Proactive spectrum access using LSTM prediction in cognitive radio-enabled IoT systems

Traditional spectrum sensing approaches of Cognitive Radio systems, being reactive in nature, often lead to excessive energy consumption, increased latency, and suboptimal throughput. This work presents a prediction-aided spectrum management framework that proactively forecasts spectrum occupancy, improving energy efficiency, throughput and responsiveness in IoT networks with cognitive radio capabilities. We propose a deep learning model based on a Long Short-Term Memory neural network to forecast the availability of spectrum from time-series data. The proposed model is simulated using MATLAB-Simulink. The performance of the system is evaluated based on the prominent sensing parameters, probability of detection and probability of false alarm, across a range of signal to noise ratios (SNRs). Simulation results establish that the proposed ML based model reliably detects spectrum holes with high detection rate at favorable SNR. Therefore, this framework paves the way for intelligent, adaptive, and energy-efficient spectrum management tools for large-scale IoT applications.