The Clustering Aglorithm Research And Application Based On Rough Set

With the Information age coming,the development and prosperity of every department needs the knowledge discovery. The size of data in repository is increasing, it is very important to extract hided or potential knowledge and regulation to influence future determination from sea-sized data, thus it requires suitable mechanism for data-mining. This paper mainly research clustering algorithm. Clustering algorithm is an un-supervised learning algorithm, the objective is to partition given data elements to homogeneous groups, and those groups are called cluster.This thesis mainly researches on rough clustering algorithm which is combined by rough set and clustering algorithm. It uses the three basic properties proposed by Professor Z.Pawlak in 1982, to improve PAM algorithm, which is a partional clustering algorithm, thus turn to rough PAM clustering algorithm. Compares to rough K-means algorithm,which is proposed by Professor P.Ligras in 2002, rough PAM clustering algorithm requires less initial parameters by two, and the results obtained by experiments of this algorithm are better.This thesis improves K-means algorithm,which clustering mixed-data proposed by Huang in 1997, thus turn to a new K-mans clustering algorithm which disposes mixed-data. This algorithm obtains approving experimental results via UCI data.In the end, this thesis designs a cluster-algorithm system (based on B/S structure) for the requirements of experiments. This system combines K-means, PAM, rough K-means, and rough PAM clustering algorithm. This system programmers by Java language, operates by UI from IE in Windows, utilizes MySQL as background database.