Performance analysis of topologies and ratios with machine learning using Cooja simulator

The prediction of Quality of Service (QoS) in wireless sensor networks (WSNs) plays a crucial role in enhancing the performance of resource-constrained IoT systems. This research focuses on utilizing machine learning (ML) techniques to predict QoS levels, categorized into high, medium, or low, based on key network performance metrics such as throughput, latency, and jitter. The study uses the Cooja simulator and different 6LoWPAN topologies, such as Random, Linear, and Ellipse. It also looks at network configurations with different types of mote (Sky, Z1, and Wismote) and transmission rates (100% and 50%). Data collected from \s in the form of packet capture (PCAP) files is processed to compute the QoS metrics, which are then used as inputs for random forest-based regression and classification models. This study uses a mixed approach that combines simulation results with machine learning models to measure performance, with the goal of improving efficiency and enabling faster, data-based decisions in real-life situations. The results show that the ML models successfully predicted QoS, achieving an accuracy of 100% in classifying QoS levels and a 98.5% accuracy in predicting continuous QoS scores. The analysis demonstrates that network topology and mote configuration significantly impact QoS performance, with optimal configurations identified for improved reliability and efficiency. These findings provide a foundation for integrating predictive modeling into real-world IoT networks, enabling intelligent decision-making in applications like smart cities, healthcare, and industrial automation.