I am wanting to set up a data set to use cross validation and want to understand train.kknn better. Does train.kknn produce all of the results for you such that the results of train.kknn do not then need to be plugged back into a test set with kknn? For instance:

model

data = crdf,

kmax = 100,

kcv = 10,

distance = 2,

kernel = "optimal",

scale = TRUE)

print(model)

This code produced results that say:

Type of response variable: continuous

minimal mean absolute error: 0.1850153

Minimal mean squared error: 0.1073792

Best kernel: optimal

Best k: 58

Since kcv was set to 10, my understanding is that automatically creates a cross validation with 10 separate folds. Is that correct? And since kmax is set to 100, it is telling it to cycle through k from one to 100. Is that correct?

After completing this, can the accuracy be tested with the following code?

k = 58

k_result

pred 0.5, 1, 0))

compare

accuracy

accuracy

From this I get .8394 which I interpret to mean that k = 58 is 83.94% accurate. When I change k to 1 or to 100, it does vary the result sufficiently to lead me to this conclusion, but I want to see if that is correct.