Research On Incremental Associative Classification Algorithm And Multi-Label Classification Approch

Utilizing association rule mining methods to construct classification models in data mining is known as associative classification (AC). AC technique integrates the two techniques association rules mining and classification well. In the last few years, associative classification algorithms showed experimentally that they generate more accurate classifiers that traditional classification approaches such as decision trees.However, in the process of researching on associative classification approaches, we find that these algorithms pay little attention to the questions that maybe arise in applications. At the same time, the necessity for application of efficient classifiers increases steadily. So, from practical application perspective, we proposed study on improving the performance of AC.Firstly, focus on the database updating in practical application, an incremental updating algorithm applied to AC technique (IUAC) is introduced in this paper. The algorithm is based on the FP-growth to mining and updating the association rules. And it applies a tree structure to store the association rules which are mined for classification. Moreover, in this paper, it adds more appropriate constraints in rule ranking procedure in order to control the number of class associative rules and improve the predictive accuracy.Secondly, focus on the multi-label classifications increasingly required by modern applications, propose a correlated multi-label lazy associative classification method, CMLAC. Based on the research on multi-label classification and lazy associative classification, CMLAC uses progressively filter to exploiting correlations among labels and getting a multi-label class association rules set, CMARs. Thus the problem that existing classification can not solve the multi-label classification is conquered.