EDA - House Prices

Goal: predict SalePrice

Load Packages

pacman::p_load(ggplot2, data.table, gridExtra)

Load Data

hpData = fread("https://raw.githubusercontent.com/augustrobo/miscellaneous/master/kaggle-house-prices/datasets/train.csv")

Visualization

Categorical

• Categorical features vs SalePrice: boxplot

# categorical features
catVars1 = names(hpData)[sapply(hpData, is.character)]
catVars2 = c("MSSubClass", "OverallQual", "OverallCond", 
             "YearBuilt", "YearRemodAdd", "GarageYrBlt",
             "MoSold", "YrSold")
catVars = c(catVars1, catVars2)


# plot categorical features vs SalePrice
catPlot = function(dt, vars, y){
  dt[, (vars):=lapply(.SD, as.factor), .SDcols = vars]
  p = list()
  i = 1
  for(x in vars){
    p[[i]] = ggplot(dt, aes_string(x, y)) + geom_boxplot()
    i = i + 1
  }
  do.call(grid.arrange, c(p, ncol = 3))
}

catPlot(hpData, catVars, "SalePrice")

Numerical

• Numerical features vs SalePrice: scatterplot

# numerical features
colnames(hpData)[names(hpData) == "1stFlrSF"] = "FirstFlrSF"
colnames(hpData)[names(hpData) == "2ndFlrSF"] = "SecondFlrSF"
colnames(hpData)[names(hpData) == "3SsnPorch"] = "ThreeSsnPorch"
numVars = setdiff(names(hpData), c(catVars, "Id", "SalePrice"))

numPlot = function(dt, vars, y){
  p = list()
  i = 1
  for(x in vars){
    p[[i]] = ggplot(dt, aes_string(x, y)) + geom_point() + geom_smooth()
    i = i + 1
  }
  do.call(grid.arrange, c(p, ncol = 3))
}
numPlot(hpData, numVars, "SalePrice")