关键词: 交通工程, 组合模型, 贝叶斯网络, 交通流, 小波分析, ARIMA算法, BP神经网络

Abstract: Bayesian network is one of the most efficient models in the uncertain knowledge and reasoning field. A method based on Bayesian networks of combination mode is put forward to solve the problem of short-term traffic flow prediction. First, several basic prediction models are used to predict the traffic flow. The prediction results and the actual traffic flow are discretized by certain step length. Then, the parameters of the Bayesian network are updated by learning those data. Through combination of reasoning, every possible value of Posterior probability of each data generated by the results of every basic prediction model can be calculated. Then the largest value of Posterior probability would be the final result of combined prediction. The prediction of traffic flow in real road indicates that the prediction results by Bayesian network combination model are more accurate than single prediction model. It thus proves that the proposed model is applicable for the real condition.

Key words: traffic engineering, combined model, Bayesian network, traffic flow, ARIMA algorithm, wavelet analysis, BP neural network