Explainable Machine Learning Model for Performance Prediction MAC Layer in WSNs

Wireless Sensor Networks (WSNs) are used to gather data in a variety of sectors, including smart factories, smart buildings, and so on, to monitor surroundings. Different medium access control (MAC) protocols are accessible to sensor nodes for wireless communications in such contexts, and they are critical to improving network performance. The proposed MAC layer protocols for WSNs are all geared on achieving high packet reception rates. The MAC protocol is adopted and utilized throughout the lifespan of the network, even if its performance degrades over time. Based on the packet reception rate, we use supervised machine learning approaches to forecast the performance of the CSMA/CA MAC protocol in this study. Our method consists of three steps: data gathering trials, offline modeling, and performance assessment. According to our findings, the XGBoost (eXtreme Gradient Boosting) prediction model is the most effective supervised machine learning approach for improving network performance at the MAC layer. In addition, we explain predictions using the SHAP (SHapley Additive exPlanations) approach.