A Machine Learning Paradigms for Metro Interstate Traffic Volume Prediction

This study introduces a new combined method for forecasting traffic volume on metro interstate highways, integrating conventional ML and DL approaches. Using a comprehensive dataset that includes weather conditions, time-based features, and simulated traffic incidents, we develop and compare four models: a tuned RF, GB, Stacked LSTM network, and a hybrid of these models. Our research contributes to the field through extensive feature engineering, advanced hyperparameter tuning, and the development of a unique hybrid model that integrates the strengths of tree-based and neural network approaches. The RF model demonstrates superior performance compared to other models based on the Mean Squared Error (MSE) and R-squared (R2) metrics. However, the hybrid approach demonstrates promising potential, particularly in its ability to incorporate both tree-based and temporal modeling techniques. This study not only provides valuable insights into the factors influencing traffic volume but also offers a robust framework for future traffic prediction models. The findings could be used in urban planning and traffic control approaches, leading to more effective transportation systems.