Tuning GSP parameters with GA

• Main Authors: Wei-ci Jheng, 鄭洧奇
• Other Authors: Ping-Yu Hsu
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/462t5e

碩士 === 國立中央大學 === 企業管理學系 === 102 === Tuning GSP parameters with GA ABSTRACT In data mining, association rules can be shown when customers buy products, which products will be purchased at the same time. Scholars use this feature to develop market basket analysis to formulate marketing strategies for business. As we know, the data are changing all the time. When new data generate, the old data will be replaced. In the database, time become a very important attribute. And new data mining method have been proposed, called generalized sequential patterns (GSP). GSP uses time stamp to find the product portfolio with sequential patterns. However, the GSP parameter is user-defined. The result of the operation may be unstable, because of the parameter setting incorrectly. Tuning the parameters used in this study combined GSP and genetic algorithm (GA) to improve the result continuously, to find the appropriate parameters. In the experiment, we use a medium-sized supermarket verify the results and found that after comparing with random input parameters, the parameters of the proposed method found significantly better than a random set of parameters. Keywords：Sequential pattern mining、GSP、GA