The Prediction Of Sinking Users On A E-commerce Platform Based On Classification Algorithm

Sinking market is a hot word now,and it is a place for the interests of various fields of the Internet,and e-commerce is no exception.J defines the sinking market as the market for people with weak purchasing power and low income,and the sinking users refer to the consumer groups in the sinking market.The identification of sinking users is conducive to fine operation of the platform,enhance users' stickiness to the platform,and thus create more value for enterprises.Firstly,based on the 22 basic attributes,behavioral characteristics and label attributes of 200,000 users in J,this paper makes a descriptive statistical analysis.Secondly,based on the understanding of the business and the characteristics of the data itself,this paper preprocesses the data.And Several models are used for supervised learning.At the same time,the grid search and Bayesian optimization are selected to optimize the hyperparameters of the models,so as to obtain the final six classification models.Indexes including Accuracy,Recall,Precision,AUC and core index Idx are used to evaluate the effect of the model.Finally,based on the optimal Stacking model,this paper describes the specific application of the model.The core conclusions are as follows: sinking users have lower user rank,lower activity and higher sensitivity to price.In addition,non-sinking users in third-tier cities and above account for a higher proportion,while sinking users in fifth-tier and sixth-tier cities account for a higher proportion.The behavioral differences between non-sinking users of the platform are larger,while those between sinking users are smaller.From the point of view of the core index Idx,Stacking has the best effect,followed by XGBoost.In terms of the Recall index that the business side pays more attention to,Stacking has the best effect,followed by RF.Therefore,this paper suggests that business parties can choose the Stacking model.When the Stacking model is applied to the home page strategy for sinking users,the model has a positive effect on the click-through rate and the order conversion rate under the condition of5% cut amount.