Research On The Methods Of Important Location Discovery And Application Of Home-work-travel Based On Massive Low-quality Trajectory Data

With the development and popularization of smartphones and mobile Internet,a huge amount of user space location information data has been generated.Analysis of a large number of mobile phone signaling data can better understand user behavior,which is of great significance for providing accurate spatio-temporal recommendation and prediction services.In recent years,user important location mining has become a research hotspot in the field of spatiotemporal data mining.However,the existing work faces many challenges,including the large scale of trajectory data and low accuracy of location.Due to the advantages of wide geographical coverage,large data sample size and low acquisition cost,mobile phone signaling data has become an important data foundation and research object in various research fields of urban transportation.Firstly,aiming at the low precision of mobile phone signaling data,this paper designs a neighboring trajectory interpolation fusion method based on the user sub-trajectory fragments.The user trajectory is divided based on the grid,and the user sub-trajectory fragments are merged and accuracy of the user trajectory,and provide the basis for subsequent important location recognition algorithms.Secondly,on the basis of summarizing the shortage of the existing user's important location recognition algorithm,the mediation matrix for group users is constructed to represent the main activities of the crowd,at the same time puts forward the improved user important location recognition Algorithm of MMA(Matrix base Mining,Algorithm).The user important location is corrected based on mediation matrix optimization to improve the accuracy of the user's important location recognition.Finally,this paper takes the desensitized mobile phone signaling data as the research object,extracts the user's living workspace distribution and completes the application analysis of the occupational residence.This paper conducts an experimental analysis on a spark cluster with 7 nodes.The experimental results show that the user trajectory fusion method proposed by this paper and the user's important location recognition algorithm have higher accuracy.