Short-time Traffic Flow Prediction with Integrated Passenger Information Using Neural Networks

Traffic flow prediction is a very important and classic problem in intelligent transportation that has received considerable attention from scholars. Previous studies on traffic flow prediction have primarily relied on taxi data. However, while such studies have focused on taxi traffic flow information, they have often overlooked the importance of passenger data. Incorporating taxi passenger information can help uncover key traffic flow characteristics, such as road traffic conditions and local road resource demand. To this end, this paper proposes a CNN-BIGRU model that leverages both taxi passenger and taxi flow information to achieve more accurate traffic flow predictions. The proposed model transforms the taxi flow and passenger data into a feature matrix, which is then fed into the model. Specifically, a convolutional neural network (CNN) is utilized to mine the spatial information of traffic flow. The extracted spatial information is then input into a bidirectional gated recurrent unit network (BiGRU) to capture the data's time sequence relationships. Furthermore, the extracted time and space features are integrated to achieve the goal of accurately predicting short-term traffic flow. Experimental results reveal that our proposed model outperforms competing models on real-world datasets, demonstrating the efficacy and superiority of combining passenger data into traffic flow forecast analyses.