Research on dynamic classification of cigarette retailers based on high-dimensional precise profiling and deep learning
[Objective]This study aims to enrich the evaluation dimensions of cigarette retailers,establish the scientific standards for dynamic identification,and assist the marketing works such as grade adjustment,classification management,and delivery strategy formulation.[Methods]this paper introduces the overall solution for cigarette retailers identification,which is based on a high-dimensional precise profiling system and the deep learning model.Some effective strategies have been put forward,such as deducting the grade for retailers,designing the indicators of portrait,building the identification models,predicting and evaluating new samples,adjusting grade dynamically.In addition,by analyzing the importance of features,the potential influencing factors of retailers are excavated.[Results]In the numerical experiments,based on marketing data from multiple launch cycles in Markets A and B,a comparison of the classification recognition effects of the nonlinear deep learning DXGBOOST model with the linear regression LR model and gradient boosting XGBOOST model is presented.Compared with the linear LR,the DXGBOOST algorithm can increase the grade identification accuracy of retailers in Markets A and B by 9.09％and 15.43％respectively.Compared to the nonlinear XGBOOST,it further enhances the accuracy of precise grades and grade intervals,reducing the degree of misjudgment bias.[Conclusion]Practice has proven that,while ensuring stability,the grade identification method based on the DXGBOOST model can greatly improve the grading accuracy of cigarette retailers,providing data and theoretical support for timely conducting scientific,precise,and intelligent marketing analysis and business decision-making.
万方数据
维普
