A classification method based on PAM algorithm and discrete preprocess

Based on using clustering method to generate training set, and applying rough set theory to discretization preprocess, the classification accuracy can be improve well. This paper presents a new classification method which combines partitioning around mediod (PAM) with discretization preprocess, it builds the train sets from original sample by using algorithm of partititioning around mediod, then discretes train sets by using discrete algorithm of combination of Boolean logic and rough set theory, and trains classifier by the discrete training sets. The experimental result shows that, in confront of same data sets, compared of getting representative data to train decision tree only through method of partititioning around mediod, the new method can produce higher classification accuracy, and use a smaller amount of the train set.