training_lending_club.R

# only evaluate w/ vintages that have come to term
df.term <- subset(df, year_issued < 2012)
df.term$home_ownership <- factor(df.term$home_ownership)
df.term$is_rent <- df.term$home_ownership=="RENT"
df.term$fico_range <- factor(df.term$fico_range)
df.term$fico_ordered <- as.numeric(df.term$fico_range)

idx <- runif(nrow(df.term)) > 0.75
train <- df.term[idx==FALSE,]
test <- df.term[idx==TRUE,]


rf <- randomForest(factor(is_bad) ~ last_fico_range_high + last_fico_range_low +
                     pub_rec_bankruptcies + revol_util + inq_last_6mths + is_rent,
                   type="classification", data=train, importance=TRUE, na.action=na.omit)