A two-step integration algorithm for discretization based on rough set theory

Discretization plays an important role in rough set theory. So far, many efficient algorithms have been proposed, but few of them can process data sets quickly with high recognition rate. In this paper, the distribution of importance of candidate breakpoints on single attribute in decision table is analyzed. Firstly, the candidate cuts will be dynamic clustered combining with Information Entropy value of candidate breakpoints. Secondly, the clustered cut points will be selected using the importance algorithm of candidate cut points. The dynamic clustering can decrease the number of condition points effectively which is useful to select breakpoints. Experiment results show that the proposed algorithm can process data sets quickly with high recognition rate.