Research On The Application Of E-commerce Customer Segmentation And Customer Churn Prediction

Due to the "Internet +" boom,China's Internet economy has experienced explosive development.Accurately identifying potential customers and high-value customers has become one of the key tasks of many Internet e-commerce enterprises.Customer segmentation technology is an important means to help e-commerce enterprises segment customer groups and provide quality services to maintain customers.E-commerce enterprises can obtain information about customer value through customer segmentation,so as to formulate corresponding marketing strategies to improve customer experience.However,the maintenance of existing customers is no longer enough to gain a great advantage in today's market competition,and how to retain the customers that will be lost has gradually become an important concern of e-commerce providers.Therefore,the prediction of customer churn based on customer segmentation has gradually become a research hotspot of major e-commerce platforms.This article will study the following two aspects of the subject:The first is the study of customer segmentation models.This paper proposes an RVMF model based on the average transaction interval,the average consumption amount,and the average number of product views.Experiments prove that the model solves the problem of collinearity between consumer purchase frequency and purchase value compared with the traditional RFM model,and applies the analytic hierarchy process to optimize the division of old customers and the segmentation of high-value customers in online stores.Optimize the division of potential customers based on product views,the improved customer segmentation model has improved the accuracy of segmenting existing customers.Secondly,to improve the accuracy and precision of customer churn prediction.This article proposes a two-level fusion structure customer churn prediction model.This model does not require one-hot encoding of the dataset in advance,avoiding dimensional disaster and data sparseness problems.The main idea is to combine several high-accuracy tree-based machine learning algorithms to form a two-layer prediction model including Stacking layer and Voting layer.After processing,the data set is input to the Stacking layer,and then the predicted result of the Stacking layer and the processed data set are passed to the Voting layer,and the Stacking layer is added to the prediction of the Voting layer,and finally the predicted result is output.Experiments show that the accuracy and precision of the two-layer fusion customer churn prediction model are improved by 8.81% and 23% on average compared with other models,and the time consumed by the model prediction is within the acceptable range.Finally,in order to solve the problem of data reduction caused by data cleaning during customer segmentation,the two-layer customer churn prediction model was applied to the input data set of customer segmentation in combination with the research contents of the two aspects of this paper,and the predicted data was carried out customer segmentation experiment.The experimental results show that with the two-layer fusion customer churn prediction model,the effective data of customer segmentation model is guaranteed and the segmentation results are more reliable.