An Efficient Generation of Candidate Itemsets and Count Algorithm for Mining Association Rules

• Main Authors: Pei-Yun Wu, 吳珮芸
• Other Authors: Bin-Yih Liao
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/39905924144178640330

碩士 === 國立高雄應用科技大學 === 電子與資訊工程研究所碩士班 === 92 === Mining association rules from transaction databases is one of important techniques in data mining. Applications of association rules extend to discovering frequent patterns in consumer behavior, marketing analysis, electronic commerce and education, and other areas. In this thesis, we developed EGC, is an efficient algorithm for mining association rules. The main improvements are EGC uses an innovative method for generating candidate itemsets by checking the numbers of the preceding frequent itemsets before joining procedure. And EGC uses the simple tree data structure for storing the candidate itemsets and counting their supports. In addition, EGC uses the database global pruning method of DCP for efficiently reducing the size of the database. The experiments show that the performance of EGC is better than Apriori and DHP,IHPwoTTP, and IHPwTTP. The execution time and memory required of EGC are less than Apriori, DHP, IHPwoTTP, and IHPwTTP. In other applications, EGC can efficiently mine interesting information from investor databases to provide the optimal portfolio for each investor of the brokerage securities firm.