K value determines how many other instances must be similar to each instance in order for it to not be considered an outlier.