A Classifier-Driven Deep Learning Clustering Approach to Enhance Data Collection in MANETs

Abstract

The conventional clustering and routing approaches used in mobile ad hoc networks (MANETs) may fail to work effectively in a dynamic network environment where nodes are highly mobile and the traffic load may also vary significantly. These limitations result in negative effects such as high packet drop rates, delays in data transmission, and low delivery rates, which make these methods unfit for modern high-density networks. To overcome these issues, this paper proposes a new deep learning-based classifier for adaptive clustering in MANETs. Through the use of machine learning algorithms, the proposed method is able to adapt to node clustering through node behavior, mobility, and content distribution in real-time, thus improving network performance. This work compares the performance of the network on networks that contain 50, 100, and 200 nodes via a clustering algorithm. The performance parameters considered include the delivery ratio, packet drop ratio, and end-to-end delay. The evaluation findings show that the developed deep learning-based classifier is far more effective than the non-clustered and conventional clustering approaches are. In particular, the classifier approach provides a delivery rate of up to 89.4%, which is significantly better than that of the baseline scenarios and decreases packet drop rates by more than 70%, especially in high-density node scenarios. In addition, the proposed approach reduces the network delay and effectively handles the inherent dynamic characteristics of MANETs.