Improved Deep Learning‐Based Clustering and Adaptive Social‐Aware <scp>CH</scp> Selection in <scp>VANET</scp>

ABSTRACT Vehicular Ad Hoc Network (VANET) is an emerging and promising wireless technology that allows vehicle‐to‐vehicle (V2V) communications between vehicles and with roadside infrastructure to improve traffic control and safety. In VANETs, stable clustering of vehicles is significant for efficient network communication. Conventional clustering techniques, such as k‐means and spectral clustering, consider spherical clusters and cannot deal with vehicles' high mobility and frequent topology changes, resulting in unstable clusters. This study presents an advanced clustering scheme based on improved Deep Autoencoder and Nonnegative Matrix Factorization (iDANMF) that aims to increase the efficiency of VANET clustering by including an adaptive social‐aware cluster head (CH) selection. The iDANMF model utilizes raw vehicular data to map to a lower‐dimensional space with the help of a modularized vehicle adjacency matrix, generating reliable embeddings for efficient clustering. To ascertain the CH, a relational mapping is calculated utilizing adaptive social‐aware parameters. This methodology enhances the precision of the selection process by integrating social network dynamics into the vehicular computational framework. Experiments are conducted over real‐time traffic data of the Connaught area of Delhi generated through SUMO simulation. Experiments demonstrate that the proposed model achieves approximately 15% higher CH stability and higher CH lifetimes than state‐of‐the‐art models, highlighting its effectiveness in VANET clustering and CH selection.