Machine Learning-Based Logistic Regression with Hybrid Loss Function for Enhanced Spectrum Sensing

In the field of radio technology, the scarcity of spectrum resources has become increasingly severe, while demands for higher accuracy and speed in spectrum sensing technologies continue to grow. To address these challenges, this paper proposes a spectrum sensing algorithm based on machine learning logistic regression. By analyzing signal feature data, the algorithm effectively identifies available spectrum through logistic regression. Specifically, it first extracts representative spectral signal characteristics as feature values, employs spectrum availability status as labels for logistic regression modeling, and designs a hybrid loss function tailored to signal characteristics. Compared with traditional spectrum sensing algorithms, the improved logistic regression approach features a loss function better adapted to signal data, simplifies the algorithmic architecture, reduces computational complexity, and enhances prediction accuracy and speed.