Data-Driven Solutions for Shuttle Bus Travel Time Prediction: Machine Learning Model Evaluation at Nnamdi Azikiwe University

Efficient campus transportation is vital for academic institutions, yet shuttle services often suffer unpredictable delays due to dynamic traffic, weather, and passenger conditions. This study investigates the application of machine learning techniques to predict shuttle bus travel time within Nnamdi Azikiwe University, Awka, Nigeria. Using a dataset collected over a one year period, including route distance, traffic level, weather condition, and passenger count. Four supervised regression models were evaluated: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Artificial Neural Network (ANN), and Extreme Gradient Boosting (XGBoost). Following comprehensive data preprocessing, model performance was assessed using RMSE, MAE, and R² metrics. Results indicate that XGBoost achieved superior predictive accuracy (R² = 0.984, RMSE = 1.16), outperforming other models in capturing the nonlinear relationships among variables. ANN followed closely, while LSTM and GRU exhibited weaker performance, likely due to limited temporal granularity in the data. Feature importance analysis revealed route distance and traffic level as the most influential predictors. These findings demonstrate that