Collaborative Filtering Method Based On RFM Model And Its Application In Personalized Recommendation

With the growing competition in the modern market, how to accurately assesscustomer value and identify the most valuable customers and then carry onpersonalized marketing according this become the key to win customers. Personalizedrecommendation method can provide customers with product recommendations, that isto say, achieve that the appropriate product recommends to the right people, thus canmeet the customer’s individual requirements and transform the potential customers intoactual buyers.Collaborative filtering is a main algorithm of personalized recommendation, andis one of the most successful recommendation technology applied in the recommendedsystem, it has many superiorities, such as handling complex unstructured objects, ahigh degree of personalization and so on. However, the inherent disadvantage has agreat impact on the recommended results. This paper presents a personalizedrecommendation method which based on the RFM customers classification, usedifferent values of customers’ behavior to predict the customers’ preferences, and wecan provide customers with an adaptive recommendation mechanisms to solve thecurrent problems of recommending inappropriately, thus can avoid the invalidrecommendation.Firstly, this paper analyzed and compared different recommendation algorithms,emphasis on the collaborative filtering recommendation algorithm, and analysis theproblems which were exposed by the collaborative filtering recommendation algorithmin practical application. As the existing problems, I proposed the collaborative filteringrecommendation method which based on the RFM model. This recommended methodis based on the history consumption records of a commodity, and then combined withthe RFM model to calculate the combined result of customers for goods R, F, M, thencombined with the weight of each indicator to calculate the comprehensive evaluationof customers for goods. This is a passive customer’s appraisal, because of taking intoaccount the comprehensive index of various aspects, so compared to the evaluationof active customers, this evaluation has a certain objectivity. After that, I use thetransaction history data of customers to calculate each customer’s R, F, M value, at lastI use the RFM model to determine the valuable level of each customer. I utilized thecollaborative filtering techniques which based on the customers’ preferences torecommend customers of different price categories with products. This paper used theundifferentiated Fuzzy C-Means Clustering to cluster the attribute information ofcustomers to compare with the effects of recommending, and then conducted the personalized recommendations to different categories of customers.In the empirical part of the paper, I used the sales record data of a large market inone year to conduct the actual operation of the above theory, and verified the results.The verification compared the effects of two recommended methods and used therelevant parameters of verifying the quality of the recommendation in therecommended method. By comparing the results of the recommendation which basedon the RFM model and the fuzzy C-means clustering, it is clear that the methodproposed in this paper has higher accuracy and F1values, according to this, I can drawthe conclusion that we are able to provide customers with the right products.