Samiya Sharma ¹ Ridhima Sharma ¹ Atif Chowhan ^2* Mudita Uppal ¹ Deepali Gupta ¹ Shilpa Saini ¹ Yasir Hamid ²

• ¹ Chitkara Institute of Engineering & Technology, Chitkara University, Punjab, Punjab, India
• ² Abu Dhabi Polytechnic, Abu Dhabi, United Arab Emirates

With the ever-increasing and ever-changing population of the world, the need for efficient transportation and a well-organized traffic system has become increasingly critical. The convergence of Machine Learning (ML) and Internet of Things (IoT) technologies is propelling the global emergence of a new era of intelligent transportation. This paper meticulously investigates different models for traffic prediction by utilizing four strategies, i.e., Random Forest, Gradient Boosting, Decision Tree and Support Vector Machine. The proposed models' accuracy and efficacy are highlighted by rigorous evaluation with metrics namely Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and R-squared (R2) score. The study addresses key challenges such as ensuring computational efficiency, maintaining model interpretability, and avoiding overfitting. Among the evaluated models, Gradient Boosting proved to be the most effective ML method, achieving an accuracy of 85.6%, with the lowest MAE and MSE, and the highest RMSE and R2 score. Conversely, the Support Vector Machine model showed the weakest performance, with the highest errors and lowest R2 score among the models evaluated. This paper's key findings demonstrate that Gradient Boosting is the most effective model for traffic prediction, significantly improving the accuracy and reliability of smart transportation systems. These results show that advanced traffic management systems have the ability to significantly boost the effectiveness and dependability of smart transportation infrastructure, paving the way for future innovation and progress in this pivotal field. Future directions involve integration of real-time data, scalability, support for autonomous vehicles, multi-modal analysis, adaptive ML models and enhanced security.