Practical Predictive Analytics

Variable importance

For classification targets you can use the random forest algorithm to determine variable importance.

For this example, a simulated sample was generated, with smoking and family history being key factors in determining heart disease among males:

set.seed(1020) #construct a 50/50 sample of Males, and Females gender <- sample(c("M","F"), 100, replace=T,prob=c(0.50,0.50))  #assign a higher probability of smoking to the Males (95%, WAY to high!) smokes <- ifelse(gender=="M",                  sample(c("N","Y"), 100, replace=T,prob=c(0.05,0.95)),                  sample(c("N","Y"), 100, replace=T,prob=c(0.45,0.55))                   )  #assume they also have a 60% chance of family history of heart disease  familyhistory <- ifelse(gender=="M",  sample(c("N","Y"), 100, replace=T,prob=c(0.40,0.60)), ...