Research On Activity-Based Analysis And Prediction Of Residents’ Travel Behavior

With the rapid development of social economy, travel demand and trip purpose exhibit a great diversity, along with much more choice and complexity of travel behavior. The traditional trip-based demand forecasting theory represented by 4-step model aim to evaluate whether the long-term transportation infrastructure can meet the traffic demand in the future. However it is incapable on problems like’how the traffic demand generated’,’how people’s travel behavior impacted by their social and economical properties’and’how a short-term or even real-time transportation system and demand management measure affect people’s travel behavior’. Therefore, activity-based demand forecasting theory provides a new direction for analysis and forecast resident’s travel behavior and evaluate short-term transportation system manage and control measures.In this context, this thesis studied the theoretical reliability and technical feasibility of the application of activity-based demand forecasting method in analyzing and forecasting domestic resident’s travel behavior with core concept of activity chain theory and research tool of Probabilistic Neural Network(PNN). Combining with the characteristics of domestic residents’travel behavior, this thesis provides a computer program to extract the whole day activity chain based on resident trip survey data of Longquanyi district. By doing statistics of residents’activity chain for typical activity pattern, length and structure of activity chain, it analyzed the basis traveling situation of Longquanyi district. On this basis, this thesis used PNN which do well in learning and pattern recognition to simulate the nonlinear mapping relationship between residents’basic properties and travel choice behavior and establish models to forecast residents’activity patterns, travel time and trip mode. Then it compared the forecast results with real survey results, and it shows the holistic hit rate of these models are all above 80% which proves the feasibility and effectiveness of PNN in forecasting residents’ travel behavior. In the end, this thesis applied these established PNN models to some problems in transportation planning and traffic demand management, and the results proves the practicability of models.