Ch. 7 Decision analytic thinking models

classifiers

accuracy= (# of correct decisions made/total # of decisions made)

negative=good, positive=bad

confusion matrix- separates out decisions made by the classifier

accuracy reduces classifier performance to a single number and is easy to measure

class confusion

true classes denoted by positive and negative, classes predicted by the model are true, false

makes no distinction between false positive and false negative erros

generalizing beyond classifiers

expected value

what is important in this application? what is the goal?

decomposes data analytic thinking into structured of problem

elements of the analysis can be extracted from data

elements of analysis need to be acquired from other sources

value calculation= p(o1)v(o1)+p(o2)v(o2)+p(o3)*v(o3)

p(oi)=probability v(oi)=value

expected benefit of targeting= Pr(X)Vr+[1-Pr(X)]Vnr

in aggregate-how well each model do: what is the expected value

(Y,N) divided by totoal number of test consumers

error rates

p(h,a)=count(h,a)/T

no difference between a cost and a benefit except for the sign. we express all values as benefits, with costs being negative benefits

false positive- classify a consumer as a likely responder and target her, but she doesn't respond

click to edit

false negative-consumer who was predicted not to be a likely responder but would have bought if offered

true positive-consumer who is offered the product and buys it

true negative-consumer who was not offered the deal and would not have bought it even if it had been offered