<<

. 123
( 137 .)



>>

descriptive models, 78

divisive clustering, 371“372

directed models, 78“79

evaluation, 372“373

estimators, 79“81

Gaussian mixture model, 366“367

association rules

geometric distance, 360“361

actionable rules, 296

hard clustering, 367

affinity grouping, 11

Hertzsprung-Russell diagram,

anonymous versus identified

352“354

transactions, 308

K-means algorithm, 354“358

data quality, 308

luminosity, 351

dissociation rules, 317

natural association, 358

effectiveness of, 299“301

scaling, 363“364

inexplicable rules, 297“298

single linkage, 369

point-of-sale data, 288

soft clustering, 367

practical limits, overcoming, 311“313

SOM (self-organizing map), 372

prediction, 70

vectors, angles between, 361“362

probabilities, calculating, 309

weighting, 363“365

products, hierarchical categories, 305

zone boundaries, adjusting, 380

618 Index


auxiliary information, 569“571
neural networks, 227

availability of data, determining,
response, methods of, 146

515“516
untruthful learning sources, 46“47

average member technique, neural
BILL_MASTER file, customer
networks, 252
signatures, 559

averages, estimation, 81
binary churn models, 119

binary classification

B decision trees, 168

back propagation, feed-forward misclassification rates, 98

neural networks, 228“232
binary data, 557

backfitting, defined, 170
binning, 237, 551

bad customers, customer relationship
binomial formula (Jacques

management, 18
Bernoulli), 191

bad data formats, data
biological neural networks, 211

transformation, 28
births, house-hold level data, 96

balance transfer programs, industry
bizocity scores, 112“113

revolution, 18
Bonferroni, Carlo (Bonferroni™s

balanced datasets, model sets, 68
correction), 149

balanced sampling, 68
box diagrams, as alternative to

bathtub hazards, 397“398
decision trees, 199“201

behaviors
brainstorming meetings, 37

behavioral segments, marketing
branching nodes, decision trees, 176

campaigns, 111“113
budgets, fixed, marketing campaigns,

behavior-based variables
97“100

ad hoc questions, 585
building models, data mining, 8, 77

aggression, 18
Building the Data Warehouse (Bill

convenience users, 580, 587“589
Inmon), 474

declining usage, 577“579
Business Modeling and Data Mining
estimated revenue, segmenting,
(Dorian Pyle), 60

581“583
businesses
ideals, comparisons to, 585“587
challenges of, identifying, 23“24
potential revenue, 583“585
customer relationship
purchasing frequency, 575“576
management, 2“6

revolvers, 580
customer-centric, 514“515

transactions, 580
forward-looking, 2

future customer behaviors, home-based, 56

predicting, 10
large-business relationships, 3“4

bell-shaped distribution, 132
opportunities, identifying

benefit, point of maximum, 101
virtuous cycle, 27“28
Bernoulli, Jacques (binomial
wireless communication industries,
formula), 191
34“35

biased sampling
product-focused, 2

confidence intervals, statistical
recommendation-based, 16“17

analysis, 146
small-business relationships, 2

Index 619


car ownership, house-hold level data,

C
96

calculations, probabilities, 133“135

CART (Classification and Regression
call detail databases, 37

Trees) algorithm, decision trees,
call-center records, useful data

185, 188“189
sources, 60

case studies

campaigns, marketing. See also

automatic cluster detection, 374“378

advertising

chi-square tests, 155“158

acquisitions-time data, 108“110

decision trees, 206, 208

canonical measurements, 31

generic algorithms, 440“443

champion-challenger approach, 139

link analysis, 343“346

credit risks, reducing exposure to,

MBR (memory-based reasoning),

113“114

259“262

cross-selling, 115“116

neural networks, 252“254

customer response, tracking, 109

catalogs

customer segmentation, 111“113

response models, decision trees

differential response analysis,

for, 175

107“108

retailers, historical customer

discussed, 95

behavior data, 5

fixed budgets, 97“100

categorical variables

loyalty programs, 111

<<

. 123
( 137 .)



>>