5GT-GAN: Enhancing Data Augmentation for 5G-Enabled Mobile Edge Computing in Smart Cities

This paper introduces 5GT-GAN, a novel approach leveraging generative adversarial networks (GANs) to create synthetic mobile Internet traffic data, particularly tailored to smart city applications. Given the challenges of data scarcity and privacy concerns in the context of 5G, generating synthetic data becomes a crucial aspect for effectively deploying AI-driven systems in real-world scenarios. 5GT-GAN integrates unsupervised GAN schemes with the ability to manage temporal dynamics through supervised autoregressive models, successfully generating large-scale synthetic mobile Internet traffic data. Our experimental results illustrate the superior performance of 5GT-GAN in terms of mean squared error (MSE) and mean absolute error (MAE) compared to traditional models. The use of "Train Synthetic Test Real" (TSTR) and "Train Real Test Synthetic" (TRTS) methodologies affirmed the model's effectiveness with (0.0023 MAE, 0.0074 MSE) and (0.0045 MAE, 0.0092 MSE) respectively. Moreover, the model's runtime complexity of O(n log n) emphasized its efficiency in handling larger datasets, an edge over traditional models. The study also identifies potential future work in augmenting data for traffic prediction and integrating self-attention mechanisms to enhance the capabilities of the model further.