A Deep Neural Network– XGBoost Approach for Efficient Resource Allocation in Distributed Edge Computing Networks

ABSTRACT

The proposed Deep Neural Network (DNN)‐XGBoost model integrated with the Pufferfish Optimization Algorithm (POA) significantly enhances the performance of edge computing across several critical metrics. The model achieves a latency of 22 ms, ensuring rapid response times for IoT‐driven applications. It maintains an impressive resource utilization rate of 95%, minimizing resource wastage while maximizing system throughput and reducing bottlenecks. Additionally, the framework lowers energy consumption to 75 J, demonstrating its effectiveness in supporting energy‐efficient operations for power‐sensitive edge environments. The execution time is reduced to 2.8 s, highlighting the model's capability to process computational tasks efficiently and meet the demands of real‐time applications. Furthermore, the proposed approach attains 98% prediction accuracy, enabling precise resource demand forecasting and facilitating dynamic resource allocation for improved scalability and system stability. The framework also achieves an overall system performance of 96%, reflecting its robust integration and reliable operation across diverse edge computing scenarios. Moreover, its adaptability rate of 98% demonstrates a strong ability to adjust to changing workloads and environmental conditions, ensuring consistent performance under varying operational requirements.