With increasing user densities in wireless networks, spectrum scarcity along with dynamic spectrum allocation are the major issues. To tackle with this problem, a novel algorithm called ANCR (Agglomerative NOMA Cognitive radio) is designed and developed in this research work. This method effectively tackles the problem of user pairing for clustering and resource allocation in a cognitive radio (CR) system by combining agglomerative clustering with NOMA (Non-Orthogonal Multiple Access). The multiuser system for implementation is created using a single base station and RSSI (Received Signal Strength Indicator) value. Using this technique, users are divided into more frequently utilized applications such as web browsing, streaming, emergency services, video calls and download backgrounds. The usage matrix is used to determine the priorities. The developed algorithm is compared with various ML (Machine Learning) models such as KNN, DT regression, Gradient Boosting and SVM based on evaluation parameters such as r2 score, MSE, RMSE values. ANCR has highest r2 value 0.9645 and lowest MSE 0.00831, RMSE 0.0074 values among all. Further, this novel technique is implemented over Kaggle dataset for 5G resource allocation for spectrum allocation to the different users based on their applications. Thus, the ANCR technique with best evaluation parameters and dynamic spectrum allocation can serve the 5G and beyond network for effective spectrum and dynamic resource management.