DTON: A Study on Dynamic Task Offloading with NGCF in an Edge Computing Environment

The rapid proliferation of IoT devices and the expansion of 5G networks have increased interest in edge computing. While edge computing offers reduced latency and optimized bandwidth usage by processing data near its source, the limited computing capabilities of IoT devices necessitate efficient task offloading. This paper suggests a novel method to the task offloading problem in wireless IoT networks by drawing parallels with user-item matching in recommendation systems. We introduce DTON (Dynamic Task Offloading with NGCF) which is based on both dynamic task offloading and an enhanced model based on NGCF(Neural Graph Collaborative Filtering). DTON integrates a temporal graph structure to reflect dynamic network state changes and an attention mechanism to learn inter-node importance, enabling effective task offloading decisions in dynamic IoT network environments. Our approach model leverages the structural similarity between task offload and recommendation systems, and the technique proposed in this study improves the performance by more than 20% in task completion and latency over traditional models.