Traffic Flow Prediction and Intelligent Road Network Optimization Under Artificial Intelligence

In response to the low accuracy of traffic flow prediction and the difficulty in adapting to real-time changes in traffic conditions, this paper adopts a combined model of visual geometry group 16 and gated recurrent unit to predict traffic flow and optimize the intelligent road network. Firstly, this article uses the visual geometry group 16 model to extract spatial features from intersection camera image information, capturing road network structure and traffic vehicle information. Then, the gated recurrent unit model is used to model the temporal relationships in traffic flow data. It fully considers the temporal relationships in the data, captures the long-term and short-term dependencies in traffic flow data, and introduces a soft attention mechanism to dynamically adjust the weights of different positions, improve attention to key positions, and reduce prediction errors. Finally, the deep Q network model is introduced to optimize the traffic lights in the road network and reduce the congestion level at intersections. The experimental results show that the average absolute error of the visual geometry group 1616-gated recurrent unit combination model is only 9.0125, which is 1.3064 lower than visual geometry group 16-long short term memory. After optimizing the road network signal lights, the congestion level is only 2.1%, reducing the prediction error and congestion level.