A Case Study for Presenting Bank Recommender Systems based on Bon Card Transaction Data

As with many other businesses, banking industry tends to digitalize its working processes and use state-of-the-art technique in the financial and commercial areas in its business. The main core of the bank business is managing communication with customers which eventually results in investment on customers. In this paper, the structure of a recommender system is described, whereby using the recommender technology the places for purchase in which so far, the customers have not used any special type of Bon cards but are probable to buy from them is estimated and proposed to the customers.Matrix factorization is a type of method for collaborative filtering based on models which is widely used for rating prediction concept. Generally, bank products are not rated by customers; these products are usually purchased or offered to customers by the bank. Therefore, to determine the rating, RFM 1 method which is an instrument for analysis in marketing is used along with clustering algorithm to determine the customer value and place. If a place does not have any value, i.e. the data have missing values, it suggests that we do not know whether the customer prefers the place for purchase or not. In this paper, a hybrid method based on dimension reduction technique is presented. This method is able to predict the missing values in data to offer recommendation to customers. Assessment of the proposed model through Root Mean Square Error 2 indicates that the architecture in this paper has less error in comparison to common collaborative filtering methods.