Towards Efficient Mobile Edge Computing: A Comparative Assessment of Offloading Techniques

Task offloading in MEC is the transfer of computation intensive tasks off to the server or cloud to accelerate the high demanding applications to be efficiently handled by the network. This process can vividly optimize the resource utilization and could achieve energy efficiency especially in the MEC paradigm. Intelligent task offloading schemes are the need of the hour, due to their adaptive and evolving behavior. Lot of AI based models in the Machine Learning and Deep Learning domains have been developed as a strategic decision-making model in task offloading. This work performs a comparative assessment of the efficacy of different models like Random Forest, Extremely Randomized Trees, Extreme Gradient Boosting, Artificial Neural Network, Deep Neural Network (DNN) and Convolutional Neural network in intelligent task offloading. The experimental evidences were evaluated on standard classification benchmark metrics. The results indicated that DNN architecture showed better performance than the other compared models because of its ability to learn complex relations among the data through its dense connections. This work can lead the way to many futuristic intelligent task offloading strategies that could handle large massive amounts of network data.