3.2 Acquisition in the automotive industry 354

13.3.3 Cross-selling in the retail industry 358

13.4 Notes and readings 360

14 Logistic and production models 361

14.1 Supply chain optimization 362

14.2 Optimization models for logistics planning 364

14.2.1 Tactical planning 364

14.2.2 Extra capacity 365

14.2.3 Multiple resources 366

14.2.4 Backlogging 366

14.2.5 Minimum lots and fixed costs 369

14.2.6 Bill of materials 370

14.2.7 Multiple plants 371

14.3 Revenue management systems 372

14.3.1 Decision processes in revenue management 373

14.4 Business case studies 376

14.4.1 Logistics planning in the food industry 376

14.4.2 Logistics planning in the packaging industry 383

14.5 Notes and readings 384

15 Data envelopment analysis 385

15.1 Efficiency measures 386

15.2 Efficient frontier 386

15.3 The CCR model 390

15.3.1 Definition of target objectives 392

15.3.2 Peer groups 393

15.4 Identification of good operating practices 394

15.4.1 Cross-efficiency analysis 394

15.4.2 Virtual inputs and virtual outputs 395

15.4.3 Weight restrictions 396

15.5 Other models 396

15.6 Notes and readings 397

Appendix A Software tools 399

Appendix B Dataset repositories 401

References 403

Index 413