MANETs are mobile nodes that self-organize into temporary networks without a base station. There is no dedicated path between source and destination nodes, these are dynamic and created only when required. Due to this, it also reacts to network changes. This versatility increases energy use, congestion, and routing overhead. In dense networks, node clustering—where cluster heads route instead of nodes is crucial. Neural network-based methods are used to create a robust MANET cluster creation and head selection mechanism. Inter-centroid distances are used to cluster nodes using the K-means method. Cluster heads with better energy and longer stability are chosen. Cluster head re-election and member node re-affiliation are reduced with this method. In each cluster, an artificial neural network (ANN) determines the head and adjusts weights based on energy, mobility, packet drop rates, and nearby nodes. Thereafter different performance parameters are then examined, experimental results showed that the proposed method reported better results under similar test conditions and outperformed the conventional state-of-the-art.