calculate_lift: calculate_lift

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

Description

Provides an overview of the binary classification (Yes, No style) model performance and plots the lift plot

Usage

calculate_lift(
  dataWithProbabilityPrediction = pred,
  levelPositive = "yes",
  responseVariable = "y",
  probabilityOfChurning = "pred"
)

Arguments

dataWithProbabilityPrediction	A data set object with predictions (probabilities)
levelPositive	Character, positive categorical true value of the binary outcome variable (e.g., 'Y', 'yes' or 'churned')
responseVariable	Character column indicating the response variable (e.g., 'churn')
probabilityOfChurning	Character column indicating the column, where the prediction probabilities are saves (e.g., 'predictions')

Value

A data frame with the Lift values as compared to the base line performance

Examples

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