@article{CRSIJ26000128,
                  author = {Mohammed Adamu Sule, Benjamin Abba-Stephen, Maryam Abdulkadir, Denise D. Jonathan, Mujittapha Idris, Shehu Abdullahi},
                  title = {Deep Learning for Spectrum Allocation in 5G Networks: A Python-Based Experimental Study Considering Regulatory Policies},
                  journal = {Cosmo Research and Science International Journal},
                  year = {2025},
                  volume = {1},
                  number = {6},
                  pages = {46-58},
                  issn = {3108-1584},
                  url = {https://cosmorsij.com/published/CRSIJ26000128.pdf},
                  abstract = {The efficient allocation of radio spectrum is a critical challenge in fifth-generation (5G) wireless networks, where dynamic and heterogeneous traffic demands require intelligent resource management solutions that simultaneously satisfy technical performance objectives and regulatory compliance constraints. This paper presents an experimental study of deep learning-based spectrum allocation techniques applied to 5G network environments, implemented and evaluated entirely in Python using Tensor Flow 2.x and the Open AI Gym framework. Three deep reinforcement learning (DRL) architectures: Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and a novel hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) models are trained and evaluated on a custom-built 5G network simulation environment incorporating regulatory constraints derived from ETSI EN 301 893, ITU-R M.2150 (IMT-2020), and FCC Part 27. The proposed CNN-LSTM model achieves a mean throughput of 487 Mbps, representing a 56.1% improvement over baseline DQN, a 21.4% improvement over PPO, and a latency reduction to 9.7 Ms under full regulatory compliance conditions. Critically, the study quantifies the performance-compliance trade-off for the first time in a 5G DRL context, demonstrating that full regulatory compliance introduces an 18.3% throughput penalty compared to unconstrained operation, a finding with significant implications for spectrum policy design. The experimental framework, training datasets, and regulatory constraint modules are made publicly available to support reproducibility.},
                  keywords = {5G networks; spectrum allocation; deep reinforcement learning; DQN; CNN-LSTM; regulatory compliance; ETSI; ITU-R; Python; Tensor Flow},
                  month = {May}
        }