Customer Value Classification And Accurate Recommendation Research For Mass Retailers

Recently,with the rapid development of "Internet Plus",users have accumulated a lot of data in various industries.Big data has become a hot topic in the Internet industry.Facing the FMCG industry,the relevant retail trade data has reached the PB level,and with the incessant depth of the reform,how to use the massive retail trade data to carry out precision marketing for customers plays a very important role in the sales process of enterprises.Therefore,based on the research of 5 million retail customers nationwide,this paper has collected,stored and processed the basic information of customers,including orders and other relevant information with the sampling method.Next,we build the value classification model from the current value,potential value and loyalty value,so as to classify the value of retail customers,and then use the improved recommendation algorithm to make personalized and accurate recommendation for different categories of retail customers.Finally,we apply the relevant theory in the actual production,build a business intelligent marketing system for retail customers,and achieve personalized and accurate marketing for retail customers.The specific research contents are as follows:1)In the "Internet plus" environment,the collection and processing of retail trade data.In view of the wide range of data sources,large amount of data,clutter and other characteristics of enterprise retail,we use customer visit,terminal collection,business system and other multi-source information perception methods to integrate the data.We further clean,integrate and transform the massive retail trade data after integration.At the same time,we use Spark,the mainstream big data processing framework,to store,process and calculate the massive retail trade data.It realizes data storage and resource sharing in big data environment.2)Research on customer value classification model based on FCM.Based on the study of customer value theory and traditional customer classification,we extract three-dimensional indicators of the enterprise's massive retail customers from the actual situation of the enterprise,and build the value evaluation index system.At the same time,we use the network analytic hierarchy process to determine the index weight and calculate the retail customer value.Then,the improved particle swarm optimization FCM algorithm is used to build a customer value classification model to achieve the value classification of a large number of retail customers.Compared the improved FCM algorithm with the traditional K-means clustering algorithm and the traditional FCM clustering algorithm,the experimental results show that the algorithm improves the accuracy and stability of classification.3)Research on the optimal recommendation algorithm based on bi-directional clustering and improved similarity.On the basis of customer value classification,a collaborative filtering recommendation algorithm based on bi-directional clustering and improved similarity is proposed,aiming at the problems of data sparsity and computational complexity existing in traditional collaborative filtering recommendation algorithm in big data environment.Firstly,the algorithm clusters attribute from two directions:user and project dimension,respectively;secondly,it uses the improved similarity calculation method to collaborative filtering and recommendation in the target user and the target project cluster;finally,it uses the balanced factor to comprehensively predict the score and form the final recommendation list.The experimental results show that in terms of the average absolute error,the algorithm(DCF)is 15.8%,8.1%and 7.3%lower than the traditional collaborative filtering recommendation algorithm(TCF),the user clustering based collaborative filtering recommendation algorithm(UCF)and the project clustering based collaborative filtering recommendation algorithm(ICF),respectively,which improves the accuracy of recommendation.4)Research and implementation of business intelligent marketing system for retail customers.The related theory proposed in this paper is applied to build a retail customer oriented business intelligent marketing system based on micro service architecture in actual production.Firstly,the application and technical architecture of the system are explained,and then the core modules of the system are designed and implemented.The application results of the enterprise show that the terminal retail customers have been better classified,the enterprise profits and customer satisfaction have been improved to some extent,which shows that the precision marketing proposed in this paper has played a good role in the practical application.