Research On Data Mining Algorithm Based On Time-Series Pattern

Data Mining has become one of the fast growing areas of research in recent years. Besides association rules mining, researchers endeavor to develop mining methods with time factor considered. Popular research topics include customers buying patterns analysis, Internet surfing time-series analysis, trend analysis, and so on. When probing the customers buying time-series patterns, most developed mining methods require repeated database scans to generate candidate patterns, which are then checked to find frequent time-series patterns.The Apriori algorithm is the method of finding Boolean association rules, but has the disadvantage in the complexity of space and time. Therefore, this dissertation introduces a new frequent-pattern (FP) growth algorithm that does not need to produce the candidate item sets. This algorithm compresses information in database to the FP-tree, then produces frequent pattern by joining suffix with prefix, consequently avoids scanning the database many times, and towers the time expense.It therefore deteriorates the performances of these methods. This dissertation presents a Frequent Pattern Adjacent Matrix (FPAM) to record intermediate length-2 patterns. After finding the frequent patterns, it only needs one more round of database scan to find all the time-series patterns by taking advantages of FPAM. Without generating unnecessary patterns, the proposed method is an efficient method for mining frequent time-series patterns from databases.In this dissertation, we embedded FPAM as the algorithm for time-series patterns mining. Time-series patterns are combined to obtain multi-dimensional sequential patterns. With this kind of approach, when given frequent patterns, we can enhance the efficiency of data mining by rescanning the database only once.