5.29 마이닝실습

install.packages("glmnet")

library(glmnet)

X<-matrix( rnorm(15), ncol=5)

B<-seq(1,0.1, length.out=5)

Y<-drop(X%*%B) +rnorm(3)

fit.1<-lm(Y~X)  #생략

solve(t(X)%*%X) #생략

G.net<-glmnet(X,Y,family="gaussian",alpha=0,standardize=F)

#알파가0이면 릿지, 1이면 라쏘???

rbind(as.vector(G.net$a0),as.matrix(G.net$beta))

########################################

X<-matrix( rnorm(1500), ncol=50)

B<-seq(1,0.1, length.out=50)

Y<-drop(X%*%B) +rnorm(30)

G.net.cv<-cv.glmnet(X,Y,family="gaussian",alpha=0,standardize=FALSE)

########################################

X<-matrix( rnorm(500), ncol=10)

B<-seq(1,0.1, length.out=10)

Y<-drop(X%*%B) +rnorm(50)

fit.1<-lm(Y~X)

fit.aic<-stepAIC(fit.1,direction="forward")

####################################

X0<-matrix( rnorm(500), ncol=10)

B<-seq(1,0.1, length.out=10)

Y0<-drop(X0%*%B) +rnorm(50,0.2)

#LASSO

G.net.cv<-cv.glmnet(X0,Y0,family="gaussian",alpha=0,standardize=FALSE)

LASSO<-sum( ( Y0-cbind(1,X0)%*%as.matrix(coef(G.net.cv)) )^2)

#OLS

Beta<-coef(summary(lm(Y~X)))[,1]

OLS<-sum((Y0-cbind(1,X0)%*%Beta)^2)