Machine Learning Based Public Transit Bus Arrival Time Prediction System

Machine learning has now become essential asset in everyday life, enabling more efficient data-driven decision-making. In transportation, it optimizes traffic management, predicts vehicle maintenance, and enhances public transit systems. By analysing large volumes of real-time data, machine learning models can accurately forecast bus arrival times, improve route planning, and reduce delays, ultimately creating an efficient and reliable commuting experience. This paper investigates the use of machine learning models to predict bus arrival time at bus stop, aiming to enhance the accuracy of public transportation systems. Seven models-XGBoost, Gradient Boosting Regressor, Light Gradient-Boosting Machine, CatBoost, Linear Regression, Markov Random Field, and Support Vector Machine-were evaluated using Root Mean Square Error (RMSE) as the error metric. Among these, XGBoost emerged as the most reliable, achieving a balance between training and test accuracy with minimal overfitting when compared to other models. While traditional models like Markov Random Field delivered weaker results, XGBoost consistently outperformed others. The findings highlight the potential of ensemble learning methods, particularly XGBoost, to optimize real-time prediction of bus arrival time, which can improve efficiency and reliability in urban transportation systems.