The incorporation of intelligence into wireless sensors is a contemporary paradigm for boosting the effectiveness of wireless sensor networks (WSNs) and the effective use of the radio frequency spectrum. The existing traditional distribution of radio frequencies leads to coexistence issues or resource underutilization. Cognitive sensors with adjustable spectrum sharing might be used to solve this problem. In this field, spectrum management includes spectrum sensing and allocation. Several techniques have been developed to address these issues but complexity and throughput remain a challenging tasks for these networks. In this work, these issues have been addressed and developed a combined approach for spectrum sensing, spectrum allocation, and cluster head selection to improve the overall spectrum utilization and improve the energy utilization. Moreover, the proposed scheme also adapts the deep learning scheme to identify the spectrum of information based on historical patterns. The comparative analysis shows that the proposed approach shows better spectrum detection performance in comparison with KMeans, GMM, KNN, DAG-SVM approaches.