Data mining is the process for achieving an interesting and potential knowledge from a large data sets, which is a focus study field. One of the important fields in data mining is the mining for classification rules, in which one can get the knowledge to predicate the future or unknown objects according to the analysis of the sample data. Attributes reduct, mining classification rules, and discretizing values of quantitative attributes are three fields in mining classification rules.This paper focuses on the three fields of mining classification rules, and presents three new algorithms:(1) The discernibility function-based of rough set theory algorithm to eliminate redundant attributes from an information system, which can get all reducts of an information system, especially the best one.(2) The algorithm of mining classification rules based on discernibility function, which can induct simple, general, and understandable rules.(3) The evolutionary computation-based algorithm for discretizing values of quantitative attributes, whose advantage is that it can finding the best cuts of a quantitative attribute.
|