In Patrikios/liftscores: Calculate and plot uplift from a binary classificatin models

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(liftscores)
library(data.table)
library(ranger)
library(caret)

data <- fread("https://raw.githubusercontent.com/just4jin/bank-marketing-prediction/master/data/bank_full.csv")
inx <- createDataPartition(data$y, list = F, p = 0.7)
train <- data[inx,]
test <- data[-inx,]
set.seed(23)
model_ranger <- ranger(as.factor(y) ~ ., data = train, probability = T)
pred <- data.table(event = test$y,
                   pred = predictions(predict(model_ranger, test))[, "yes"]
)

calculate_lift(
  dataWithProbabilityPrediction = pred, 
  levelPositive = "yes",
  responseVariable = "event", 
  probabilityOfChurning = "pred"

)

Patrikios/liftscores documentation built on Dec. 18, 2021, 6:42 a.m.