For d = 10

Since the accuracy is almost constant for small k. K  = 1 can give least misclassification rate.