Research on Parallel Frequent Pattern Discovery Based on Ontology and Rules

Frequent pattern discovery is a hotspot of knowledge discovery, which is based on the hybrid knowledge base that consists of ontology and rule logic. However, the mining efficiency of it is too low and unsatisfying, since the reasoning process needs large quantities of computing. Thus, we need to develop the parallel algorithm and architectures to solve this problem. In this paper, we come up with a parallel Inductive Logic Programming(ILP) frequent pattern discovery method based on MapReduce Framework, and then establish the Trie Tree by node expansion algorithm. Finally, we distribute the various computational pattern detection processes to each parallel computing nodes in the Frequent Pattern Tree(FPT) and get the frequent pattern. This way can also eliminate the semantic redundancy, which is judged by support degrees. Consequently, by the verification in experiments, we find this frequent pattern discovery algorithm can improve the efficiency of pattern discovery in a large degree.