Includes a data frame of 1994 US census income from 48,842 people
divided into a training set of 32,561 and an independent test set
of 16,281. The training outcome variable y (yt for test) is
binary and indicates whether or not a person’s income is greater
than $50,000 per year. There are 12 predictor variables x (xt
for test) consisting of various demographic and financial
properties associated with each person. It also included estimates
of \(Pr(y=1|x)\) obtained by several machine learning methods:
gradient boosting on logistic scale using maximum likelihood (GBL),
random forest (RF), and gradient boosting on the probability scale
(GBP) using least–squares.
censuscensus
A list of 10 items.
training data frame of 32561 observations on 12 predictor variables
training binary response whether salary is above $50K or not
test data frame of 16281 observations predictor variables
test binary response whether salary is above $50K or not
training GBL response variable
test GBL response variable
training GBP response variable
test GBP response variable
training RF response probabilities
test GBP response probabilities