Research On Closed Sequnential Patterns Mining Algorithm Based On Constraints

Existing closed sequential patterns mining algorithms can efficiently mine closedsequential patterns. However, following the applications becoming more and morecomplex, user has more and more requirement in mining result. So, the constraints that areconverted by user-specified requirements or interest are pushed into closed sequentialpattern mining process to restrict closed sequential patterns mining. This will be theresearch focus in sequential patterns mining area. The paper has mainly focused on how tomine closed sequential patterns with constraints. These researches have broad applications,including analysis of customer behaviors, analysis of web access patterns, and DNAanalysis, etc.Firstly, an algorithm for mining closed sequential patterns based time constraints ispresented. This algorithm pushes the time constraints into closed sequential patternsmining process. In the pattern-growth framework, it uses the se time and the timeline tomine closed sequential patterns with time constraints in the projected database of prefixpattern. Furthermore, the algorithm adopts the bi-directional closurechecking technique tomine closed sequential patterns. The algorithm outperforms CTSP algorithm for miningclosed sequential pattern with time constraints in large database.Secondly, an algorithm for mining closed sequential patterns based regularexpression constraints is presented. This algorithm pushes the regular expressionconstraints as the user-specified requirements into the closed sequential patterns miningprocess to realize reactive mining. The memory index method is adopted and a hashstructure is used to do the closurechecking of sequential patterns. The algorithm canefficiently mine closed sequential patterns and satisfy the requirements of user.Finally, an algorithm for mining weighted closed sequential patterns based on regularexpression constraints is presented. This algorithm pushes the weight and regularexpression constraints into closed sequential patterns mining process. It uses the memoryindex method and calculates the minimum weighted support count of every sequence. Thesequential patterns satisfying the minimum weighted support count are found in the index set. A hash table is utilized to do the closurechecking of weighted sequential patterns toimprove the mining efficiency. The performance of the algorithm outperforms WCSpanalgorithm.The experimental results show the algorithms that are presented in this paper canefficiently resolve respective problems. Their performance outperforms obviously existingsimilar problems and achieves the expectant goal.