University Students' Travel Mode Detection Research Based On Smartphone GPS

With the rapid development of the scale of higher education in our country,the relocation of unversities to the suburbs has become a common phenomenon.But the transport infrastructure of the suburbs is difficult to meet the travel need of teachers and students of these universities.In order to reduce the impact of a large number of university students' travel on the surrounding traffic and even the urban transport network,it is necessary to carry out the optimization work of the transportation network and conduct a survey for university students.Because the nature of the traditional travel survey methods is dependent on the visitors' itinerary memory and subjective cognition,there are many problems in the survey data,such as the heavy burden of the respondents,the low response rate,the poor data quality,fail to report and miss report and so on.Travel survey based on handheld GPS device has two shortcomings,such as high survey costs and travel data missing due to forgot to carry handheld GPS devices.With the continuous development of mobile technology,GPS positioning system has become a standard of the smartphone.Now,it has become a daily habit for university students to carry their own smartphone when they go out.Thus,using the smartphone GPS to investigate university students' travel can not only reduce the cost of investigation,but also reduce the occurrence of missing data.A new way to analyze the travel behavior of university students is to extract the travel information and identify the travel mode from the trajectory data recorded by smartphone GPS.(1)This paper studies the investigation program and implementation plan of travel modes of university students,and carefully analyzes its advantages and disadvantages.These parts can provide the data guarantee for the processing of trajectory data of university students and travel mode detection.(2)Processing the trajectory data of university students recorded by smartphone GPS:firstly,preprocessing the trajectory data,including data filtering and data format conversion;secondly,selecting two parameters(the dwell time and the average velocity)in the GPSsignal loss case and selecting three parameters(the critical distance,the minimum dwell time and the maximum dwell time)in the GPS signal normal case,and using the five parameters to identify travel segments based on hybrid method;thirdly,extracting travel characteristic variables and using box plot method and tests of equality of group means to verify effectiveness of selected variables.(3)The travel modes of university students are detected.Considering the shortcoming of premature convergence of Particle Swarm Optimization,the Improved Particle Swarm Optimization is used to optimize SVM.And the proposed model is used to identify the six modes(walking,bike,electric bike,campus bus,bus and taxi).The linear kernel function,the polynomial kernel function and the radial basis function are selected as the kernel function of SVM respectively.The accuracy of the travel mode detection of IPSO-SVM model under different kernel functions is obtained,and the highest accuracy is selected as the best accuracy of IPSO-SVM model.The best accuracy of IPSO-SVM is compared with other three common travel mode detection models.The results show that IPSO-SVM proposed in this paper has a better accuracy in travel mode detection of university students based on smartphone GPS.And it has far-reaching significance for the promotion of smartphone GPS in the travel research field of university studenties,scientifically analyzing university students' travel regularity,diagnosing traffic problem around the universities and developing transporation system around the universities.