An exorbitant source of data is easily available, but the actual task lies in using this data efficiently. In this article, the aim is to analyse the significant information embedded in the customer purchase behaviour to recommend new products to them. The proposed scheme is a two-fold approach. First, the authors retrieve various product correlations from the vast library of user transactions. Based on these product correlations, utility-based association rules are learned which depict the customer purchase behaviour. These rules are then applied in a recommender system for novel product suggestions to the customers. With improved utility-based mining, the paper tries to incorporate the usefulness of an item set like cost, profit, or any other factor along with their frequency. In this paper, the authors have deployed the rules discovered from both the conventional frequent item set mining and improved utility-based mining on an e-commerce platform to compare the accuracy of the algorithms. The obtained results establish the efficacy of the proposed algorithm.
展开▼
