R program that implements K-Nearest Neighbour with K-Fold Cross Validation. Although the program analyzes "Appointment1000" dataset but can easily be tweeked for other datasets.
The data set is divided into 10 different parts and the KNN model is fitted with the k-1 parts (training set) and predicted using 1 part (test set). Following this, the test mean square error is calculated and the above procedure is repeated 10 times for each part to be treated as a validation set. In the end, the test error rate is calculated as the average of all 10 errors. Finally for all the values of K (0-200 in this case), 10 folds are fited and error rates are determined therfore giving the optimal K with minimum error.
KNN Cross-Validation error rates against all values of K and folds
