Integrating collaborative filtering and association rule mining for market basket recommendation

Abstract

This paper proposes a market basket recommendation algorithm based on association rules and collaborative filtering. It solves the problem that traditional association rule recommendation algorithms cannot generate association rules from cold commodity items under big data environment. An implicit semantic model based on historical transaction data of all users is constructed to represent potential features of commodities and measure similarities among commodities. The missing unknown elements in the implicit semantic model are complemented by the least square method. Association rules on unpopular commodities are obtained by the similarity of the commodities, thereby improving the recommendation accuracy. Experiments with real supermarket sales data demonstrate its effectiveness.