question 2.2 - knn
#install.packages("kknn")
library(kknn)
Read the file
data <- read.table("/Users/marylin/Downloads/hw1/data 2.2/credit_card_data-headers.txt",header=TRUE)
Set up predictors and response
x<- data[, -11]
y<- data[,11]
Call kknn
k<- 8
y_pred <- vector("integer",length(y))
#exclude i in the nearest neighbor
for (i in 1:nrow(data)) {
train_x <- as.matrix(data[, -11])
train_y <- data[,11]
train_data <- data.frame(train_x, R1=train_y)
test_X <- data[i, -11, drop=FALSE]
test_data <- data.frame(test_X)
#kknn
knn_model <- kknn(R1 ~., train_data, test_data, k=k, kernel="rectangular")
#predict
y_pred[i] <- as.integer(fitted(knn_model))
}
Calculate how good the classifier is
accuracy <- sum(y_pred == y) /length(y)
confusion_matrix <- data.frame(y, y_pred)
precision <- confusion_matrix[2,2]/sum(confusion_matrix[,2])
recall <- confusion_matrix[2,2]/sum(confusion_matrix[2,])
f1 <- 2*(precision *recall) /(precision +recall)
Print the output
1
