Design And Implementation Of Topic Search Based Campus User Behavior Analysis System

The number of network users has increased year by year with the development of network technology and services,and the user behavior information is contained in the network.The analysis of the user behavior information of the campus network through data mining can overcome the shortcomings of the traditional university management mode that is inadequate and non-depth.The traditional campus user analysis system mainly includes topic model establishment and association mode mining.The ability of semantic representation of existing theme models are weak,and it is easily to extract popular but useless features by traditional association mode mining,which is not customized.In order to solve the problems,this paper introduces a unique identification method in campus network,and designs and implements a campus user behavior analysis system based on topic search.Compared with other systems of users' behavior analysis,this system adds an improved keyword extraction algorithm and an improved association mode algorithm.The specific research contents are as follows:1.An LDA2Vector based topic word extraction algorithm is proposed based on structural collaboration.This method uses the structure score as the edge weight to build the graph of TextRank algorithm,and adds the semantic information of LDA2Vector on it,so that the extracted keywords have the ability of topic semantic representation.2.An improved algorithm for frequent pattern mining of warning clue words is proposed based on the multi-support of positive and negative association items.The method uses multiple minimum support to improve the accuracy of the association mode,adds negative associations to increase the potential item set in the analysis,and adds a warning clue word filter based on the FP-Growth algorithm,so that the sensitive keyword preset by system managers can be effectively extracted by the system.3.The user behavior analysis system based on topic search is designed and implemented.The system includes four modules:data storage,data processing,algorithm analysis,and UI display.The algorithm analysis module includes topic extraction and feature vector correlation algorithms.The results of analysis will be displayed through the UI display layer.The experimental results show that the system can effectively extract the topics corresponding to the links in the user's visit record,and can correctly correlate with other user feature vectors.Finally,the results of the algorithms are displayed in the UI layer of Web pages,which can provide the basis of management decisions of network administrators.