Machine Learning-Based Short-Term Traffic Flow Prediction Model for Metro Systems

Municipal governments in China have made significant investments in building urban rail transit networks in an effort to reduce traffic congestion caused by the country’s fast economic expansion and hastened urbanization process. Shortterm metro passenger flow prediction, which aims to forecast passenger volumes at metro stations within imminent time frames, constitutes a crucial component of intelligent urban rail transit operations. Addressing the deficiency in existing methods that fail to fully utilize station-specific inflow and outflow passenger volume data, this study employs three machine learning approaches as predictive methodologies: linear regression model, random forest model, and support vector machine model. According to experimental results, the random forest model performs exceptionally well in terms of prediction accuracy, which makes it especially appropriate for application scenarios that call for a compromise between processing speed and precision. This model is therefore recommended as the preferred solution.