Corrplot -For Correlation matrix plot

  1. Use library corrplot to draw the correlation matrix
  2. Take out the ID column and the medv column (as this is the target/ dependent- variable)

The objective here is to see if Indepemdent varaibles have correlation between each other, this can lead to a problem called multicolinearity, which gives rise to

-wrong extimation of coeffcients
-overfitting

library(data.table)

Data=read.csv("BostonHousing.csv")


suppressPackageStartupMessages(library(corrplot))
correlations <- cor(Data[,-c(1,15)]) # ID and medv  


corrplot(correlations, order = "hclust")

Plot corrplot to see results after

suppressPackageStartupMessages(library(caret))
highCorr <- findCorrelation(correlations, cutoff = 0.50) # Cut off can be higher too like 0.7 
      

reduced_Data = Data[,-c(highCorr)]


corMatfterFiler <- cor(reduced_Data) 


###After the independent variables with high correlation is removed
corrplot(corMatfterFiler, order = "hclust")

See the data set after dropping varaibles with high correlation - using DT

library(DT)


datatable(reduced_Data)