Design And Application Of Customer Value Model Based On Massive Data

With more than 8 million retail stores,each retailer generates a large amount of data if it sells only one specification,so the amount of data that all retailers sell across multiple sizes is very large,TB level has now been reached.Therefore,this article uses Hadoop as a platform for big data processing and analysis to improve data storage capacity and processing efficiency.For enterprises,mastering the dynamic retail market can be a good understanding of the market dynamics,the classification of customers can make the enterprise targeted investment resources,to avoid unnecessary waste.However,existing customer classification methods have great drawbacks.First of all,the existing customer classification method is not ideal for the classification of massive data,and the data execution efficiency is very low.Second,the existing customer classification algorithm has a strong subjectivity,classification criteria are more single.In this paper,starting from the needs of enterprises,the retail value of more than 800 million retail customers to study the overall value of retail,create objective and scientific customer value analysis model to improve the efficiency of business-to-customer relationship management.The main research contents are as follows:1)Through market visits,terminal acquisition,system entry and other means of data collection,mass data collected after mass analysis,to take the steps of cleaning,processing and standardization of data processing,the use of HDFS features to build a distributed data Store models,use Hive data warehouse to process data,reduce dimensions and simplify data.2)Based on the analysis of the concept of customer value and the more traditional customer classification algorithm,this paper proposes a set of scientific customer value subdivision model from the perspective of enterprise needs: selecting appropriate value indicators and considering the impact of different indicators on customer value Degree,analyze and calculate the weight of each index,give the detail rules and establish a set of customer value scoring system.Based on the scoring system,a customer value model is constructed:(1)Value classification model to guide the enterprises to differentiate their marketing strategies for different value customers and achieve precise customer positioning and customer relationship maintenance;(2)Loyalty warning model to calculate the loyalty of retail customers Degree,guide enterprises to analyze changes in retail loyalty,prevent customer loss.3)FCM fuzzy clustering algorithm is used to classify customer value.However,FCM fuzzy clustering algorithm has the disadvantages of being easy to fall into local optimum and sensitive to initial clustering center.Algorithm(GWO)and Particle Swarm Optimization(PSO)to improve it,and select an improved algorithm with the best clustering effect through experimental comparison as the final customer value clustering algorithm.4)Enterprise data size is very large,so the use of MapReduce parallel computing framework,the improved FCN fuzzy clustering algorithm for parallel processing and improve the efficiency of the algorithm to handle massive data.5)The customer value model put forward in this paper put into practical application,build a customer value model analysis application system based on big data platform of the enterprise,and provide a visual management platform for the enterprise.