短时交通量时间序列的小波分析-模糊马尔柯夫预测方法

基于短时交通量时间序列的随机波动特征，提出一种小波分析和模糊马尔柯夫结合的预测方法．首先对交通量时间序列进行多分辨率小波分解，然后对低频部分和高频部分分别进行重构，对重构后的基本信号和干扰信号建立模糊马尔柯夫模型，最后对多个预测结果进行合成，从而得到交通量的预测结果．此外，根据灰色系统理论的新息优先原理，实时更新马尔柯夫预测模型中的状态转移矩阵，进一步提高预测精度．通过对苏州某交叉口短时交通量预测，表明小波分析和模糊马尔柯夫结合的预测方法具有良好的抗干扰能力和容错能力．

Short-term traffic flow time series forecasting based on wavelet analyses-fuzzy Markov prediction model

Based on the dynamic and stochastic characteristic of short-term traffic volume, an approach combined wavelet analysis and fuzzy Markov forecasting model is put forward. First, wavelet multi-resolution decomposition is applied to the original traffic flow time series, and low frequency signal expressing basic trend and several high frequency signals expressing disturbance are reconstructed, then several Markov forecasting models based on these reconstructed signals are built, that are low frequency signal and high frequency signals. Finally, the forecasting results obtained by the above Markov models are integrated to get the final result. In addition, according to new information priority principle in grey system theory, state transform probability matrixes in Markov forecasting model are update dynamically, therefore to improve the precision. The approach was employed to forecast short-term traffic flow of an intersection in Suzhou city, and the result of experiment shows that this hybrid model of wavelet analysis and fuzzy Markov forecasting has favorable capability of resisting environment disturbance and favorable fault tolerance.