Research And Implementation Of User Profile Based On College Network Behavior Log

With the rapid development of Internet technology,user profile has been widely used in various fields.Its core is to abstract user's characteristics into the form of labels by analyzing various user behaviors.Based on the real college network behavior log,an in-depth study of college network user profile is made in this thesis.In view of the fuzzy description of students' behavior types in college network behavior log,a behavior type label prediction model that fuses compressed DOM tree structure vectors is proposed in this thesis,and it predicts the web pages' types that students browse by fusing the text feature vectors and the DOM tree structure vectors.At the same time,in order to effectively predict students' grade,a FP-growth grade warning label prediction model based on the time series weight of behavior is proposed in this thesis.By mining the strong association rules between students' behavior types and grades in different time periods,the probability estimation of students' grade types can be realized.Based on the above two models,the college network user profile core system is constructed in this thesis.In data acquisition module,the Scrapy-Redis distributed crawler architecture is used to obtain the unknown network behavior data of student users.In behavior type label prediction module,based on the behavior type label prediction model,the function of type prediction is realized.In grade warning label prediction module,based on the grade warning label prediction model,the Spark framework is used to accelerate the mining process of frequent items and association rules.In data statistics module,the multi-dimensional data are statistically summarized based on the data of network behavior and grade level,which provides support for the secondary development.In order to realize the visualization of user profile data,college network user profile platform based on Express and vue.js is designed,and visually displays various data,and ports the platform to We Chat small program based on Kbone framework.The experimental results show,compared with the existing methods,behavior type label prediction model improves the accuracy of category prediction,the grade warning label prediction model has some reference value,the core system of user profile has stable function and performance,and the user profile platform has superior performance and good visualization effect.