gini.R

rm(list=ls())

# Read data from disk

car_data = read.csv("../car_data.csv")
# Create tables with freqs for CID
absfreq = table(car_data[, c(1, 5)])
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

GINI_CUSTOMERS = numeric(20)
GINI_ID = 0
for (i in 1:20) {
    GINI_CUSTOMERS[i] = 1 - freq[i, 'No']^2 - freq[i, 'Yes']^2
    GINI_ID = GINI_ID + freqSum[i] * GINI_CUSTOMERS[i]
}

gini_process <-function(absfreq,splitvar = NULL){
  freq = prop.table(absfreq, 1)
  freqSum = rowSums(prop.table(absfreq))
  row_1 = rownames(freq)[1]
  row_2 = rownames(freq)[2]
  GINI_1 = 1 - freq[row_1, "No"]^2 - freq[row_1, "Yes"]^2
  GINI_1
  GINI_2 = 1 - freq[row_2, "No"]^2 - freq[row_2, "Yes"]^2
  GINI_2
  GINI = freqSum[row_1] * GINI_1  + freqSum[row_2] * GINI_2
  return (c(GINI_1, GINI_2, GINI))
}

# Create tables with frequencies for Sex
absfreq = table(car_data[, c(2, 5)])
list[GINI_MALE, GINI_FEMALE, GINI_SEX] = gini_process(absfreq)
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

# Calculate GINI index of Sex
GINI_Male = 1 - freq["M", "No"]^2 - freq["M", "Yes"]^2
GINI_Female = 1 - freq["F", "No"]^2 - freq["F", "Yes"]^2
GINI_Sex = freqSum["M"] * GINI_Male  + freqSum["F"] * GINI_Female

##########################
# Types: Sedan / Family / Sport
# Create tables with frequencies for CarType, multisplit
splita = car_data[,c(3,5)]
splita$CarType = as.character(splita$CarType)
splita$CarType[splita$CarType == 'Sedan'] <- 'FamilySedan'
splita$CarType[splita$CarType == 'Family'] <- 'FamilySedan'
absfreq = table(splita)
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

# Calculate GINI index when splitting Family-Sedan vs Sport
GINI_FamilySedan = 1 - freq["FamilySedan", "No"]^2 - freq["FamilySedan", "Yes"]^2
GINI_Sport = 1 - freq["Sport", "No"]^2 - freq["Sport", "Yes"]^2
GINI_SplitA = freqSum["FamilySedan"] * GINI_FamilySedan  + freqSum["Sport"] * GINI_Sport

############### Family-Sport vs Sedan

splitb = car_data[,c(3,5)]
splitb$CarType = as.character(splitb$CarType)
splitb$CarType[splitb$CarType == 'Family'] <- 'FamilySport'
splitb$CarType[splitb$CarType == 'Sport'] <- 'FamilySport'
absfreq = table(splitb)
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

# Calculate GINI index when splitting Family-Sedan vs Sport
GINI_FamilySport = 1 - freq["FamilySport", "No"]^2 - freq["FamilySport", "Yes"]^2
GINI_Sedan = 1 - freq["Sedan", "No"]^2 - freq["Sedan", "Yes"]^2
GINI_SplitB = freqSum["FamilySport"] * GINI_FamilySport  + freqSum["Sedan"] * GINI_Sedan

### SPORT SEDAN VS FAMILY
splitc = car_data[,c(3,5)]
splitc$CarType = as.character(splitc$CarType)
splitc$CarType[splitc$CarType == 'Sport'] <- 'SportSedan'
splitc$CarType[splitc$CarType == 'Sedan'] <- 'SportSedan'
absfreq = table(splitc)
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

# Calculate GINI index when splitting Family-Sedan vs Sport
GINI_SportSedan = 1 - freq["SportSedan", "No"]^2 - freq["SportSedan", "Yes"]^2
GINI_Family = 1 - freq["Family", "No"]^2 - freq["Family", "Yes"]^2
GINI_SplitC = freqSum["SportSedan"] * GINI_SportSedan  + freqSum["Family"] * GINI_Family

##################### Budget

############### Family-Sport vs Sedan

splitb = car_data[,c(3,5)]
splitb$CarType = as.character(splitb$CarType)
splitb$CarType[splitb$CarType == 'Family'] <- 'FamilySport'
splitb$CarType[splitb$CarType == 'Sport'] <- 'FamilySport'
absfreq = table(splitb)
freq = prop.table(absfreq, 1)
freqSum = rowSums(prop.table(absfreq))

# Calculate GINI index when splitting Family-Sedan vs Sport
GINI_FamilySport = 1 - freq["FamilySport", "No"]^2 - freq["FamilySport", "Yes"]^2
GINI_Sedan = 1 - freq["Sedan", "No"]^2 - freq["Sedan", "Yes"]^2
GINI_SplitB = freqSum["FamilySport"] * GINI_FamilySport  + freqSum["Sedan"] * GINI_Sedan