A Study of Genetic Algorithm Based Web User Session Clustering Technology

• Main Authors: Kuo-Chih Chuang, 莊國志
• Other Authors: Shing-Hwang Doong
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/63745759985979742955

碩士 === 樹德科技大學 === 資訊管理研究所 === 90 === Clustering analysis is a common tehchnology used in data mining. It is based on some criteria to separate a set of data into serveral clusters such that intra cluster data have high similarity and inter cluster data have high dissimilarity. There are user clustering, page clustering and session clustering when we apply clustering analysis in web mining. The user clustering is made on users’ demographic data; the page clustering is based on the attribute or content of web pages; the session clustering is to cluster users’ browsing sessions of web pages. This thesis proposes a new clustering technique based on Genetic Algorithms (GAs). It uses uers’ sessions as the data to be clustered, and employs a visiting order based formula to measure the similarity bewteen two sessions. The variance of similarity distribution is used as a refernce to choose the proper number of clusters. Assocition rules and prediction by partial matching are used to verify the effectivenss of the proposed clustering technique.