Spreading Factor Classification In Lora Communication Based on Machine Learning

The proliferation of IoT has greatly impacted daily life by offering numerous solutions, with key requirements for devices including long battery life, extensive range, and low deployment costs. Low-power wide-area networks (LPWAN), particularly long-range (LoRa), are essential for applications requiring long-distance data transmission. However, one of the critical challenges in LoRa networks is the optimal selection of spreading factor (SF) values, which significantly influences network performance. Inappropriate SF selection can lead to increased interference, suboptimal network capacity, and reduced battery life for IoT devices. This study seeks to investigate the optimal selection of SF values based on distance by utilizing a machine learning model. The dataset employed comprised real-world data obtained from a rice field area in Banyumas, Central Java, Indonesia, which has a tropical climate. Accordingly, this study examined the performance of various classification algorithms, including k-NN, Random Forest, and Decision Tree, in the context of SF assignment in LoRa networks. The simulation results reveal that the proposed model achieves the highest accuracy of 96.88% for the random forest algorithm. These promising results, derived from lightweight learning methodologies, suggest a feasible approach for determining appropriate SF values corresponding to different distances.