Feasibility Analysis of Enhanced Driving Behavior Recognition via IoV-based Deep Learning Framework

With the rapid development of the Internet of Vehicles (IoV), the integration of vehicular networks and deep learning has become a promising approach for enhancing driving behavior recognition. This paper presents a preliminary analysis of an IoV-based deep learning framework designed to improve the accuracy and efficiency of driving behavior recognition. We leverage real-world driving data provided by a globally renowned enterprise, fully utilizing the advantages of IoV in real-world environments, non-intrusive data collection, natural driving behavior patterns, and real-time processing. Experimental results demonstrate significant improvements in recognition accuracy, achieving a macro-average F1-score of $\mathbf{9 1. 2 \%}$, which surpasses simulation-based methods by $\mathbf{1 4. 6 \%}$. This work establishes a new benchmark for naturalistic driving behavior analysis using deep learning.