| Summary: | 碩士 === 國立交通大學 === 資訊工程系 === 89 === Catering the buying behaviors of customers becomes more and more important by the popularization of E-Commerce recently. How to find the association rules efficiently from the transaction records is one of the most interesting topics to be investigated. In this thesis, at, first, we propose en efficient algorithm, named Early Pruning Partition algorithm (EPP), with extending the concept of Partition algorithm and using an early pruning technology to improve the performance of mining frequent itemsets under single minimum support. Then we add the checking of multiple thresholds in EPP algorithm to construct our Multiple Thresholds Early Pruning Partition algorithm (MTEPP). Our MTEPP algorithm can find more effective frequent itemsets corresponding to some events of buying behavior. For evaluating our algorithm, we also implement a simulation environment to verify it. According to our evaluations, our algorithms perform a well performance and find the more useful frequent itemsets indeed. The detailed descriptions of our algorithms will be given in the contents.
|
|---|
