Research On Evaluation Of Acquisition Effect Of Travel Information Based On Cellphone GPS Trajectory Under Different Traffic States

In recent years, with the speeding up of urbanization in our country and the concentration of population, traffic congestion problems become increasingly highlighted, transportation basic data collection are the key important means to solve the urban traffic problems. Resident travel survey is not only a basic approach to grasp the traffic demand of urban residents, but also the important basis for the related traffic departments to make scientific transportation planning schemes and traffic policies. Cellphone GPS technology as a new type of resident travel survey technology, compared with the traditional resident travel survey, the survey data has the characteristics of real-time, dynamic, and greatly improve the investigation efficiency and data quality, thus cellphone GPS survey has obvious advantages, and has become a hotspot among research scholars at home and abroad.In this paper, cellphone GPS technology is used to collect individual travel trajectory data under different traffic states, then one pattern recognition algorithm is applied to evaluate the effect of data collection. Firstly, this paper designs individual travel experiments under the free flow state, general congestion and severe congestion, besides a method is proposed to use Internet map to investigate real-time road traffic states information, meanwhile record the actual travel logs in travel survey, in order to lay a solid foundation for subsequent travel information identification and data collection evaluation.Secondly, GPS trajectory data characteristics are extracted based on the mathematical statistics and data mining algorithms, and support vector machine algorithm is put forward to track and identify individual travel information, combined with the theory of support vector machine algorithm and the mathematical model, then this paper studies step length, the selection of characteristic parameters and model optimization of kernel function parameter values, thus build the optimal travel information identification model, and the model is used to identify travel mode via GPS data under different traffic states, on this basis, traffic transfer information is extracted between different transport modes.Finally, this paper mainly makes an evaluation on two aspects regarding transportation mode and transfer point recognition error:under different traffic states, the support vector machine algorithm can accurately identify walking-bus-walking of the combination data, however, only by using the support vector machine algorithm for identifying walking-car-walking combination data, the transportation mode recognition error is bigger, but map matching algorithm can modify the result, and the final recognition accuracy is increased greatly. In addition, the study is shown that different traffic states have no significant effect on transfer point recognition, the transfer point recognition error is within a reasonable range compared with the traditional resident travel survey.