Research On E-commerce User Portrait Based On Machine Learning Method

Since the 1960s,with the rapid development of computers,the Internet and big data technology in recent years,all kinds of applications and mobile phone apps are constantly changing our daily life from different directions and from multiple dimensions.At the same time,all kinds of application software,websites,apps and so on are also constantly recording our information,which also shows the explosive growth of multi-dimensional,super volume and multi-depth,Multi-dimensional,super-volume and multi-depth data have opened the transformation of our new era.Entrepreneurs in various fields have gradually realized the deep value of data,and data and software developers have also begun in-depth research on how to use massive data to extract the deep derivative data of more practical value for enterprises.Data products such as guess your favorite products of various mobile e-commerce apps and personalized recommendation of Tik Tok emerge at the right moment.These data models make use of users’ daily consumption behaviors,purchasing preference habits and other data,and use a coefficient data analysis method to continuously iterate the data model to build user portraits,and then provide strong support of digital intelligence and automation tools for enterprise fine operation.For each major e-commerce platform,as well as the application software APP merchants to seize the data initiative,will seize the new era of business opportunities.However,in the past,traditional data collection tools,construction tools,and methods for analyzing the characteristics of target user groups had large deviations in data analysis results and low computational analysis speed.In recent years,scholars at home and abroad have carried out in-depth research and innovation on these problems and iterated out various methods to study accurate user portraits based on a variety of algorithm models.The sample data in this paper are derived from real desensitized user data of an e-commerce APP,and the basic concept,model construction method and optimal algorithm combination model of user portrait are studied.Starting from a single experimental model,compared GBDT algorithm model,Xgboost algorithm model,Random Forest algorithm model,and then combine these three models to form a combination model,through the experiment,the most ideal combination algorithm model,and then used to predict the future e-commerce platform customer purchase intention.According to this,customized portrait of user consumption target group is provided for e-commerce APP platform.Through the experiment,the combination of GBDT algorithm model-Xgboost algorithm model-Random Forest algorithm model can achieve the best user portrait construction,user purchase intention prediction effect.This hybrid model can be applied to e-commerce APP platform to achieve accurate online marketing.