Mining Sequential Patterns Based on Cluster-Decomposition-Matrix

• Main Authors: Chang-Hsing Li, 李章興
• Other Authors: Yuh-Jiuan Tsay
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/40855376023206137328

碩士 === 國立屏東科技大學 === 資訊管理系 === 93 === Recently, the technology of database and information has been made speedy progress. Data mining technology is applied to find out the information which is useful to operate a enterprise. Discovering sequential patterns is very important model of application in data mining. Most developed mining methods are based on the Apriori algorithm to improve. However, it is very costly to generate candidate sets since it tediously and repeatedly scans the database. In order to improve the drawback. This thesis presents a new method without candidate itemsets and scanning the original database only twice, named CDM (Cluster, Decomposition and Matrix) algorithm. The CMD algorithm clustered the length of customer sequence. When we want to mining the sequential patterns of length-k, decompose only the clustering sequence which length greater than k, and fill in the matrix which is composed of the sequential pattern of length-(k-1) and length-1. If the entry greater than minimum support in the matrix, the sequential patterns of length-k was found. Through comprehensive experiments, the CDM algorithm gains a significant performance improvement over the PrefixSpan algorithms.