This article examines the problem of predicting tram passenger load using machine learning methods. The relevance of the study stems from the need to improve passenger information and optimize public transport use. The paper proposes an integrated approach to data collection and preparation, using vehicle tracks, meteorological data, commuter train schedules, information on public transportation hotspots, and the results of field passenger counts using a tabular method. Feature processing included time stamp decomposition, stop geocoding, and the calculation of derived parameters. During the exploratory analysis stage, key passenger flow patterns were identified: daily and weekly dynamics, the influence of commuter train arrivals, and the correlation of load with day type and stop location. To solve the forecasting problem, three models were trained and compared: linear regression, random forest, and gradient boosting. The random forest model demonstrated the best results and was selected for industrial implementation. A software implementation was developed in Python and deployed on an Apache server using MySQL for data storage. The model is integrated into a user-friendly web service: the congestion forecast is visualized as an intuitive scale, making it easier for passengers to understand the information.

machine learning, passenger flow forecasting, public transport, tram, random forest, data analysis, smart city

Artificial intelligencе and big data