This study investigated the use of machine learning techniques for the classification of the propagation environment in vehicle-to-vehicle (V2V) communication systems under varying environmental conditions, to improve communication reliability and energy efficiency. To do this, field measurements were first conducted in Trabzon and Gümüşhane, Turkey, for different environments, including rural, highway, suburban, and urban environments, to analyze the propagation characteristics. Then, the classification performance of five machine learning models—K-Nearest Neighbors (KNN), Random Forest (RF), Artificial Neural Networks (ANN), Support Vector Machines (SVM), and AdaBoost—was evaluated and compared. Key input features, including distance, path loss, obstacle types, modulation schemes, and weather conditions, were weighted and utilized for classification. Here, the KNN model achieved superior performance with an accuracy of 99.84%, precision of 99.83%, recall of 99.84%, and an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 99.97%. RF and ANN models also demonstrated high AUC-ROC values, underscoring their robustness in complex propagation environments. To understand the contribution of individual features to the classification performance, an Out-of-Bag Permutation Feature Importance analysis was performed. The results indicated that weighted building obstacles were the most influential feature, accounting for 20.09% of the total feature importance. These findings highlight the potential of adaptive power adjustments based on real-time environmental classification to improve both the reliability and efficiency of V2V communication systems, thereby supporting the development of safer and more efficient intelligent transportation networks.
Sağır et al. (Fri,) studied this question.