Designing an e-commerce recommender system based on collaborative filtering using a data mining approach

E-commerce recommender systems have been converted to a very important decision-making helper for customers, and provide online personalised recommendations using information technology and customers' information. In the meantime, collaborative filtering (CF) recommender systems are one of the key components of successful e-commerce systems. Despite the popularity and successes of CF, these systems still face a series of serious limitations, including cold start, sparsity of user-item matrix, scalability and change of user interest during the time, that impede exact recommendations to customers. Although much research has been presented to overcome these limitations, no comprehensive model is yet offered to reduce them: 1) customer segmentation based on LRFM variables in the level of product category to evaluate the length of customer relationship with the company, recency, frequency, and monetary of purchasing product categories; 2) extracting association rules based on user-category matrixes in the level of each cluster; 3) customer segmentation according to demographic variables; 4) change of user-item matrix and reduction of its dimensions; 5) developing a new similarity function by weighted combination of results of segmentation methods and CF. According to the gained results, the proposed system of this research has resulted in the removal of traditional CF constraints and presents more accurate and appropriate recommendations for the preferences of customers.