Research On Log Analysis System Based On User Behavior

At present,with the continuous development of the Internet,the e-commerce platform is growing constantly from the initial simple online shopping to the Internet business.At the same time,it is obvious that the user will focus more on the requirement for a better user experience ever than before.Therefore,the urgent issue is how to better analyze the user's preference through the behavior and finally construct a more efficient and intelligent recommendation system.This dissertation will use user behavior log as the basic data set to complete the digital research of user need by studying the user's personal preference and analyzing the user's potential purchase need.In the research process,two kinds of analysis algorithms will be used to analyze and practice user behavior.On the first aspect,a content-based recommendation algorithm will be created by attribute-based K-means clustering analysis of nearly 10,000 products in a duty-free shopping mall in Hainan Province.Secondly,a behavioral matrix-based collaborative filtering algorithm analysis will be conducted through a three-month user order and browse behaviors of a duty-free shopping mall in Hainan Province.Then this dissertation will establish two mathematical models on user purchasing behavior and product attribute relationship respectively through two different analytic algorithms.In addition,based on the above two different types and dimensions of clustering analysis algorithms,this dissertation will deal the limitation caused by a single algorithm better.Furthermore,Content-based algorithm will effectively solve the problems associated with collaborative algorithms and will provide similar relationships between the products.Finally,the recommended service on the e-commerce platform will be achieved by utilizing the above two algorithms with the Spark platform.The above two algorithms are optimized in the performance of the specific e-commerce platform through the practice in the production environment and constantly modifying the algorithm parameters in practice.At the same time,in order to better monitor the application effect of the algorithm,we analyze the results of the execution of the algorithm,and obtain the effect of the algorithm applied to the platform.As the business grows and changes,it can adapt to new business requirements and meet new business changes by modifying algorithm parameters at any time.Currently,in the production environment,the recommendation service function is implemented only by historical behavior data which lacks the real-time nature of data.To solve the above problems,and in order to provide a better user experience and more efficient system performance,we add real-time flow data analysis and statistics functions in the production environment.Its purpose is to provide users with a better experience.In this article,we will describe the classification of the algorithm and choose two different types of algorithms to implement the recommended services in different services.The results will be analyzed and the effect will be displayed.