The Internet of Things (IoT) has experienced rapid growth and adoption in recent years, enabling applications across diverse industries, including agriculture, logistics, smart cities, and healthcare. Long Range Wide Area Network (LoRaWAN) has emerged as a leading choice among IoT communication technologies due to its long-range, low-power, and cost-effective capabilities. However, the rapid proliferation of IoT devices has intensified the challenge of efficient resource management, particularly in spreading factor (SF) allocation for LoRaWAN networks. In this paper, we propose a Machine Learning-based Adaptive Data Rate (ML-ADR) approach for SF management to address this issue. A Long Short-Term Memory (LSTM) network was trained on a dataset generated using ns-3 for optimal SF classification. The pre-trained LSTM model was then utilized on the end-device side for efficient SF allocation with newly generated data during simulation. The results demonstrate improved packet delivery ratios and reduced energy consumption.
Nisar et al. (Wed,) studied this question.