AI-DRIVEN SPECTRUM MANAGEMENT IN 5G NETWORK TECHNOLOGY TO IMPROVE NETWORK EFFICIENCY

Spectrum management has become a burning problem in the Fifth-Generation (5G) wireless networks as a result of growing traffic load, scarcity of spectrum, and very dynamic network conditions. The conventional approach to the allocation of the spectrum, static, cannot adapt to these obstacles and leads to poor utilisation of the available resources and worsens Quality of Service (QoS). This paper will suggest an AI-based spectrum management system that combines Extreme Gradient Boosting (XGBoost) to predict network traffic and spectrum demand with Deep Q-Network (DQN) to predict dynamic spectrum allocation. A practical network traffic dataset was deployed to study and test the proposed system in a virtual 5G environment. The experimental findings indicate that the XGBoost model has a high prediction accuracy as evidenced by a high R2 score of 0.93 that will be used to conduct proactive spectrum planning. The DQN-based allocation strategy demonstrated effective learning, achieving a spectrum utilization of 87.6%, compared to 62.3% obtained using static allocation. There were also noticeable enhancements in the average throughput, latency, packet loss as well as overall QoS. The results affirm that AI-based spectrum management is an efficient and dynamically adaptable tool that can be used to optimise spectrum utilisation and improve the network in 5G networks.