EFFICIENT MINING OF WEIGHTED QUANTITATIVE ASSOCIATION RULES AND CHARACTERIZATION OF FREQUENT ITEMSETS

Abstract — In recent years, a number of association rule mining algorithms were developed. In these algorithms, two important measures viz., support count and confidence were used to generate the frequent itemsets and the corresponding association rules in a market basket database. But in reality, these two measures are not sufficient for efficient and effective target marketing. In this paper, a weighted frame work has been discussed by taking into account the weight / intensity of the item and the quantity of each item in each transaction of the given database. Apriori algorithm is one of the best algorithm to generate frequent itemsets, but it does not consider the weight as well as the quantity of items in the transactions of the database. This paper consists of two phases. In the first phase, we propose an algorithm Apriori-WQ, which extends the Apriori algorithm by incorporating the weight and quantity measures and generates Weighted Frequent Itemsets (WFI) and corresponding Weighted Association Rules (WAR). The rules are filtered based on a new measure called Minimum Weight Threshold (MWT), and then prioritized. Some itemsets may not be frequent but they satisfy MWT. Such sets are also generated. In the second phase we analyze the transactions {Ti}, which form the frequent itemsets and the customer characteristics (i.e., attributes) of those transactions {Ti}. Experiments are performed to establish a relationship between frequent itemsets and customer characteristics. 3D-graphical reports are generated, which helps the marketing leaders for making better predictions and planning their investment and marketing strategies