Research on Traffic Flow Prediction and Traffic Light Timing Recommendation Technology Based on Vehicle Data Analysis

In view of the fact that traditional traffic signal systems cannot provide dynamic and flexible timing schemes for modern high-volume urban road traffic, this paper predicts road traffic flow from a global perspective and provides reasonable strategies for traffic signal timing based on this. By analyzing data to predict future road traffic flow and providing reasonable strategies for corresponding traffic signals, this paper proposes a time series prediction method based on recurrent neural network(TSPR). To reduce prediction errors, multiple segmented predictions were performed, and the selection of relevant parameters was determined through simulation analysis. The accuracy of the TSPR algorithm was demonstrated by comparing its prediction results with those of SVR [1], CART, and BPNN [2], and the rationality of multiple segmented predictions was demonstrated by comparing them with one-time multi-segment predictions. Based on the TSPR prediction results, in order to rationally set up traffic lightsGreen time ratio to improve the overall income, this paper combines the prediction results with the DQN [3] algorithm and applies it to the field of traffic light control, proposing a traffic light timing recommendation model based on prediction. Compared with the traditional DQN algorithm, the overall return of the DQN algorithm can be improved after the traffic light timing is recommended by TSPR prediction, thereby achieving an increase in benefits.