Design Of User Portrait System Based On Big Data

The rapid growth of Internet users means that we have entered the era of big data.Along with the growth of Internet users,there is also a large amount of user data.Enterprises are eager to analyze valuable content from user data.In this paper,through the analysis of user behavior and consumption habits through big data,this paper constructs the intermediate medium of user portrait,so that personalized recommendation,precision marketing,data mining,effect evaluation,customized service,etc.can be achieved.With the continuous growth of user data,the scope and dimensions of user portrait models will become more diverse.This paper implements multi-dimensional labels such as rule matching,statistical and data mining based on the data of a mall to form user portraits.The contents of this study are as follows:1.This paper analyzes the construction process of user portrait,selects reasonable technical components to form the overall structure of the project,and provides technical support and development direction for the subsequent functional development of user portrait.2.This paper proposes a naive Bayesian classification algorithm based on the improved EM algorithm.The scheme of replacing the Euclidean distance with the grey relational analysis method makes the selection of the initial value more reasonable,and the filling effect of missing data is better,thereby improving the classification accuracy..3.Construct basic attribute model and user behavior model through improved Naive Bayes algorithm and analyze the prediction results.4.The Spark computing engine and Scala language are used to realize the development of user tags.It mainly introduces RFM combined with Kmeans algorithm to realize user value label,and Newton's cooling law combined with TF-IDF algorithm to realize user interest preference label.