<<

. 131
( 137 .)



>>




Team-Fly®
Index 633


classification, 9

nearest neighbor techniques

combination function, 222

classification, 9

components of, 220“221

collaborative filtering

continuous values, features with,

estimated ratings, 284“285

235“237
grouping customers, 90

coverage of values, 232“233
predictions, 284“285

data preparation
profiles, building and comparing,

categorical values, 239“240
283“284

continuous values, 235“237

social information filtering, 282

decision trees, 199

word-of-mouth advertising, 283

discussed, 211

memory-based reasoning (MBR)

estimation tasks, 10, 215

case study, 259“262

feed-forward

challenges of, 262“265

back propagation, 228“232

classification codes, 266, 273“274

hidden layer, 227

combination function, 258, 265

input layer, 226

customer classification, 90“91

output layer, 227

customer response prediction, 258

generic algorithms and, 439“440

democracy approach, 279“281

hidden layers, 221, 227

distance function

historical data, 219

fraud detection, 258

history of, 212“213

free text responses, 258

implementation, 212

historical records, selecting,

inputs/outputs, 215

262“263

neighborliness parameters, 250

medical treatment applications, 258

nonlinear behaviors, 222

new customers, 277

OR value, 222

relevance feedback, 267“268

overfitting, 234

similarity measurements, 271“272

parallel coordinates, 253

training data, 263“264

prediction, 215

weighted voting, 281“282

real estate appraisal example,

negative correlation, 139

213“217

neighborliness parameters, neural

results, interpreting, 241“243

networks, 250

sensitivity analysis, 247“248

neural networks

sigmoid action functions, 225

activation function, 222

SOM (self-organizing map), 249“251

AND value, 222

time series analysis, 244“247

automation, 213

training sets, selection consideration,

average member technique, 252

232“234

bias sampling, 227

transfer function, 223

biological, 211

validation sets, 218

building models, 8

variable selection problem, 233

case study, 252“254

variance, 199

categorical variables, 239“240

634 Index


new customer information
Open Database Connectivity
gathering, 109“110
(ODBC), 496

memory-based reasoning, 277
operational errors, 159

profiles, building, 283
operational feedback, 485, 492

new start forecast (NSF), 469
operational summary data, OLAP, 477

nodes, graphs, 322
opportunistic sample, defined, 25

nonlinear behaviors, neural
opportunities, good response

networks, 222
scores, 34

non-response models, mass
optimization

mailings, 35
generic algorithms, 422

normal distribution, statistics, 130“132
resources, generic algorithms,

normalization, numeric variables, 550
433“435
normalized absolute value, distance
training as, 230

function, 275
OR value, neural networks, 222

NORMDIST function, 134
Oracle, relational database

NORMSINV function, 147
management software, 13

NSF (new start forecast), 469
order characteristics, market based

null hypothesis, statistics and, 125“126
analysis, 292

NULL values, missing data, 590
ordered lists, 239

numeric variables
ordered variables, measure of, 549

data correction, 73
organizations. See businesses

distance function, 275
out of time tests, 72

measure of, 550“551
outliners

splits, decision trees, 173
data correction, 73

data transformation, 74

O output layer, feed-forward neural

Occam™s Razor, 124“125 networks, 227

ODBC (Open Database outputs, neural networks, 215

Connectivity), 496
outsourcing data mining, 522“524

one-tailed distribution, 134
overfitting, neural networks, 234

Online Analytic Processing (OLAP)

P
additive facts, 501

data mining and, 507“508
parallel coordinates, neural
decision-support summary data,
networks, 253

477“478
parsing variables, 569

<<

. 131
( 137 .)



>>