| Summary: | 碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 103 === In this thesis, a frequent closed sequential patterns with time constraints generation algorithm is developed. The proposed method uses the developed early termination criteria to speed up the process of generating patterns. By determining a set of valid items, which are used for generating type-1 or type-2 patterns, this method performs forward and backward checking to find frequent closed patterns. From the experimental results, compared with CTSP, we can find the presented approach can reduce the computing time by 51.2% using the sequential databases, generated from T10I4D100k data set, with number of data sequences = 5000, minimum gap = 5, maximum gap = 16, sliding window = 3 and minimum support threshold = 0.5%. Using the same data set, the proposed method can reduce the computing time of CTSP by 38.54% with number of data sequences = 5000, minimum gap = 5, maximum gap = 16, sliding window = 3 and minimum support threshold = 1%. These results show that this method is more remarkable when a larger data set with smaller minimum support threshold is used.
|
|---|
