Enhancing Spectrum Prediction via Efficient Channel Selection and Adaptive Correlation Extraction

Efficient spectrum utilization is crucial for meeting the demands of modern wireless communications. Spectrum prediction aids this by forecasting future availability. Existing methods face challenges in large-scale environments, particularly in accurately handling inter-channel correlations, which hinders prediction performance. This paper introduces an enhanced spectrum prediction method leveraging efficient channel selection and adaptive correlation extraction. The method first obtains initial predictions using a channel-independent approach. Then, a proposed Copula-Max Dependency (CMD) method selects relevant channel subsets. An adaptive module extracts correlations from these subsets to refine the final predictions. Experiments show our method achieves significantly improved prediction accuracy and effectively resolves issues related to inaccurate inter-channel correlation modeling compared to state-of-the-art baselines.