Summary: | 碩士 === 淡江大學 === 資訊工程學系 === 92 === Mining sequential patterns is an important issue of data mining. Mining sequential patterns is to discover the relativity of time between distinct records in database. Most mining sequential patterns algorithms derive all sequential patterns, and then pick up the interesting patterns. Under the condition that the goal is specific, these traditional algorithms are non-efficient. Besides, most algorithms ignore quantitative information in transaction records, and can’t deal with fuzzy specific goal. To solve these problems, a new algorithm that calls Fuzzy Goal-Oriented Sequential Patterns is proposed in this paper. This algorithm has three characters: (1) Under the condition that the goal is specific, this algorithm derives only sequential patterns with goal related, and no other redundant patterns. (2) The specific goal of this algorithm is fuzzy. (3) This algorithm can deal with quantitative information in transaction records. By the experiment results, our proposed new algorithm outperforms traditional sequential pattern algorithms.
|