The Application Research On User Roaming Group Behavior In Campus Wireless Network Based On Clustering Method

In recent years,with the rapid promotion of education informationization,the gradual construction of the wlan in colleges and universities,as well as the wide use of mobile intelligent terminals,the number of wireless users in colleges and universities has shown a surge trend,which accumulates massive wireless network data.Different from the wired network,wlan users have more obvious mobile features.And current wireless network supports user seamless roaming,which will generate a large amount of roaming trajectory data.Extracting valuable information from these data to help make a decision becomes a huge challenge.Therefore,studying the application of wireless user roaming group's behavior is extremely important.Based on the authentic data of campus wireless network(Beijing Jiaotong University),this paper analyzes the mobile users' roaming behavior by data mining and reaches a conclusion of mobile users' characteristics and clustering results.First of all,this paper analyzes current research status of the mobile users' behavior trajectory based on wireless LAN and introduces the commonly used data collection methods and data pre-processing technology.Then,expands the definition of mobile user roaming trajectory vector.This paper proposes a method to calculate the similarity of wireless users based on the total number of AP encounters.Clustering algorithm based on Hierarchy is chosen to cluster wireless users,and the process of clustering algorithm is designed and implemented.Secondly,a real campus wireless network measurement platform is designed and implemented,mainly adopting these three collecting methods--syslog,SNMP and terminal emulation telnet.Data is collected from a number of wireless vendors AC.After structured processing,data is stored in MySQL database,which will present the status of wireless network operation and user's mobile behavior characteristic graphically.Finally,we selected Dartmouth College wireless network data and BJTU wireless data as the experimental subjects.After pre-data screening criteria and filtering,the two data sets were respectively calculated based on users' similarity.Based on users' mobile roaming trajectory vector,clustering algorithm is used to analyze experimental results among WLAN users.The applicability of clustering algorithm based on wireless user roaming trajectory vector is verified.The behavior of wireless users is studied and analyzed according to the clustering result.The example of wireless subscriber data in Weihai Campus of BJTU verifies the cohesive clustering algorithm based on mobile roaming trajectory vectors of wireless users.Similarly,wireless users are clustered into clusters of different sizes to achieve the purpose of verification.This study helps to understand the operational status and rules of campus wireless network and provides a reference for wireless network management,network performance optimization and later network planning.Based on cluster analysis of history data for campus wireless network roaming,similar users can be dug out based on user group's mobile behavior,which can push personalized and intelligent services for similar users precisely.