ML Using R Section36 Kernel PCA

Section 36: Kernal PCA

df <-  read.csv("G:\\RStudio\\udemy\\ml\\Machine Learning AZ\\Part 9 - Dimensionality Reduction\\Section 45 - Kernel PCA\\Kernel_PCA\\Social_Network_Ads.csv")
df <-  df[,3:5]
head(df)

Split dataset into training and test set (300 training, 100 test)

library(caTools)
set.seed(123)
split <- sample.split(df$Purchased, SplitRatio = 0.75)
training_set <- subset(df, split == TRUE)
test_set <- subset(df, split == FALSE)

For KernelPCA ,it is better to do feature scaling (normalization)

# Feature Scaling 1 age, 2 is salary
training_set[,1:2] <-  scale(training_set[,1:2])
test_set[,1:2] <-  scale(test_set[,1:2])

Applying PCA to the data

# install.packages("kernlab")
library(kernlab)
kpca <-  kpca(~., data = training_set[-3], kernel = 'rbfdot', 
              features = 2)

# add the depenent variable into the pca 
training_set_pca <-  as.data.frame(predict(kpca, training_set))
# add the dependent varaible Purchased to the training set
training_set_pca$Purchased <-  training_set$Purchased
# do the same for the test dataset
test_set_pca <-  as.data.frame(predict(kpca, test_set))
# add the dependent varaible Purchased to the test set
test_set_pca$Purchased <-  test_set$Purchased

Building the classification model

classifier <- glm(formula = Purchased ~ .,
                  family = binomial,
                  data = training_set_pca ,
                  )
summary(classifier)

Call:
glm(formula = Purchased ~ ., family = binomial, data = training_set_pca)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-2.8679  -0.4607  -0.1230   0.3191   2.4451  

Coefficients:
            Estimate Std. Error z value Pr(>|z|)    
(Intercept) -1.27556    0.21592  -5.907 3.47e-09 ***
V1           0.47918    0.05873   8.159 3.39e-16 ***
V2           0.14365    0.03280   4.380 1.19e-05 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 390.89  on 299  degrees of freedom
Residual deviance: 191.09  on 297  degrees of freedom
AIC: 197.09

Number of Fisher Scoring iterations: 6

Predicting the test set results

prob_pred <-  predict(classifier, newdata = test_set_pca[-3]
                   )
y_pred <-  ifelse(prob_pred >0.5,1,0)

Evaluate the prediction using confusion Matrix.

# Making the confusion matrix
# [3] refers to the outcome
cm <- table(test_set_pca[,3], y_pred)
cm

   y_pred
     0  1
  0 58  6
  1 13 23

Plot

# install.packages("ElemStatLearn")
library(ElemStatLearn)

Warning message:
In scan(file = file, what = what, sep = sep, quote = quote, dec = dec,  :
  EOF within quoted string

set <- test_set_pca
X1 <- seq(min(set[,1]) - 1, max(set[,1]) + 1, by = 0.01)
X2 <- seq(min(set[,2]) - 1, max(set[,2]) + 1, by = 0.01)
grid_set <- expand.grid(X1, X2)
colnames(grid_set) = c('V1','V2')
prob_set = predict(classifier, type = 'response', newdata = grid_set)
y_grid = ifelse(prob_set > 0.5, 1, 0)
plot(set[,-3],
     main = 'Logistic Regression(Test Set)',
     xlab =  'Age', ylab = 'Estimated Salary',
     xlim = range(X1), ylim = range(X2))
contour(X1, X2, matrix(as.numeric(y_grid), length(X1), length(X2)), add = TRUE)

points(grid_set, pch = '.', col = ifelse(y_grid == 1, 'springgreen3', 'tomato'))
points(set, pch = 21, bg= ifelse(set[,3] == 1,'green4', 'red3'))