User Behavior Pattern Mining Methods In Mobile Context-Aware Environment

In recent years,with the rapid development of positioning and tracking technologies,such as global positioning systems,wireless mobile communication technologies,wireless networks,and the popularization of smart terminals,people’s work and life patterns continue to innovate and change greatly,and at the same time,massive and diverse data resources have been produced.These data contain a lot of information,so it is urgent for researchers to analyze them effectively,apply the mining knowledge to the actual decision-making process.This thesis focuses on the mining methods of user behavior patterns in mobile context-aware environment,the main research contents and contributions include the following aspects:(1)User’s behavior patterns mining in mobile context-aware environment.In order to solve the challenges caused by the diversification and dynamics in individual user’s behavior patterns mining,this thesis proposes a behavior pattern mining algorithm:in a mobile context-aware environment.This method uses a nested key-value model to efficiently fuse and store multi-source heterogeneous mobile context-aware information,and builds a rule-based multi-dimensional sequential pattern mining algorithm MSP and its improved version UTDMSP.The MSP and UTDMSP can discover globally and locally frequent user behavior patterns in interaction behaviors,discover users’ long-term interest preferences and short-term new behavior trends.Experimental results on real data sets demonstrate that the models and algorithms proposed in this thesis can effectively discover users’ long-term patterns and short-term changes,which will help companies carry out personalized precision marketing and promote the sustainable development of context-aware service business models.(2)Top-N high utility individual user’s behavior patterns mining in mobile context-aware environment.In order to solve the problem of ignoring the impact of personal income in mining individual user’s behavior patterns,this thesis takes the mining of high-utility operation behavior patterns of online taxi drivers as an example,and proposes a Top-N high-utility order sequences mining algorithm,which can combine drivers’ current context to recommend the next order of Top-N high expected revenue.Through spatiotemporal clustering of the location coordinate data of the passengers’ pick-up and drop-off points in historical taxi operation data set,the temporal and spatial distribution characteristics of the passengers in the city are identified.The proposed method first constructs a high-utility sequence tree with the current passenger origin point as the root node,and then uses two pruning strategies,node utility and path utility,to reduce the generation of candidate sets.This thesis constructs a dynamic and updated order sequence recommendation algorithm combined with high-utility sequence tree and real-time context to promote the expected revenue of taxi drivers.The experimental results on the real data set show that the proposed high-utility sequence mining algorithm can effectively identify the high-utility order sequence,which can provide a decision-making basis in improving the expected revenue for taxi drivers.(3)Group users’ behavior patterns mining in mobile context-aware environment.In order to solve the problem of ignoring the close relation between group users’ behavior patterns and urban spatial structure in mining users’ behavior patterns,this thesis takes urban residents’ commuting behavior pattern mining as an example,proposes a framework for analyzing the commuting pattern of urban residents,and studies the commuting pattern of urban residents contained in taxi operation trajectories and the relationship between commuting pattern and urban spatial structure.The grids are divided based on the latitudes and longitudes of the origins and the destinations of the taxi passengers,and a density peak clustering algorithm based on the grid heat,named GridH-DPC,is constructed to cluster the origins and the destinations of the taxi drivers.By designing the work residence index,this thesis first divides the clustered regional functional characteristics into residential areas,neutral areas and working areas,and then analyzes the correlation between the commuting patterns of urban residents and the urban spatial structure.Based on the statistical analysis of commuting distance and commuting frequency,the commuting of urban residents is divided into four patterns:high-frequency short-distance,high-frequency long-distance,low-frequency short-distance and low-frequency long-distance,then the spatiotemporal characteristics of urban residents’ commuting behavior patterns are analyzed.The experimental results on real data set show that the analysis framework and clustering algorithm GridH-DPC can effectively mine urban residents’ commuting behavior patterns from multi-dimensional heterogeneous taxi operation trajectory data,and provide decision-making basis and reference for urban planning and urban traffic supervision.(4)Community users’ behavior patterns mining in mobile context-aware environment.In order to solve the problem that traditional trajectory clustering algorithms ignore the semantic relationship between trajectories in users’behavior pattern mining,a semantic trajectory clustering algorithm based on network community detection,named STCCD,is proposed,which can better measure the trajectory similarity from the perspective of the network,capture the local and global relations between trajectories,and obtain higher quality trajectory clustering results.The STCCD first define a generalized semantic similarity function to measure the similarity between semantic trajectories and then construct a similarity matrix.Secondly,by using the similarity matrix,the STCCD builds a network so as to convert the semantic trajectory into a semantic trajectory network,and then utlilizes the community detection algorithms to divide the trajectory network.The experimental results on real semantic trajectory data sets show that the proposed similarity measurement method and trajectory clustering algorithm can effectively capture the similarity of semantic trajectories,obtain the clustering results of semantic trajectories,discover the activity rules and interest preferences of the same interest enthusiast community,and the behavior patterns of community users,and provide assistance for intelligent decision-making such as recommendation based on social networks such as smart tourism.Aiming at the characteristics of trajectory data,such as heterogeneity,sequence,hierarchy,spatiotemporal synchronization and semantics,this thesis proposes some key methods for mining user behavior patterns in context-aware environment,which include trajectory data representation,similarity measurement methods and utility function,frequent sequence pattern mining,high-utility pattern mining and clustering algorithms.The research results for human behavior patterns mining can support companies in making marketing strategies,assist the government in formulating policies and guidelines,and improve the intelligent supervision of social service facilities.