Analysis Of Urban Residents’ Trip Behavior Based On Smart Phone

With the rapid growth of China’s economy and the continuous expansion of the city’s scale,people’s living standards have gradually improved.The number of urban motor vehicles is also increasing year by year.Such rapid growth has brought convenience to people’s work and life,but also brought many problems to urban transportation.Resident travel survey is an important part of urban traffic planning survey,and it is also an important means to obtain residents’ activities-travel behavior rules.In the past,residents’ trip surveys mostly used paper surveys.The survey results were greatly influenced by the subjective cognition of the survey questions,and the data accuracy was low.In addition,due to the large number of survey items,respondents need to spend a lot of time and energy to fill out the survey report,resulting in higher rejection rate and lower accuracy of answer.With the rapid development of positioning technology,the GPS data survey method based on smart phones provides new data acquisition methods and data processing methods for analyzing residents’ trip behavior.The content of the thesis mainly includes the following three parts:(1)Introduce the survey method of residents’ trip based on smart phone and GPS network.(2)Trip mode identification based on GPS data.Through the R language programming,the collected GPS data of millions of residents are processed,and the trip time,trip distance,speed related features and acceleration related features are extracted as the characteristic indicators of the inferred trip mode;then,BP neural network is used.The algorithm identifies six modes of trip such as walking,bicycle and car,and obtains 90.2%overall recognition accuracy and 85.3%average recognition accuracy.Finally,the algorithm is compared with the recognition accuracy to verify that the BP neural network algorithm has higher accuracy for trip mode identification.(3)Identification based on the purpose of trip of residents’ society and trip information.Combine the collected GPS data information,and observe and determine the land use type of the trip endpoint on the“Map Carefree”software,and select the land use information,trip mode selection and volunteer related information as the characteristic indicators for inferring the purpose of the trip;There are too many characteristic indicators.The Boruta algorithm is used to screen the characteristic variables,and the random forest algorithm is used to identify the five trip destinations of going to work,going home,going to school,general education and leisure entertainment,and obtaining 88.5%of the overall recognition accuracy and 85.8%of the average.Recognition accuracy;Finally,the algorithm is compared with the recognition accuracy to verify that the random forest algorithm has higher accuracy for the purpose of trip identification.The innovation of the thesis is maiuly reflected in the following aspects:(1)Selecting more comprehensively the feature indicators for identifying the way of trip.Based on the research and summary of the previous trip mode identification methods,based on the existing data,this paper selects 10 kinds of feature indicators for identifying trip modes,and each feature index has a clear distinction for the way of trip.It provides the possibility to improve the recognition accuracy of the model.(2)Select the feature variables using the Boruta algorithm.Reasonable feature selection can improve the accuracy of model recognition.In this paper,the Boruta algorithm is used to select the most effective 15 feature variables from the 28 indicators that represent the purpose of trip,which are used for the research of relatively few trip destinations.