The Research Of Personalized Recommendation For The Entity Retail Business Based On The RFM Model

The online retail industry have brought a certain impact to the entity retail businesses,and the number of customers spending in the entity retail businesses has decreased,resulting in poor sales performance.As the concept of the spending of customer changes,the concerns of spending have shifted from the price of goods to the quality of goods and personalization gradually.In this context,it is necessary to apply a personalized recommendation method to the retail business.The traditional collaborative filtering algorithm based on the evaluation of the product recommendation for customers.However,the entity retail businesses cannot use this method for customers to recommend directly for they cannot get the evaluation information of customers.In this paper,aiming at the status quo of real retail businesses,a hybrid recommendation approach that is suitable for real retail businesses is proposed.Based on the Recency Frequency Monetary Values(RFM)model,the Recency Frequency Profit(RFP)model of qualitative and quantitative combination analysis is proposed,then approach integrates demographics characteristics of customers,collaborative filtering algorithm and rank learning algorithm into the model,and verified the effectiveness of the algorithm through actual data experiments,to give the recommendation for customers on the condition that entity retail businesses cannot get the evaluation of product information.Finally,this paper develops a personalized recommendation strategy for different value customers to help companies get the highest profit with limited cost as soon as possible.The main contents of this paper are organized as follows:First,aiming at the problem that the entity retail business cannot obtain the customers' evaluation of the commodity,this paper calculate the value scores of goods from customers on the basis of the RFM model.Although the RFM model can express the customers' shopping preference by calculating the value scores from customers,the RFM model uses the AHP to define the parameters,which is subjective and cannot be in view of the data.What's more,when the RFM model expresses customers' shopping preferences,it cannot accurately identify those customers who purchase the same amount but bring different profits to the business.Therefore,this paper uses the qualitative and quantitative analysis of the RFP(Recency,Frequency,Profit)to calculate value scores of the goods instead of giving a mark for the foods from customers,and then use the collaborative filtering algorithm to predict and recommend.Through actual data experiments,it is verified that the RFP model based on the combination of qualitative and quantitative analysis can reflect the customer's shopping preferences better,and the recommendation accuracy is higher.Second,in order to improve the accuracy of recommendation products,this paper extracts four statistical characteristics of customer's age,sex,occupation and education level from the basic information of customer registration,and calculates the similarity of the comprehensive features among customers and then combine it with the similarity of customer value which is calculated by RFP model based on qualitative and quantitative combination analysis.The mix similarity of customer is formed,and then the recommendation of collaborative filtering is proposed.The experimental results show that the proposed collaborative filtering algorithm based on hybrid similarity is better than the collaborative filtering algorithm based on the RFP model of qualitative and quantitative analysis.Third,the end result of the recommendation method is to show a list of recommendation products or a kind of ranking recommendation products for the customers.Therefore,based on the above-calculated collaborative filtering algorithm with hybrid similarity,the rank learning algorithm is incorporated.The list of recommendation products based on the collaborative filtering algorithm shall be optimized respectively by Schultz learning algorithm and greedy rank learning algorithm.Finally,the final recommendation list of products shall be given.Experiments show that the rank learning algorithm can improve the recommendation accuracy to a certain degree,and Schultz rank learning algorithm is better than greedy rank learning algorithm.Finally,this paper shall identify different value of customers based on the RFP model.As for different value of customers,it shall propose different recommendations of customers to achieve a truly personalized recommendation in content and form.Moreover,it can help companies determine the input costs on the different value customers by classifying the customers,and then get as high profit with limited cost as soon as possible.