<p>Vehicular ad hoc networks (VANETs) represent a transformative technology that enhances road safety, optimizes traffic flow, and enables efficient communication between vehicles and infrastructure. However, vehicular environments’ dynamic and unpredictable nature presents significant challenges, including frequent cluster head (CH) changes, increased communication overhead, and limited network lifetime. These issues degrade overall network performance and hinder the practical deployment of VANETs in real-world scenarios. This paper introduces IMICLiVAN, an enhanced clustering mechanism designed to address these challenges. The proposed method integrates the K-means algorithm with the silhouette score to dynamically determine the optimal number of clusters, ensuring compact and stable cluster structures. Additionally, a weighted formula for CH selection is employed, designed to balance multiple metrics to improve cluster stability and reduce unnecessary cluster head changes, and minimize reformation overhead. The performance of IMICLiVAN was evaluated through simulation-based experiments under varying traffic densities. Simulation results demonstrate that IMICLiVAN significantly outperforms existing clustering methods across key metrics, including WTCHS, ECBLTR, and EKSGA. These findings establish IMICLiVAN as a practical and effective solution for enhancing VANET performance in dynamic environments. From a supercomputing perspective, the per-step clustering quality evaluation (silhouette-based selection over multiple candidate-K values) and distance-driven computations become increasingly demanding in dense or city-scale VANETs; therefore, IMICLiVAN is designed to be amenable to parallel and distributed execution (e.g., RSU/MEC or cloud/HPC backends) to help meet strict latency constraints in real-time operation.</p>