Semantic Mining For Travel Information

In recent years, along with the increasing of urban population, urban traffic management and planning is faced with great challenge. Residents’ travel activity is a very important part of urban traffic. Residents’ travel activity information is the most important reference basis of traffic planning, traffic management and residents’ travel behavior research. At present, the residents’ travel information mainly comes from residents’ travel questionnaire investigation. This method has a lot of problems, such as: taking a long time, spending more money, getting inaccurate information and so on.Along with the development of mobile phones, radio technology, GPS(Global Positioning System) and GIS technology, collecting a large number of residents travel data rapidly and accurately has become possible. Travel information semantic mining method based on GPS data arises at the historic moment. In this paper, the research of the travel information semantic mining is mainly studied from three aspects: Trip identification, Based on the multiple spatial scales deduce trip purpose, Based on the Markov models predict the next stop.(1)Trip identificationThe purpose of trip identification is to transform the original trajectory into stops and moves, which can be directly identified. This paper combines the geographic information system and the trajectory. On the basis of DBSCAN algorithm, some variables have been changed to identify the traveler’ travel trajectory and low velocity area. Let the low velocity area match the GIS map to make further judgment.(2) Based on the multiple spatial scales deduce trip purposeOn the basis of the theory of multilevel space scale, GPS trajectory is analyzed in the aspects of micro level. The algorithm was implemented by identifying the sub-stops from track stops, mining the semantic information of sub-stops, and quantifying the information through using the characteristic parameters of activity points(such as time, speed, corner). Additionally, the types of sub-stops activity was obtained by contrasting the characteristic parameters’ value to the knowledge database based on the statistical results of a large number of data.(3) Based on the Markov model predict the next stopThe existing studies of traveler’ information semantic mining is mainly on the analysis of existing data and has rarely research on the future travel prediction. This paper uses Markov model to excavate travel semantic information to predict the future of the traveler’ travel conditions. The main idea of this method is to divide the travel track state level and establish a state transition probability. And then the next stop position of travelers can be predicted.