Research On Retail Banking Product Recommendation Model Based On Collaborative Filtering And K-means Clustering

With the popularization of the "Internet+" age,major banks are intensively transforming into retail finance,and the competition in major financial markets is fierce.This paper dedicated to obtain more customer resources,they have to innovate in marketing strategies and channels.It is a sales channel to explore and adopt a combination of online and offline channel marketing strategies,and try to establish a banking retail business product recommendation model that relies on big data to digs out and predict customer preferences to carry out big data precision marketing.innovative direction.Based on customer historical behavior data mining customer demand preferences and hobbies,the bank formulates differentiated marketing strategies and provides personalized services.On the basis of the previous researches,based on the traditional collaborative filtering algorithm and K-means clustering algorithm,this paper constructs a product recommendation model for bank retail regular savings products,wealth management products,loan products,and firstly realizes the intelligent recommendation and big data precision marketing of bank retail products through practice.The main research work is as follows:First of all,in order to achieve full-customer full-retail product recommendation,in the face of the bank’s tens of millions of customers,this paper mainly adopts the commonly used project-based collaborative filtering algorithm as the main algorithm of the product recommendation model.There will be regular savings products,wealth management products,Loan product customers who have purchasing behavior are the target customers of collaborative filtering.For customers who have not purchased regular savings,wealth management,and loan products,the K-means clustering algorithm is used to group customers,and customers who have no historical purchase behavior are recommended through the preference products of customers with historical purchase behavior in the customer group,which can solve part of the problem.Client cold start problem.Secondly,in order to realize the cross-selling recommendation of regular savings,wealth management,and loan products,and increase the cross-holding rate of customer products,we will meet the problem of cold start of items.In the face of new products on the shelves,this article will classify the products,and new products will also belong to them.In the first category,the project-based collaborative filtering algorithm is used to calculate the classified product categories,which can solve the cold start problem of some products,reduce the amount of calculation to a certain extent,and solve some expanded problems.Finally,according to the experimental data of a city commercial bank in 2020,this paper uses RFM binning to construct a user-item scoring matrix.According to the preference indicators including R,F,and M,it is used to calculate the user score to obtain the user’s preference score for the item.R represents the time interval from the customer’s last purchase of a product to the present,F means the customer’s cumulative transaction frequency in the past year,and M is the customer’s cumulative transaction amount in the past year.Through the experimental results,the evaluation results of the collaborative filtering algorithm and K-means clustering algorithm applied to the product recommendation model of bank retail business are obtained,which shows that the product model design scheme given in this paper is feasible.To some extent,the proposed model overcomes the problems of data sparsity problem,cold boot problem and expanded problem of traditional collaborative filtering algorithms.