MSD-Apriori: Discovering borderline-rare items using association mining

Abstract

Ecommerce business constantly decides innovative strategies to increase their sales and hence earn profit. They mainly strive to boost the sale of those items that are rarely purchased. There are few borderline-rare items that lie just below the minimum support threshold and may have a strong correlation with frequent items. The minimum support threshold is the user-defined minimum support value for an item. If these borderlinerare items are strategically placed in the market then it can help the e-commerce industry to improve their sales further. In this paper, we propose a hybrid approach, MSD-Apriori to discover borderline-rare elements which are below but close to minimum support threshold and have strong correlation with frequent items. The hybrid approach is formed by integrating MS Apriori with Dynamic Apriori. MS Apriori finds the borderline-rare item sets from the web logs and Dynamic Apriori discovers those items among these that share strong correlation with the frequent items by association rule mining. The proposed method is evaluated on Kosarak, a real dataset that gives encouraging results.