Traffic Congestion Prediction and Mitigation Strategies Combined with Big Data

Traditional traffic flow prediction models have limitations in dealing with complex road network structures and dynamically changing traffic flows, and cannot fully capture the spatial and temporal correlation characteristics of traffic flow. Therefore, this article explores a new method for predicting traffic flow on urban main roads, which combines a local sensitive discriminant analysis model and an improved deep learning structure. Firstly, based on the local sensitivity discriminant analysis model, the feature extraction of the similarity matrix of the urban main road traffic network is studied, and high-dimensional traffic data is mapped to a low dimensional manifold to achieve effective extraction of traffic features. Subsequently, a deep learning model combining spatiotemporal graph convolution and improved Convolutional Neural Networks (CNN) is developed, which improves the accuracy of traffic flow prediction by analyzing the spatiotemporal correlation of upstream and downstream traffic data. The experimental results show that this method exhibits high prediction accuracy under different traffic conditions and can effectively capture the dynamic trends of urban traffic flow. The research results of this article provide strong support for the design and optimization of intelligent transportation systems. As the training period increases, the accuracy of the training set gradually improves, increasing from 75.2% to 84.7%, indicating that the training performance of the model continues to improve.