An incremental algorithm for mining classification rules in incomplete information systems

One of important research areas in data mining is to develop methods to update knowledge by using the existing knowledge, since it can generally enhance mining efficiency, especially for very large database. Rough set, a new mathematical tool to deal with vagueness and uncertainty, has been successfully applied in data mining. Presently, several approaches based on rough set have been proposed for mining tasks when knowledge updates. However, it only aims at complete information systems. In fact, many information systems are incomplete in practical application. Therefore, in order to support more effective data mining tasks, it is meaningful to develop approaches to update knowledge in incomplete information systems (US). A method for incremental updating approximations of a concept was proposed by T. R. Li, et al (2003), which may realize adding and deleting a few attributes simultaneously at a time. Based on this method, we develop an incremental algorithm for mining classification rules from IIS. Complexity analysis of the algorithm and example show that our method can realize updating knowledge effectively.