infaddil/krr

Implement Kernel Ridge Regression in R with Advertising.csv dataset.

Kernel Ridge Regression

---

1. Load and split Advertising dataset

> ad <- read.csv("Advertising.csv")
> X <- ad[,2:4]
> Y <- ad[,5]
> set.seed(0506)
> train <- sample(1:nrow(X),2*nrow(X)/3)
> trainX <- X[train,]
> trainY <- Y[train,]
> testX <- X[-train,]
> testY <- Y[-train,]

2. fit KRR using polynomial

get the testMSE = 10.45449, with d=3, lambda=0.01

> method = 2
> d = 3
> lambda = 0.01
> predY1 <- krr(method, d, trainX, trainY, testX, lambda)
> MSE1 <- mean((predY1-testY)^2)

3. fit KRR using radial

get the testMSE = 10.3372, with r=0.001, lambda=0.1

> method = 2
> r = 0.001
> lambda = 0.1
> predY2 <- krr(method, d, trainX, trainY, testX, lambda)
> MSE2 <- mean((predY2-testY)^2)