Real-Time Short-Term Traffic Flow Forecasting Based on Process Neural Network

Abstract

Existing short-term Traffic flow forecasting models have not fully considered the characteristics of spatio-temporal process and online analysis, we imported the process neural network which can model spatio-temporal process well into short-term traffic forecasting. The model use wavelet radix as weighted function expanding radix of process neurons to deal with the inputs on multi-scale. By using principal component analysis to consider the space affect of traffic flow, the model was optimized. In addition, online learning algorithm of the model was proposed. The experimental results show that the forecasting accuracy of the model is better than ordinary neural networks, and the model can meet the demand of real-time forecasting of short-term traffic flow.