Mining Main Points of Knowledge in Relational Databases for Classification

Class association rules can be used not only for concept description but also for classification, so mining class association rules has drawn wide attention in data mining field. Specially, associative classification, which is based on class association rules, has been a hot spot in data mining research and machine learning community. At present, it is widely accepted that in general, associative classification has higher accuracy and better robustness than decision tree classification. However, associative classification has some deflects such as slow performance, huge memory usage and large-size classifying model. In this paper, a new idea is proposed to mine main points of knowledge in relational databases based on class association rules, which can be successfully used for fast, accurate and robust classification.