| Summary: | 碩士 === 銘傳大學 === 資訊工程學系碩士班 === 107 === Mining frequent itemsets is to find frequently purchased items from the transaction data. Enterprises can combine these items to increase the customer's purchase rate, but some products may be concentrated in a certain period of time. For example, ice products may be purchased frequently in July and August because of the hot weather; Long-sleeved and warm-selling bags may be purchased frequently in January and December because of the cold weather. Some products may be purchased at some time, but not in the entire data set. In this paper, we study how to find which frequent itemsets at which time periods. We partition the transaction time to some time periods and find out which consecutive time periods exist frequent itemsets. The previous approaches divided the transaction data according to the time periods and find the frequent itemsets in each time period. If the itemsets in each successive time period are frequent, then these itemsets are considered as frequent in the continuous time period. However, they will lose the information about the itemsets which is almost frequent in this time period and is frequent in the front or next time periods, such that the itemsets cannot become frequent in the continuous time period. Our approach can reduce the information loss and find frequent itemsets in a long continuous time period.
|
|---|
