Scripts for Criteo Challenge

hist.R

setwd('~/kaggle/criteo')

library(ggplot2)
library(scales)

probs <- read.csv('submit/submit-8.csv')

num_less <- length(which(probs$Predicted < 0.5))
num_more <- length(which(probs$Predicted > 0.5))
num_less / (num_less + num_more)
num_more / (num_less + num_more)

plot.histogram <- function(data, title) {
  ggplot(data, aes(x=Predicted)) +
    ggtitle(title) +
    geom_histogram(binwidth=1/20) +
    theme(plot.title = element_text(size=17, vjust=1)) +
    xlab("Predicted Probability") +
    ylab("Frequency") +
    scale_x_continuous(breaks=c(0, 0.25, 0.5, 0.75, 1)) +
    scale_y_continuous(labels=comma)  
}

plot.histogram(probs, "Histogram of a Submit That Scored 0.47971")

ideal.probs <- probs
ideal.probs$Predicted[probs$Predicted > 0.5] <- 1.5 - probs$Predicted[probs$Predicted > 0.5]

plot.histogram(ideal.probs, "Ideal Histogram?")