It's a Super K-Nearest Neighbor(SKNN) classification method with using kernel density to describe weight of the distance between a training observation and the testing sample. Comparison of performance between SKNN and KNN shows that SKNN is significantly superior to KNN.