SVM_-_Boston.R

data(Boston)

## Warning in data(Boston): data set 'Boston' not found

library(MASS)
library(dplyr)

## 
## Attaching package: 'dplyr'

## The following object is masked from 'package:MASS':
## 
##     select

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

bostonsvm <- select(Boston,medv,crim)
names(bostonsvm)

## [1] "medv" "crim"

plot(bostonsvm,pch=16)
model <- lm(medv ~ crim, bostonsvm)
abline(model)
predictedY <- predict(model, bostonsvm)
points(bostonsvm$medv, predictedY, col = "blue", pch=4)

rmse <- function(error)
{
  sqrt(mean(error^2))
}

error <- model$residuals  # same as data$Y - predictedY
predictionRMSE <- rmse(error)   # 5.703778
library(e1071)
class(Boston)

## [1] "data.frame"

model <- svm(medv ~ crim , bostonsvm)

predictedY <- predict(model, bostonsvm)

points(bostonsvm$medv, predictedY, col = "red", pch=4)

error <- bostonsvm$medv - predictedY
svrPredictionRMSE <- rmse(error)
svrPredictionRMSE

## [1] 8.387847

predictionRMSE

## [1] 8.467038

tuneResult <- tune(svm, medv~crim,  data = bostonsvm,
                   ranges = list(epsilon = seq(0,1,0.1), cost = 2^(2:9))
)
print(tuneResult)

## 
## Parameter tuning of 'svm':
## 
## - sampling method: 10-fold cross validation 
## 
## - best parameters:
##  epsilon cost
##      0.6    8
## 
## - best performance: 67.21342

# best performance: MSE = 8.371412, RMSE = 2.89 epsilon 1e-04 cost 4
# Draw the tuning graph
plot(tuneResult)

tuneResult <- tune(svm, medv~crim,  data = bostonsvm,
                   ranges = list(epsilon = seq(0,0.2,0.01), cost = 2^(2:9))
) 

print(tuneResult)

## 
## Parameter tuning of 'svm':
## 
## - sampling method: 10-fold cross validation 
## 
## - best parameters:
##  epsilon cost
##     0.18    8
## 
## - best performance: 71.00509

plot(tuneResult)

tunedModel <- tuneResult$best.model
tunedModelY <- predict(tunedModel, bostonsvm) 

error <- bostonsvm$medv - tunedModelY  

# this value can be different on your computer
# because the tune method  randomly shuffles the data
tunedModelRMSE <- rmse(error)  
# 2.219642  
tunedModelRMSE

## [1] 8.360493

plot(bostonsvm,pch=16)