knn

knnn.R

library("caret")
library("mice")
library("kknn")

 
data = read.csv2("~/Desktop/data.csv", sep = ";", dec = ",")

data = data[, c(-1, -3)] 
data[, 1] = as.factor(data[, 1])
class(data[,1])
class(data[,2])

 
data$DEFAULT = make.names(data$DEFAULT)
data[,1] = as.factor(data[,1])
data[,1] = relevel(data[,1],"X1")
zero_var_index = nearZeroVar(data) 
data = data[-zero_var_index]
md.pattern(data)
index_train = createDataPartition(data$DEFAULT, p = 0.70, list = FALSE) 
 
training_set = data[index_train,]
testing_set = data[-index_train,]
 
preprocessParams = preProcess(training_set[,-1], method = c("center", "scale")) 
 
training_set_normalised = predict(preprocessParams, training_set[, -1]) 
testing_set_normalised = predict(preprocessParams, testing_set[, -1])
 
training_set_normalised$DEFAULT = training_set[,1] 
testing_set_normalised$DEFAULT = testing_set[,1]
 

training_set = training_set_normalised
testing_set = testing_set_normalised
train_control = trainControl( method ="cv",
                               number = 10,
                              classProbs = TRUE,
                              summaryFunction = twoClassSummary)


knn_model = train(DEFAULT~., 
          data = training_set,
          trControl = train_control, 
          method = "kknn",
          metric = "Sens")
prediction = predict(knn_model, testing_set) 
conf_matrix = confusionMatrix(prediction, testing_set$DEFAULT)
print(conf_matrix)