On The Cell-ID Positioning Data Based Work/Residence Location Identification And Commuter Trajectory Extraction Techniques

The spatial-temporal distribution and commuting route mining of urban residents has always been a hot research subject.It is of great significance in many applications,such as traffic flow analysis,urban planning,travel characteristics analysis and path recommendation.In the traditional research,the main data source of the urban residents' work and residence is questionnaire survey.However,large-scale questionnaire survey has many problems,such as time-consuming,high data acquisition cost,and the accuracy of the survey is greatly affected by the cooperation willing of the interviewees.With the gradual advancement of traffic information facilities,more and more information acquisition devices can provide us data sources.At present,the rapid popularization of mobile phones,the continuous development and maturity of data mining and large-scale computing technology,mobile cellular network becomes an ideal platform to analyze the travel characteristics of urban residents.In this thesis,the positioning data of Cell-ID based localization system are cleaned and clustered to extract the residents' work and residence places,which will be utilized to restore and analyze the commuting track information of residents.The research work in this thesis mainly includes three parts.Firstly,the sampling and error characteristics of Cell-ID positioning data are analyzed,and the invalid data and noise data are cleaned;Secondly,aiming at the problem of low sampling rate and inaccurate positioning of mobile phone signaling data,a two-step clustering algorithm based on multiday data is proposed.The DBSCAN algorithm is used to cluster the significant places of users by using the regularity of people's travel.After extracting the staying features,the fuzzy CMeans algorithm is used to cluster all the significant places of all users,and the estimation of home and work location is completed.Practical mobile call detail records data were utilized to verify the proposed scheme and the experiment results confirm us that,the proposed scheme provides us an efficient method to identify work and residence location from the mobile phone signaling data.Thirdly,a representative commute trajectory extraction algorithm is proposed for the sparse problem of commute trajectories positioning points.The algorithm calculates the similarity matrix of the user's multi-day commute trajectories according to the relative merging distance,and then uses the improved CATC algorithm to cluster the trajectories to obtain the main commute trajectory cluster.Then a trajectory aggregation method is proposed to fuse the trajectories in the cluster into representative commute trajectory,which can better characterize the user's commuting behavior.To validate and evaluate the techniques proposed,both real and synthetic datasets are used.Experimental results show that proposed trajectory cluster algorithm has better performance compared with other methods.The trajectory aggregation algorithm can effectively improve the number of trajectory positioning points and obtain a representative trajectory closer to the real path.The work of this thesis provides a feasible research idea and technically sound route for the residence and work location,the commute characteristics of urban residents from the CellID based positioning data of the mobile network,as well as the research and application of the Cell-ID based resident travel feature extraction technology.Moreover,this study provides useful insights for the future research and related applications in this area.