A vehicular network based intelligent transport system for smart cities using machine learning algorithms.

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• Traffic congestion is a major urban challenge, particularly in areas with limited infrastructure and connectivity.
• Existing traffic monitoring solutions often rely on extensive physical infrastructure and reliable internet, which are not feasible in developing regions.
• Internet traffic analysis offers potential solutions for real-world problems like urban mobility.

Purpose of the Study:

• To propose an intelligent transport system for predicting traffic congestion in smart cities using Internet-of-Vehicles (IOVs).
• To evaluate the effectiveness of various machine learning models, specifically tree-based algorithms, for IOV traffic analysis.
• To determine if feature selection enhances the performance of these machine learning models in predicting traffic congestion.

Main Methods:

• Utilized ensemble learning with tree-based machine learning strategies: decision trees, random forests, extra trees, and XGBoost.
• Implemented feature selection (FS) techniques to identify crucial features for traffic prediction.
• Employed a Stacking approach, averaging feature selection, to enhance detection accuracy and minimize computational cost.

Main Results:

• Tree-based machine learning approaches combined with feature selection demonstrated superior performance for IOV-based vehicular network traffic prediction.
• The proposed system achieved high detection accuracy with minimal computational overhead.
• The Stacking approach achieved the highest accuracy of 99.05%, outperforming KNN (96.6%) and SVM (98.01%).

Conclusions:

• Ensemble learning and feature selection are effective in developing intelligent transport systems for smart cities.
• The proposed system offers a practical solution for traffic congestion prediction in IOV networks, even with limited infrastructure.
• Machine learning, particularly tree-based methods, provides a robust framework for intelligent transportation solutions.