| The rapid urbanization process is an important achievement of China's economic construction and industrialization.However,the expansion of the city,the growth of the urban population and the increase of the car ownership have made the traffic congestion become an important social and economic problem of the contemporary city.Urban rational distribution,road planning and construction,traffic policy formulation and implementation of the current problem is the main means to solve the problem of traffic congestion.The residents' activities and travel patterns determine the traffic demand and have a significant impact on the state of the transport system.It is an indispensable part of urban layout,road planning and policy making to provide theoretical support for improving the efficiency of transport and service level by studying the activities and modes of travel,summarizing and summarizing people's activities/travel behavior characteristics and laws.In last century,travel behavior analysis is mainly based on the manual traditional traffic survey data,the investigation efficiency is low,the number of samples is small,the data quality is poor,the reliability of the data model based on these data is not high,therefore,the development of traffic policy and road planning effect is not good.With the continuous development of technology,road camera,vehicle or portable GPS equipment is widely used,traffic research to further extend to the microscopic field,however,the individual residents of the activities of the chain of fine study,depending on the electronic equipment traffic survey method still can not meet the research needs.In this paper,based on the rapid popularization of mobile phone platform in recent years,this paper designs a new multi-day activity chain node identification method,activity node attribute discrimination method and resident multi-day activity/travel behavior analysis method based on machine learning model for large-scale GPS trajectory data acquired through smart phone.This paper mainly includes the following aspects:(1)Data acquisition and preprocess of resident activity/travel based on smartphone.Recent years,smart phone is wide spread for its natural nature that easy to carry,not easy to forget,always boot,at the same time,smart phones with a strong interactive and computing power to be able to simultaneously complete the automatic GPS track survey and activity chain information acquisition,significantly reduced the burden of investigators and volunteers greatly improved the accuracy of the survey results.This paper makes full use of the above advantages of smart phones,designed based on the smart phone travel survey and data access strategy,developed a GPS track based on i OS and Android platform acquisition applications,the development of the online PR(Prompted Recall)survey tool.This paper analyzes the data of large-scale travel survey,and puts forward the pretreatment method which accord with the characteristics of smartphone activity/travel trajectory data,and summarizes the principle of large-scale data cleaning.(2)GPS track data driven activity chain node identification method.It is necessary to study the activities of residents' activities and travels based on largescale GPS trajectory data,and construct the residents' multi-day travel activity chain from the large-scale GPS trajectory data to excavate the residents' activities/travel behaviors.The first step of constructing the "activity chain" is to effectively identify the active chain nodes,that is,from the continuous,stateless GPS trajectory data,the objects to be investigated are classified as "active" and "travel" track points.In this paper,data-driven method directly from a large number of sample data analysis of the status of GPS track data and travel status GPS track point data difference.Firstly,the characteristics of the degree,velocity distribution and direction change of each trajectory are calculated,and then the binary classification of the trajectory point of the active trajectory and the trajectory is carried out based on the random forest model,and the classification result is analyzed empirically.(3)Activity chain node identification method based on GPS trajectory data.Only know that the respondents in a specific time interval and geographical scope of the activities/travel status,still can't be directly used for downstream residents' activities/travel behavior research.In the era of large data,these activities or travel with what kind of attributes and characteristics,reflecting the survey of what habits and intentions and so on,human investigation is difficult to do.The attribute information of these activity chain nodes also needs to be analyzed from large-scale GPS track data.In this paper,the main problems of attribute identification of active chain nodes are discussed.The characteristics of velocity distribution,number of stops and trajectory distribution range of each activity/travel are calculated by GPS trajectory data.Based on the support vector machine model,the travel mode and activity content Identify and identify the results of the empirical analysis.(4)The structural equation model analyzes the pattern of multi-day activities/travel time use pattern.With the improvement of urban construction,the improvement of traffic conditions gradually from the construction of transport infrastructure into travel demand management policy.On the basis of the residents 'multi-day activity chain information,the author analyzes the principle of residents' daily travel behavior and understands that the activity/travel time distribution mechanism can help the traffic management department to forecast the travel demand and evaluate the transportation policy.This paper examines the data of individual travel behavior for multi days,deeply studies the changes of people's travel time distribution,uses the variance analysis to describe the time distribution of residents in activities and travel behavior intuitively,and model analysis by multi-group structural equation model.The relationship between the social and economic attributes of the survey object and the distribution of the multi-day travel time.The main innovation of this research work lies as following:(1)Based on the large-scale GPS track data acquired from smart phone,a preprocessing method which accords with the characteristics of the smartphone activity / travel trajectory data is put forward,and the principle of the large-scale activity / travel path data cleaning collected by the smartphone is summarized.(2)This paper proposes a method to identify the active chain nodes driven by GPS trajectory data.This method can reduce the subjective intervention of the researcher as much as possible,summarize and reflect the characteristics of the large number of sample data itself,and more through the knowledge extracted from the data,rely on traditional experience for activity chain node identification.(3)This paper proposes a method of attribute identification of active chain nodes based on support vector machine model,which integrates travel identification and activity content identification into the same framework,which lays a solid foundation for the analysis of downstream residents' activities and travel behaviors.(4)Through the detailed statistics and the modeling analysis of the residents' multiday activities / travel time allocation,it is found that the "typical day" of the residents' multi-day trip does not exist,which proves that this article studies the residents' activities through the multi-day travel data/travel habits and rules,is the behavior of residents travel behavior inevitable development direction.The research results of this paper can give full play to the potential of large data of residents' activities/travel GPS trajectories,and more objectively excavate the activity chain information of individual residents and carry on the fine analysis,so as to greatly promote the residents' activity/travel trajectory data in urban layout,traffic planning,road construction,policy formulation and implementation. |
PDF Full Text Request