A Characteristic Rules Mining Scheme Based On Weighted Incidence Matrices

摘要

Many researchers have weighted the mining of information and knowledge from large electronic databases as a key research topic in database management and machine learning, and many hi-tech industrial companies have also valued it as an important future opportunity of major revenues. The goals of data mining methods are to achieve tasks such as data clustering, association, classification and so on. One specific task is classification rules discovery, which has drawn great attention in the data mining community during the past few years because it has been demonstrated that classification rules can contribute a lot to many real-life businesses and engineering cases. Quite a few algorithms have been proposed for finding classification rules so far, and yet most of them require multiple database scans, which increase both the I/O cost and the time consumed. To improve from the drawbacks of those methods, an efficient classification rules mining method, the weighted incidence matrix, is presented in this paper that will reduce both the number of database scans and the rules extracting time.