BROWSE
Detail
Journal Article
ML-Based 5G Traffic Generation for Practical Simulations Using Open Datasets
Cited 8 time in 
Share 
Share
- Authors
- Yong-Hoon Choi, Daegyeom Kim, Myeongjin Ko, Kyung-yul Cheon, Seungkeun Park, Yunbae Kim, Hyungoo Yoon
- Issue Date
- 2023-09
- Citation
- IEEE Communications Magazine, v.61, no.9, pp.130-136
- ISSN
- 0163-6804
- Publisher
- Institute of Electrical and Electronics Engineers
- Language
- English
- Type
- Journal Article
- Abstract
- As artificial intelligence (AI) and machine learning (ML) technologies continue to advance, they are creating innovations in various industrial sectors and are increasingly influencing the development of wireless communication networks. Research advances in ML for wireless communications and networks depend essentially on the availability of datasets for testing results and attempting generalizations, but the availability of datasets obtained from real-world networks or experimental testbeds is currently very limited. To address this issue, we created a 5G traffic dataset by measuring various packet traffic and made it publicly available for anyone to use. The dataset is 328 hours long, making it difficult to handle directly and requiring preprocessing to be used for a specific purpose. To overcome these limitations, this study implements two types of traffic generators that can generate 5G traffic by improving ML models that have been verified and are widely used in academia. The traffic generator trained on the collected dataset has millions of parameters, but the total size of the model is no more than 118.12 MB, making it lightweight and easily deployable. To evaluate the difference between the generated traffic and the real traffic, the Jensen-Shannon divergence (JSD) and maximum mean discrepancy (MMD) metrics are observed, and it shows that the generated traffic is very similar to the real traffic. Our implementation and pretrained models are available on IEEE Code Ocean and GitHub.
- KSP Keywords
- Industrial sector, Jensen-Shannon divergence, Machine learning (ml), Maximum mean discrepancy, Real-world networks, Traffic Generation, Wireless communication Network, artificial intelligence, experimental testbed, open datasets, traffic generator