An Approach of Mining Multiple-Level Sequential Patterns in Large Databases

• Main Authors: Bo-Chang Lin, 林伯昌
• Other Authors: Guanling Lee
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/33269805992799478456

碩士 === 國立東華大學 === 資訊工程學系 === 93 === Sequential pattern mining is an important problem with broad applications in the field of datamining research. Such as the prediction of customer purchase behavior, Web access pattens, Natural disasters, DNA sequences and so on. But almost of the researches are mining with single level. Actually, in the real life, almost of the informations are implied with the “Hierarchy Taxonomy”. And if we consider the concept of hierarchy taxonomy, the best advantages is that we can gain the generalized knowledge or more concrete and detailed informations depending on the users. In this paper, we apply the concept of hierarchy taxonomy to the sequential pattern mining to gain the multiple-level sequential patterns. The mining approach is the extending of the Apriori algorithm mining with single level. The method we implement the concept of hierarchy taxonomy is to construct the correlative tree with items in the database. Then we transform the original database to the type of coding and mining . We generate our synthetic database with IBM data generator. And we prove that we can speed up the mining process with using of the mining results of last mining.