Author(s): Mohammad Shihab Ahmed

Abstract: City traffic congestion continues to be a significant issue, resulting in financial losses, higher emissions, and diminished quality of life. Conventional traffic management systems, typically reactive and rule-oriented, find it challenging to adjust to changing conditions. This study presents a smart, data-centric method utilizing machine learning (XGBoost, Random Forest, LSTM) and diverse data sources (GPS, IoT sensors, weather) to forecast congestion in real time. Our model reaches 71.2% accuracy, highlighting temporal features (hour, day of week) and weather conditions (rain, snow) as significant predictors identified via feature importance analysis. Comparative analysis indicates that XGBoost surpasses other algorithms, achieving a balance of accuracy (71.2%), computational efficiency (11ms latency), and interpretability, which is essential for practical application. The research emphasizes practical uses, such as real-time traffic signal enhancement and preventive traffic jam reduction, while tackling issues like class imbalance and immediate data handling. This work enhances Intelligent Transportation Systems (ITS) by integrating predictive analytics with operational traffic management and creates a scalable framework for smart cities. Future pathways involve combining social media data with edge computing for city-wide applications. This study enhances sustainable urban mobility, providing officials with an economical, data-supported approach to lessen congestion and improve the commuting experience.

DOI: 10.33545/2707661X.2025.v6.i2a.134

Pages: 01-08 | Views: 130 | Downloads: 48

Download Full Article: Click Here