tdmModVote2Target: Analyze how the vote fraction corresponds to reliability of...

In TDMR: Tuned Data Mining in R

Description

This function analyzes whether in different vote bins the trained RF makes predictions with different reliability. Only for RF-prediction in case of binary (0/1) classification.

Expected result: The larger the fraction of trees voting for class 0 is, the smaller is the percentage of true class-1- cases in this vote bin. This function is somewhat specialized for the DMC2010-task.

Usage

tdmModVote2Target(vote0, pred, target)

Arguments

vote0	vector: which fraction of trees votes for class 0?
pred	vector: the predicted class for each record (0/1)
target	vector: the true class for each vector (0/1)

Value

a data frame with columns

vcut	vote cut v
count	number of cases with vote fraction in [v[i-1],v[i]]
pred0	fraction of 0-predictions
pCorr	fraction of correct predictions
pR	fraction of true 1-cases

Author(s)

Wolfgang Konen [e-mail]

TDMR documentation built on March 3, 2020, 1:06 a.m.