glm_binomial_logit.R

# kaggle titanic 
# https://www.kaggle.com/c/titanic-gettingStarted/data


t <- read.csv("../../Downloads/train.csv")
names(t)<-tolower(names(t))

idx<- sample(nrow(t),(nrow(t)*0.7))
t.tr <- t[ idx,]
t.te <- t[-idx,]

table(t.tr$survived)/nrow(t.tr)
table(t.te$survived)/nrow(t.te)

head(t.tr)
g <- glm(data=t.tr, survived~pclass+sex+age+sibsp+parch)

p <- predict(g,t.te)
t.pred <- data.frame(t.te, p)
head(t.pred)
t.pred$p_factor <- cut(t.pred$p,breaks=c(-1,0.55,2))
table(t.pred$p_factor,t.pred$survived)
sum(diag(table(t.pred$p_factor,t.pred$survived)))/nrow(t.pred)