论文部分内容阅读
将神经网络与Markov链理论应用于随机波动的交通流预测,提出一种交通流实时滚动预测方法TDFNM。该方法采用BP网络构建交通流基准预测曲线,使用SOM网络划分残差的Markov链状态,计算各状态加权中心及状态转移概率矩阵,以此预测未来状态,并以加权中点修正计算得到精度较高的预测值,同时实现实时滚动预测。采用方法TDFNM对实测交通流量进行仿真实验,结果表明,该方法比常规BP网络具有更高的准确性,而且具有较强的适应性。
The neural network and Markov chain theory are applied to the forecast of random fluctuation traffic flow, and a real-time traffic prediction TDFNM is proposed. This method uses BP network to construct the traffic flow baseline prediction curve, and uses the SOM network to divide the residual Markov chain state, and calculates the state weighting centers and state transition probability matrices to predict the future state, and calculates the accuracy with the weighted midpoint correction High predictive value, while real-time rolling forecast. The method TDFNM is used to simulate the measured traffic flow. The results show that this method is more accurate than the conventional BP network and has a good adaptability.