Research On Multi-dimensional Mobile User Behavior Pattern Mining And Application

Mobile user behavior analysis is one of the most popular research areas.In recent years,the rapid development of mobile telecommunications,smart devices and appli-cation softwares introduced 4V(Volume,Variety,Velocity,and Value)characteristics to cyberspace,and caused several serious issues such as information overloaded,dimen-sionality curse,etc.On the other hand,the rising mobile internet generates abundant of spatial-temporal and content data,making it possible for user preference analysis at a fine-grained and systematic level.UDRs(usage detailed records)not only include plen-tiful spatial-temporal information,but also contain users;activities in content space,which makes it favorable for the research of human behavior dynamics.User serves as subject in various online activities.Temporality,spatiality and content are fundamental elements to describe relevant behavior.To solve the problems mentioned above,we should mine users' behavior patterns from massive data both in physical word and cyberspace,and build a ternary unified information system including human society,physical word and cyberspace.Therefore,mobile user behavior analysis should take all these fundamental elements into account.In particular,mobile user behavior analysis should mine the patterns among temporality,spatiality and content,and the hidden relation among those basic dimensionality.Therefore,this thesis conducts mobile user behavior analysis based on UDRs.Specifically,we investigate multi-dimensional behavior patterns from the aspect of tem-porality,spatiality and content,and the relation among them at user and collective level respectively.Practically,based on the process of mobile online activity,we study how to find users' required content in massive data and the inverse process that how to de-livery the required content to mobile users.The main research contents and innovations of this thesis are listed as follows:Investigating mobile users' behavior patterns from multi-dimensional perspective.Based on a novel multi-dimensional pattern mining framework,we figure out the patterns and the difference of two typical users on temporality,spatiality and content.The learned patterns can benefit mobile network operators,smartphone manufacturers,and internet service providers.· Modeling the relation among temporality,spatiality and content on user level.Based on the fact that human behaviors are highly predictable and centralized,we propose a feature Transformation method based Central Behavior(TCB)to construct informative feature sets.TCB is informative and effective compared to traditional spatial-temporal features.The trained classification model bridges temporality,spatiality and content,which can be used to user interest acquisition in the era of mobile big data.· Investigating the relation among temporality,spatiality and content on collective level.To meet this challenge,based on the physical interpretation of the target of fusion,we propose a multi-source data fusion method to integrate users3 similarity on temporality,spatiality and content into a single similarity network.The pro-posed method can reflect the significance of original dimensionality by adjust the distance between respective similarity matrix to the fused one.Compared to clas-sical method,the proposed method can cover more original information without increasing the spatial complexity in computation.Moreover,this thesis provides a practical application concerning the fused similarity matrix.We utilize the com-munity structure learned from multi-source data fusion to obtain the collaborative base stations in one community.After an optimization process,the cache files on each base station can be confirmed.The proposed caching strategy can promote hit ratio as well as caching efficiency.Since the performance promotion is stationary on temporality,the proposed caching strategy is favorable for mobile big data.This thesis concentrates on basic elements in user behavior,namely temporality,spatiality and content,and systematically investigate the mobile user behavior from multi-dimensional perspective both on user and collective levels.Our works provide new insight for human behavior dynamics,and the designing and application of future mobile telecommunication networks.