Research On Customer Classification Of Food Chain Sales Based On Data Mining Technology

With the development of China’s economy and growth of resident consumptionlevel, people’s consumption ways become more and more diverse and recreationaland the leisure food much more popular among the daily food consumption."SnackFamily" is the largest leisure food chain store situated in Zhuhai. During themarketing process, the enterprises should not only analyze the "quantity", but also"quality". Therefore, it is of great importance to subdivide the customers of the foodchain sales and analyze their types and characteristics, which will play a significantrole for the enterprises in figuring out the strategy of customized marketing andmaintaining the customers.Based on analyzing the three important customer classification models bothhome and abroad, it is proposed that customer Average Monetary, Frequency andHold time can be used as segmentation variables to assess the customers’ value.Moreover, with the new AFH customer classification models created, customertrading data of the food chain sales already preprocessed and created databasewarehouse for the theme of customer classification completed by means of the SQLserver And A-value, F-value, H-value and AFH synthetical values will be studiedthrough the cluster analysis from the perspectives of both current value andappreciation potentials. In conclusion, there will be not only completed customervalue matrix based on their customer life profit (CLP), but also four different types ofcommercial strategies to be put forward in the paper.To efficiently eliminate the randomness factors in the process of the traditionalK-means cluster initialization and enhance the stability of the customer’s cluster, theresearch first takes advantage of hierarchical cluster technology (e.g. Two Step) toinitialize the K-means cluster centre and then finishes the data mining of the customerclassification through K-means cluster calculation during the cluster analysis. It isfound in the experiment result that new AFH customer classification models possessgreat evident characteristics, which can highly reflect the current value (contributionlevel) and appreciation potentials (loyalty level) of the customers to provide thoseenterprises with useful information for their decisions.