[Create base version of Information Value calculations] #R

information value.R

## Implement Information Value in base?

iris_iv <-
  iris %>%
  mutate(IsSetosa = ifelse(Species == "setosa", 1, 0))

## Information
infoTables <- 
  Information::create_infotables(
    data = iris_iv,
    y = "IsSetosa",
    bins = 10,
    parallel = FALSE
)

# WOE Table

infoTables$Tables$Sepal.Length

# Sepal.Length  N    Percent       WOE        IV
# 1     [4.3,4.7] 11 0.07333333  0.000000 0.0000000
# 2     [4.8,4.9] 11 0.07333333  2.197225 0.3515559
# 3       [5,5.1] 19 0.12666667  2.367124 1.0380218
# 4     [5.2,5.5] 18 0.12000000  1.145132 1.2097916
# 5     [5.6,5.7] 14 0.09333333 -1.098612 1.2976806
# 6       [5.8,6] 16 0.10666667 -2.014903 1.5596180
# 7     [6.1,6.2] 10 0.06666667  0.000000 1.5596180
# 8     [6.3,6.4] 16 0.10666667  0.000000 1.5596180
# 9     [6.5,6.8] 18 0.12000000  0.000000 1.5596180
# 10    [6.9,7.9] 17 0.11333333  0.000000 1.5596180


## Self implementation

compute_quantile <- function(variable, bins){
  
  quantile(variable,
           probs = c(1:(bins - 1) / bins),
           na.rm = TRUE,
           type = 3)
  }

q <- compute_quantile(iris_iv$Sepal.Length, bins = 10)
cuts <- unique(q)

agg_var <- function(predictor, outcome, cuts){
  
  cut_table <-
    table(
    findInterval(
      predictor,
      vec = cuts,
      rightmost.closed = FALSE),
    outcome) %>%
    as.data.frame.matrix()
  
  data.frame(
    "N" = rowSums(cut_table),
    "WOE" = log((cut_table$`1`*sum(cut_table$`0`))/
                  (cut_table$`0`*sum(cut_table$`1`)))
  )
  
  
  
}

agg_var(predictor = iris_iv$Sepal.Length, 
        cuts = cuts,
        outcome = iris_iv$IsSetosa)