Short-term passenger flow prediction of subway based on XGBoost multi-feature fusion

The aim of this study is to use the XGBoost algorithm for short-term metro passenger flow prediction and to improve the prediction accuracy by combining several features. The study collected historical twenty-five days of swipe data from Hangzhou metro stations and considered other factors that may affect passenger flow, such as time of day, weather and holidays. After data pre-processing and feature engineering, suitable features were screened out and the prediction of passenger flow at different times of the day in the future was finally achieved. In the model training stage, the XGBoost algorithm is used to construct the passenger flow prediction model, and the model performance is optimized by adjusting hyperparameters and cross-validation. To further improve the prediction accuracy, a multi-feature fusion approach, including feature combination, feature crossover and feature weighting techniques, is used to capture the correlation and degree of influence between features. By evaluating the training and test sets, the study assesses the prediction performance of the model and uses the mean absolute error (MAE) metric to measure the error between the prediction results and the actual passenger flow. In addition, a comparative analysis with the random forest algorithm is performed to evaluate the performance of the XGBoost algorithm. The experimental results show that the XGBoost multi-feature fusion-based metro short-term passenger flow prediction model performs well in terms of accuracy and stability. The fusion of multiple features can better capture the influencing factors of passenger flow changes and improve the prediction accuracy. The study also analyzes the importance of the features, which helps to understand the explanatory power of the model. In summary, the XGBoost multi-feature fusion-based metro short-term passenger flow forecasting method proposed in this study has the potential to provide accurate passenger flow forecasts in practical applications to support metro operation management and passenger travel decisions.