setwd("current_working_directory")
df <- read.csv("creditlimit26aug.data",header = FALSE) ## training dataset
#tdf <-read.csv("APIOutcome.csv", header = TRUE) ## test dataset
tdf <- read.csv("creditlimit31augedit.data",header = FALSE) ## test dataset
colnames(df) <- c("SNo","Salary","LoanAmt","Level")
colnames(tdf) <- c("SNo","Salary","LoanAmt","Level")
plot(df$Salary,df$LoanAmt)
plot(tdf$Salary,tdf$LoanAmt)

library(class)                             #For knn
library(caret)                             #For confusion matrix

#Normalization Function
# normalize<-function(x){return ((x-min(x))/(max(x)-min(x)))}

dependent_col=4             #####Change 5 to dependent variable's column     number
predictor_col=c(2,3)    #####Change 1,2,3,4 to predictor variable's column numbers
#dataset_normalized<-as.data.frame(lapply(dataset[,predictor_col],normalize))          #Here we are normalizing the predictor columns

set.seed(150)                    #seed is set as 12
training_data<-df[,-dependent_col]       #Using every column from normalized       dataset except the dependent variable for training
testing_data<-tdf[,-dependent_col]        #Using every column from normalized dataset except the dependent variable for testing
train_target<-df[,dependent_col]       #train target contains dependent variable values of training dataset
test_target<-tdf[,dependent_col]       #test target contains dependent variable values of testing dataset

model<-knn(training_data,testing_data,cl=train_target,k=5)    #Don't change anything,here we are creating the model
table(test_target,model)                        #No change required
cm<-confusionMatrix(test_target,model)          #No change required

Accuracy <- cm$overall[1]                       #To extract accuracy
Accuracy                                   #To print accuracy

testing_data$actual = as.character(test_target)
testing_data$predicted = as.character(model)
rmat = as.matrix(testing_data[testing_data$predicted != testing_data$actual,])
rmat