Research On Evaluation Of Activity Location Recognition Using Cellular Signaling Data Under Different Base Station Density

With the continuous upgrading of mobile communication network,the geographical coverage of base station is becoming more complete and broad,and the base station density is increasing.Cellular signaling data can more comprehensively record the information generated in the process of residents’ trip,which makes it possible to extract trip characteristics.At present,many scholars and organizations have extracted and analyzed the traffic characteristics of cities at the macro level based on the cellular signaling data.However,due to the fact that the real location in the process of trip can not be obtained,the lack of benchmark data makes the reliability and accuracy of trip characteristics extraction questionable,and limits the development of related applications based on celluar signaling data towards the direction of refinement.As a key step of refined extraction of trip characteristics,accurate extraction of activity location recognition is the basis of subsequent trip characteristics analysis.The positioning accuracy is different where the base station density is inconsistent.The accuracy of activity location recognition is closely linked with the positioning accuracy of the point.Therefore,it is very important to take base station density into account in the procedure,which plays an important role in improving the accuracy.It is also useful for the future development and application of celluar signaling data in the field of transportation.Based on the real individual communication behavior and wireless communication principle,this paper constructs and optimizes "traffic-communication" simulation platform of G city.Secondly,based on analysis of base station distribution and positioning accuracy,this paper proposes a method to measure and calculate the base station density,and designs a simulation trip experiment across different base station density areas.This paper selects STDBSCAN algorithm as the extraction algorithm of activity location,combined with the characteristics of celluar signaling data,it determines that parameter closely related to the base station density in space is the cluster radius parameter Eps.Based on genetic algorithm,the optimal cluster radius parameters under different base station densities are determined respectively,and the law corresponding to "base station density-optimal cluster radius parameter " is fitted and loaded into the original algorithm,so that the optimal radius parameters can be automatically selected when using algorithm to extract the activity location.This paper carries out empirical experiments,GPS data and celluar signaling data are collected synchronously,meanwhile trip log data is recorded.Real celluar signaling data is compared with the simulation one in terms of data volume,occurrence time interval and switching sequence to prove platform’s reliability.Taking the activity location information set in the platform and recorded in the trip log as the benchmark data source,the activity location is recognized by ST-DBCSAN algorithm before and after optimization.Results reflect that ST-DBSCAN can recognize activity location effectively.In addition,the extraction accuracy of proposed optimization method based on different base station density is 84.8%,which is10.1% higher than original algorithm.The multi recognition rate is 3%,which is 2% lower than original one.The average distance error is reduced by 70 meters and the average time error is reduced by 3 minutes.The effect of activity location recognition has been improved in all aspects.