Spectrum sensing is a crucial process in the dynamic management of radio frequencies, particularly in cognitive radio networks. A major challenge lies in developing efficient techniques for simultaneously monitoring multiple frequency bands to detect and confirm the presence of licensed signals. The goal is to optimize spectrum utilization while ensuring fair and efficient radio resource management. This research examines the effectiveness of four supervised learning models in spectrum detection: Support Vector Machine (SVM), Naïve Bayes Classifier (NBC), Decision Trees (TREE), and K-Nearest Neighbors (KNN). The signal used is Binary Phase Shift Keying (BPSK), a widely employed digital modulation scheme. The models are tested on a synthetic dataset containing BPSK signals with varying noise levels (SNR ranging from \(-20\) dB to \(20\) dB). Performance is analyzed using probability of detection ( \(P_d\) ), probability of false alarm ( \(P_{fa}\) ), and total error rate ( \(TER\) ). The results indicate that SVM and NBC outperform the other models, offering the best trade-off between high detection accuracy, low false alarm rate, and minimal total error rate, making them promising approaches for intelligent spectrum sensing.