R code for MLR for prediction of agricultural water consumption

# MOdel selection, AIC, BIC, AdjR2, RSE

library(Metrics)
library(scales)
library(leaps)
library(car)
library(caTools)
MyData1 <- read.csv(file="data for agriculture water prediction.csv",sep=",")
MyData=MyData1[1:45,1:9]
set.seed(123)
split = sample.split(MyData$AgriWC, SplitRatio = 0.9)
training_set = subset(MyData, split == TRUE)
test_set = subset(MyData, split == FALSE)
MyData2=MyData1[46:47,2:9]

#Model selection
refit=regsubsets(AgriWC~.-ï..Year,training_set,method='forward')
sumry=summary(refit)
names(sumry)

par(mfrow=c(2,2))
plot(sumry$rss,ylab='rss',type='l')
p=which.min(sumry$rss)
points(p,sumry$rss[p],pch=17,col='red')
plot(sumry$adjr2,ylab='adjr2',type='l')
p=which.max(sumry$adjr2)
points(p,sumry$adjr2[p],pch=17,col='red')
plot(sumry$cp,ylab='cp',type='l')
p=which.min(sumry$cp)
points(p,sumry$cp[p],pch=17,col='red')
plot(sumry$bic,ylab='bic',type='l')
p=which.min(sumry$bic)
points(p,sumry$bic[p],pch=17,col='red')
#sumry$which
se=coef(refit,6)
#Fiting selected model
#Scaling data

# Feature Scaling
training_set = data.frame(scale(training_set))
test_set = data.frame(scale(test_set))
Fulldata=data.frame(scale(MyData))
names(MyData)

#Fitting model
lm.fit=lm(AgriWC~.-ï..Year-Food.production.index..2004.2006...100., data =MyData)
summary(lm.fit)

#Prediction on test,train and entire data (to plot predicted vs original data)
predicttest=predict(lm.fit, test_set)
rmsetest=mse(test_set$AgriWC,predicttest)

predicttrain=predict(lm.fit, training_set)
rmsetrain=mse(training_set$AgriWC,predicttrain)
vif(lm.fit)

predictfull=predict(lm.fit, Fulldata)
rmsefull=mse(Fulldata$AgriWC,predictfull)


# Rescaling data
vector1= c(min(MyData$AgriWC),max(MyData$AgriWC))
pred2=data.frame(rescale(predictfull, to=vector1))