A new supervised classification technique based on the joint use of K-nearest neighbors and weighted K-means to discover new patterns in the data