housing_linearregression.R

# pricing prediction with linear regression
# we use linear regressions to predict the price of house
# then use RMSE to evaluate the model

library(arimo)

housing_ddf = arimo.getDDF('housing')
housing = head(ddf, nrow(ddf))

splitdf <- function(dataframe, ratio=0.8, seed=NULL) {
  if (!is.null(seed)) set.seed(seed)
  index <- 1:nrow(dataframe)
  trainindex = sample(1:nrow(dataframe), size=ratio*nrow(dataframe))
  trainset <- dataframe[trainindex, ]
  testset <- dataframe[-trainindex, ]
  list(train=trainset,test=testset)
}

rmse = function(y0, y1) {
  sqrt(mean((y0 - y1) ^ 2))
}

split = splitdf(housing)

fit = lm(price ~ lotsize + bedrooms + bathrms + stories, data=split$train)
summary(fit)

fitted(fit) # predicted values
residuals(fit) # residuals (actual - predicted)

predictedPrice = predict(fit, split$test)
rmse(split$test$price, predictedPrice)

# r-squared
rq = function(y0, y1) {
  y_mean = mean(y0)
  ss_tot = sum((y0 - y_mean) ^ 2)
  ss_res = sum((y0 - y1) ^ 2)
  
  1 - (ss_res / ss_tot)
}
rq(split$test$price, predictedPrice)