Research On Real-time Retail Customer Value Analysis Based On Clustering Algorithm

With the development of the economy and the advancement of science and technology,the retail industry has also ushered in unprecedented changes and challenges.Retail enterprises'research on customer value has always been a hot topic.Due to the characteristics of retail products,research on retail customers' consumption habits and consumption behaviors can help retail enterprises grasp important information of retail customers.This consumption-related information can be effective.The local auxiliary retail enterprise makes business decisions and formulates relevant corporate strategies.With the increasing popularity of enterprise informatization,after many retail enterprises have completed the digital transformation,how to use the advanced tools provided by information technology to empower enterprises and guide the daily operation of retail enterprises is a very important challenge for many enterprises.The research work of this paper is as follows:The first work is the study of the current customer value analysis model.Aiming at the current customer value analysis model,it is not enough to characterize the customer's consumption behavior and characteristics,and the customer value analysis model has the problem and limitation of the customer classification method.A customer value analysis model based on clustering algorithm is proposed.The main innovations of this work include:First,by examining the three-tier architecture model of the enterprise system at the present stage,a clustering method of flat architecture is proposed,which simplifies the transmission steps of data in the application server by simplifying the analysis process.The existing analysis process;secondly,the improved customer value analysis model:Clustering-based recency frequency monetary model(C-RFM)model,which improves the recency frequency monetary(RFM)model's discrete processing of customer data in the process of customer classification,resulting in loss of data information;Thirdly,the introduction of two quantitative indicators of the squared sum of error and the contour coefficient in the clustering algorithm improves the limitation of the lack of objective evaluation index of the RFM model.The experimental results show that after the introduction of the quantitative indicators,the C-RFM model has a guiding role in determining the important parameters of clustering and the selection of clustering algorithms.The second work is based on the C-RFM model proposed in this paper,combined with column storage technology,to achieve an analysis system based on memory computing.The main innovations of this work include:first,the clustering method for flat architecture proposed in the previous paper is realized;secondly,the parameter operation of the client and database server is optimized in the analysis process of the flat architecture,and further improved the analysis process;third,the HANA platform is used as the basis for implementing the prototype system.Combined with the experiment results,the improved memory computing system using the HANA platform is compared with the mainstream data analysis engine.Experiment shows that HANA platform performs better in processing clustering tasks and has shorter running time.