Research On Incremental Learning Algorithm Based On Decision Logic

With the development and application of database technology, the various social departments accumulated the massive data, which is still increasing every day. The data mining is an effective tool to be used to discover the potential meaning knowledge in those data. But, if we need recalculate rules from all data to mine knowledge after updating the database, it will consume massive resources, which causes the urgent demand of the incremental mining algorithm. Unifying the algorithm and the database update in together would gradually update the knowledge in the database, revising knowledge and strengthening what have already discovered, which adapts revised knowledge base to updated database, and need not mining all data.The rough set theory is one of the data mining methods, which is a mathematical way to process fuzzy and uncertain knowledge. Rough Sets has already been applied in the artificial intelligence, the knowledge discovery, Pattern Recognition, intelligent control, intelligent decision, conflict analysis , troubleshooting, etc. But, at present, the data mining algorithm which is based on the rough set theory mainly aims at the static data sets. Considering the demand of practical application and the present theory situation of rough set, this article mainly conducts the research of incremental algorithm in the framework of the rough set theory. But the decision logic in Rough Sets, as a model of extracting knowledge from the information system, uses the symboltool of logic deduction, which can effectively discover the dependence of knowledge and reduce it. Furthermore, it has the capability of the accurate description to the new incremental instances. So the theory of the decision logic is considered as the basic theory in this paper.First, we carry on an analysis to the research of incremental algorithm based on the decision logic theory in paper [15], and point out the mistakes which exist in the classification method to incremental instances, and design the new classification method to the new incremental instances through modifying the theory, and make the detailed discussion to the computation of complete-minimum decision algorithm based on the classification method, and propose the new incremental learning theory based on the decision logic. Second, in terms of the proposed new incremental learning theory, we design the incremental learning algorithm, and analyze the time complexity of the algorithm. It is proved theoretically that the incremental algorithm has the very big superiority than classical algorithm on the time complexity. Finally aiming at the same situation in the time complexity, we examine it in the aspect of time performance through the experiment.