论文部分内容阅读
为了充分利用交通流的时空过程特性,进行交通流的实时预测,将过程神经元网络和数据流在线学习技术引入到短时交通流预测中。充分考虑交通流的日周期、周周期等内在特性,结合过程神经元网络和小波变换,实现对历史数据的多尺度过程特征处理。构建了路网整体预测过程神经元网络模型,并采用主成分分析方法,利用交通流空间相似性的影响对模型进行优化。基于Harr小波技术提出具有自适应和实时性预测特征的在线学习算法。试验结果表明:该模型的预测准确性优于普通神经网络,平均百分比相对误差降低6%～8%,预测时间至少降低67%,具有较高的性能,能满足短时交通流实时预测的需求。
In order to make full use of the spatio-temporal process characteristics of traffic flow, real-time traffic flow prediction is carried out, and the process neuron network and data flow online learning technology are introduced into short-term traffic flow prediction. Considering the intrinsic characteristics of traffic flow, such as daily cycle and weekly cycle, combined with process neural network and wavelet transform, multi-scale process characteristics of historical data are processed. The neural network model of the overall forecasting process of road network is constructed, and the principal component analysis method is used to optimize the model by using the similarity of traffic flow. Based on Harr wavelet technology, an online learning algorithm with adaptive and real-time prediction features is proposed. The experimental results show that the model has better prediction accuracy than ordinary neural network, the average percentage relative error is reduced by 6% -8%, the prediction time is reduced by at least 67%, and the model has high performance and can meet the demand of real-time forecasting of short-term traffic flow .