Research For Traffic Flow Forecasting Method Based On De And Wnn

With the ever-increasing pace of economy development and urbanization, economic loss caused by traffic congestion demands accurate traffic flow forecast and reasonable distribution of existing road resource. Over the years, we can see intelligent transportation system has emerged in major cities of China. To implement the system, accurate traffic flow forecast is one of the keys. Hence experts and scholars around the world have been dedicated in developing models applied to short-time traffic flow forecast which brings out essential practical significance.On the basis of Chaos theory, this thesis analyzes the chaos characteristics and identification methods of short-time traffic flow, which introduces phrase space reconstruction. Built on C-C method, time delay τ and embedding demission m are confirmed which provide the data relationship for further study. Meanwhile, small data method is brought in to conduct experimental simulation. The measured data verified the chaos existence and predictability of short-time traffic flow.The thesis analyzes the WNN as the foundation on which Phrase Space-WNN Model (WNN based on Phrase Space for traffic flow forecast) is built. Serving in distinguishing the traffic flow forecast cases between holidays and workdays through the experimental simulation of measured data, and comparing with Phrase Space-BPNN Model and Time series-WNN model, Phrase Space-WNN Model proves to possess better predictability.Since the evolutionary algorithms are widely used in complicated fields of forecast, it’s brought in traffic flow forecast and shows high complexity and nonlinearity. This thesis improves the Differential Evolution and proposes a Bi-Self-adaptive Differential Evolution (BSDE) algorithm whose performance is tested by standard functions. Combining BSDE with WNN, BSDE-WNN Models for traffic flow forecast is built. Comparison among Phrase Space-WNN Model, Phrase Space-DE-WNN Model and Phrase Space-SaDE-WNN Model also indicates that Phrase Space-BSDE-WNN precedes in excellent predictability, which offers a new method and idea to short-time traffic flow forecast.