Short-Term Traffic Flow Forecasting Methods Based On Combination Model

Nowadays with the rapid development of economy, traffic congestion happens more and more often and the situation is getting worse. Travel time increases because of the traffic jam which is causing economic loss and environmental pollution. In large cities with millions of people, traffic jams not only slow down the economic growth, but also impact the society development. To address this problem, Intelligent Transport System (ITS) was introduced to China in 1994. ITS uses information technology, communication transmission technology, electronic control technology, and computer data processing technology to collect, process and compute the traffic information. The purpose of ITS is to reflect the traffic condition in real time and accurately. It helps to put the traffic condition under a scientific control.Two important components of ITS are advanced traffic management system and advanced transportation information service system. Short-term traffic flow forecasting is a key technology of them. This thesis studies on original date process and the building of the forecasting models. The contribution of this thesis is as follows:(1) Analyzing the characteristics of the original data collecting from the road coil, proposing a method to discrimination and repair the original data which is wrong or lost.(2) Wavelet-analysis theory is adopted to process traffic flow data in both time and frequency domains. A wavelet noise reduction model is built to recognize and filter-out noise. The effect of using different noise threshold and noise reduction method is discussed in this thesis.(3) A loose-type short-term traffic flow forecasting model and a tight-type short-term traffic flow forecasting model are created in this thesis. Both of them are based on the wavelet-analysis theory and the neural network theory.(4) The models'feasibility and reliability are proved by the experiment using real time data collected from the high way in a metropolis.