Market Basket Analysis using Improved FP-tree

Abstract

The Market Basket Analysis helps in identifying the purchasing patterns of customers such as, which products are purchased more and which products are purchased together. This helps in decision making process. For example, if two or more products are frequently purchased together then they can be kept at the same place so as to facilitate the customer, to further increase their sale. The price of products that are not frequently purchased can be reduced in order to enhance their purchase. Additionally the promotion of one product will also increase the sales of other products which are purchased together with the product being promoted. The traditional Apriori algorithm based on candidate generation cannot be used in Market Basket Analysis because it generates candidate sets and scans database regularly for the generation of frequent itemsets. The FP-growth algorithm cannot be used despite of the fact that it does not generate candidate sets and scans the database only twice because, it generates a lot of conditional trees recursively. Therefore, an efficient algorithm needs to be used. In this paper an efficient algorithm is used for development of market basket analysis application. This efficient algorithm neither generates candidate sets nor conditional FP- tree; like FP-growth scans the database twice