Multi-Source Correlation Data Traffic Prediction Based on TextCNN and BiLSTM Model

Predicting traffic flow at different highway nodes can provide operational staff with basis for control measures. Current model only predicts the traffic flow for a single node and cannot capture the traffic flow information of different related nodes.To tackle this issue, this paper presents a model that combines the TextCNN and BiLSTM architectures while incorporating features derived from multi-source correlation data. Through online testing on the Hangzhou-Shaoxing-Taizhou highway (Taizhou Section) in Zhejiang Province,The experimental results demonstrate that this model effectively captures traffic flow characteristics from neighboring nodes. which is better than other models. Specifically, compared to the Historical Average Algorithm, the RMSE result improved by approximately 36.29%, and the MAE result improved by about 31.59%.