Product Recommendation based on Shared Customer's Behaviour

Today consumers are exposed to an increasing variety of products and information never seen before. This leads to an increasing diversity of consumers' demand, turning into a challenge for a retail store to provide the right products accordingly to customer preferences. Recommender systems are a tool to cope with this challenge, through product recommendation it is possible to fulfill customers' needs and expectations, helping maintaining loyal customers while attracting new customers. However the huge size of transactional databases typical of retail business reduces the efficiency and quality of recommendations. In this paper a hybrid recommendation system that combines content-based, collaborative filtering and data mining techniques is proposed to surpass these difficulties. The recommendation algorithm starts to obtain similar groups of customers using customer lifetime value. Next an association rule mining approach based on similar shopping baskets of customers of the same cluster, in a specific time period is implemented in order to provide more assertive and personalized customer product recommendations. The algorithm was tested with data from a chain of perfumeries. The experimental results show that the proposed algorithm when compared with a base recommendation (made solely on past purchases of customers) can increase the value of the sales without losing recommendation accuracy.