Analysis And Prediction Of Marketing Campaigns' Influence Based On E-commerce Platform

Since the outbreak of the novel coronavirus,various e-commerce platforms have been rapidly filling people's lives with the new marketing method of "live commerce",and e-commerce platforms,supported by the unique openness and speed of the Internet,have gradually replaced traditional offline trading and shopping as the main setting for people's consumption.At the same time,the e-commerce platform is committed to planning a variety of marketing campaigns to profit,the impact of marketing campaigns also reflects the number of consumers,as well as the planning scheme is good or bad.Therefore,accurate prediction of user behavior and comprehensive analysis of marketing campaign trends can help e-commerce platforms to screen high-quality marketing campaigns,which is crucial to the development of e-commerce platforms.In this thesis,we model,analyze and study the trend changes of marketing campaigns at the individual and group levels respectively,starting from the entity relationship and behavioral characteristics of the information space of marketing activities in e-commerce platforms.The main research work and contributions of this thesis are as follows:1.At the individual level,a model for predicting user behavior under marketing campaigns is proposed.First,an II2 vec algorithm is designed to represent the usercampaign hidden information and construct a hidden information network for complexity of entity relationships in the e-commerce platform.Second,for the diversity of behavioral features under marketing campaigns,effective hidden features are mined through compressed interaction networks to build user derived behaviors.Finally,for the timeliness of the development process of marketing campaigns,a DE-CNN-based user behavior prediction model is proposed based on the construction of hidden information network by II2 vec,filling in the information network of derived behaviors through time slice overlay,and combining the advantages of convolutional neural network for processing local spatio-temporal characteristics.2.At the group level,a model for predicting the influence trend of marketing activities is proposed.First,the multi-label nature of the platform's users is addressed by using textual representation learning to unify the representation of users.Then,considering the carryover power of key users in e-commerce platforms,an influence propagation network is constructed by learning key information in the marketing campaign-user network.Finally,to address the time-series nature of influence development,the influence propagation network is fused,and a model for predicting the influence trend of marketing campaigns based on key information is proposed through a recurrent neural network algorithm.Finally,through the data set of Chongqing Xinhua Bookstore e-commerce platform:Yuetao to verify.The experimental results show that the model proposed in this thesis is not only able to accurately predict user engagement behavior under marketing campaigns,but also can dynamically depict the trend changes of marketing campaign influence.Therefore,this study plays an important role for e-commerce platforms to grasp user behavior patterns,as well as monitoring and screening of marketing activities.