Short-term Traffic Flow Prediction Based on Recurrent Neural Network

Effective traffic flow prediction has important guiding significance to solve the problems of intelligent transportation system. Based on the historical traffic flow data of observation points and deep learning model, this paper proposes a short-term traffic flow prediction model based on recurrent neural network (RNN). In this model, we use three different RNN structural units (LSTM, GRU, SAEs) to realize the model construction and training. Through experiments on real traffic data, we found that the cyclic neural network has a good prediction effect on specific road traffic flow, which can achieve the purpose of short-term prediction, and has a better practicability. Among the three cyclic neural network structures, LSTM has a smaller prediction error.