Residual Energy and Quality of Service Parameters based Optimization of Congestion-Aware Machine Learning Algorithms

This paper presents a pioneering approach employing machine learning techniques to optimize routing algorithms in wireless networks, focusing on dynamic route adaptation while considering residual energy and quality of service (QoS) parameters. The proposed algorithm, Congestion-Aware Routing Optimization (CARO), utilizes a supervised learning model integrated with a hybrid decision-making framework to predict residual energy and prioritize routes accordingly. CARO employs a multi-layer perceptron (MLP) for energy prediction and a random forest model for QoS parameter optimization, ensuring robust decision-making under varying network conditions. Through extensive experimentation, the algorithm achieved a high accuracy of 90% for residual energy prediction, with a mean squared error (MSE) of 0.0752 and an R-squared value of -0.0084. For QoS parameter prediction, CARO demonstrated an MSE of 0.0852 and an R-squared value of 0.0024. These findings underscore the effectiveness of CARO in enhancing network performance by intelligently managing residual energy levels and maintaining QoS standards, offering significant advancements in congestion-aware routing optimization.