Research On The Mining Algorithms Of Association Rule

Data mining is the process of abstracting unaware, potential and useful information andknowledge from plentiful, incomplete, noisy, fuzzy and stochastic data result of practicalapplication. Association rules mining is an important branch of research in the field of datamining. It researchs and finds mainly the interesting relationship among attributes in thosedata result of practical application.By reading the documents at home and abroad, some association rules mining algorithmswere researched.In this paper, the research is focused on the following two points.Firstly, a new algorithm, named Distributed Mining Global Maximum Frequent Pattern(DMGMFP), for mining distributed global maximal frequent patterns from databases wasproposed. DMGMFP had the local mining phase and the global mining phase. During thelocal mining phase, first DMGMFP created respectively the tree named Improved FrequentPattern tree (IFP-tree) on each node and used figure sequence to store frequent patterns, thenit discovered the local maximal frequent patterns after scanning the local databases. Duringthe global mining phase, DMGMFP used to share with all nodes in the local maximal frequentpatterns and broadcasted patterns information for sets communication, so that the globalmaximal frequent patterns was mined. DMGMFP was implemented and evaluated itsperformance for various cases. It demonstrated better performance than other algorithms. Theexperimental results showed that the proposed algorithm was efficient.Secondly, considering the timeliness of data stream, we propose an algorithm of datastream frequent closed pattern mining, named Mining Frequent Closed Pattern Algorithm(MFCPA-stream), which combined both the transaction sliding window and the time decayingmodel. By dynamically constructing the full merged sorted frequent pattern tree with headtable (HMSP-tree), the algorithm captured the latest patterns’ information timely in the datastream. Moreover, the algorithm used pruning and merged method to maintain the full mergedsorted frequent pattern tree with head table. The experimental results showed that the proposed algorithm was efficient.