Research On The Multi-modal Trip Distribution Model In Mobile Phone Signaling Data Environment

At present,the analysis and application of big data such as mobile phone signaling data in traffic problems has become a research hotspot.But it is currently limited to medium and macro analysis,and has limited application in the extraction of refined travel information and the research of traffic models.Traditional traffic distribution models are also disturbed by influencing factors such as data quality and model parameters.It is difficult to accurately describe the travel habits of urban residents in today’s complex travel background conditions.So,this paper aims at optimizing the traditional trip distribution model in the mobile phone signaling data environment.The core work of the thesis includes three aspects: extraction of travel trajectory information,construction of travel mode identification model and construction of multi-modal trip distribution model.The specific research content and main conclusions are as follows.Obtaining accurate and reliable travel trajectory information is the foundation for constructing a multi-mode trip distribution model.First,the parking point were recognized based on the spatial-temporal clustering algorithm.On the basis of defining a trip,the travel information was identified and extracted according to the spatial-temporal characteristics.Second was the point to point map matching based on probability statistics.Based on the road frequency and spatial distance,the user’s real travel trajectory on the road was obtained by matching the base station and intersection.Finally,the sample label was acquainted based on the association rules mining.According to the association rules like gender,age,departure and arrival time,the trajectory sample data with travel mode labels were obtained,which solved the sample problem in the latter travel mode identification.Fusion of multi-source data to extract feature parameters and build a random forest classification model to extract travel OD of different modes of transportation is the key to building a multi-mode trip distribution model.Firstly,a total of 20 feature parameters were selected from the aspects of trajectory spatial-temporal features and path navigation features,and the filtering method of chi-square test was used to retain the feature parameters.Then a random forest model was built to identify the travel mode into five types: walking,bicycling,electric bike,bus and car.The overall recognition accuracy was achieved at 90.2%.Finally,the travel OD was extracted and expanded.On the basis of discussing the internal relationship between travellers,land layout,traffic mode and traffic distribution,a multi-mode trip distribution model was constructed.First select the appropriate basic model of traffic distribution.Then,after analyzing the impact of land layout and traffic mode on traffic distribution,select appropriate influencing factor indicators to quantify,so as to construct trip distribution models of different traffic modes.Finally,use the OD obtained after the sample expansion to divide the mode OD for parameter calibration and model verification.The experimental results show that the model accuracy of the multimodal traffic distribution has been improved.