Integrating Econometrics and Artificial Intelligence to Assess the Impact of Trade on Nuclear Proliferation

Assessing nuclear proliferation risk from open-source international trade data with machine learning, deep learning, and econometric methods

Nuclear Technology 2025

Chansuh Lee1,2, KyungTae Lim3,4, Man-Sung Yim2,†

1Korea Institute of Nuclear Nonproliferation and Control (KINAC) 2Department of Nuclear and Quantum Engineering, KAIST 3Department of Applied Artificial Intelligence, Seoul National University of Science and Technology 4Graduate School of Culture Technology, KAIST

†Corresponding author

Paper (Nuclear Technology) PDF

Abstract

Given the rise in global interest in nuclear energy, the spread of nuclear technological capabilities and their potential impact on nuclear nonproliferation are of significant interest. This study examines the utility of open-source international trade data along with demand and supply-side data as a means by which to assess the potential nuclear proliferation risk related to nuclear power development. The proliferation risk assessment involves the use of machine learning, deep learning, traditional econometric methods, and big data. The results of the analysis indicated that using trade data can assist with nuclear proliferation risk predictions. Key items of importance in relation to nuclear trade were found to be the Harmonized Commodity Description and Coding System (HS) code 360300 (explosives for signaling, the most significant feature), followed by HS codes 282590 (inorganic bases) and 841350 (reciprocating positive displacement pumps for liquids). Other important items were HS codes 722810 (stainless steel products), 391721 (tubes, pipes, and hoses of plastic), 840120 (nuclear reactors and their parts), and 722830 (bars and rods of alloy steel).

Keywords Nuclear nonproliferation Proliferation risk assessment International trade data HS (Harmonized System) codes Econometrics Machine learning Deep learning Big data / open-source analysis

Main Findings

As stated in the published abstract:

  • Trade data helps predict proliferation risk. Combining open-source international trade data with demand- and supply-side data, and analyzing it with machine learning, deep learning, and traditional econometric methods, assists nuclear proliferation risk prediction.
  • Most significant feature: HS 360300 (explosives for signaling), followed by HS 282590 (inorganic bases) and HS 841350 (reciprocating positive displacement pumps for liquids).
  • Other important trade items: HS 722810 (stainless steel products), HS 391721 (tubes, pipes, and hoses of plastic), HS 840120 (nuclear reactors and their parts), and HS 722830 (bars and rods of alloy steel).

BibTeX

@article{Lee_2025,
  title={Integrating Econometrics and Artificial Intelligence to Assess the Impact of Trade on Nuclear Proliferation},
  volume={212},
  ISSN={1943-7471},
  url={http://dx.doi.org/10.1080/00295450.2025.2462494},
  DOI={10.1080/00295450.2025.2462494},
  number={1},
  journal={Nuclear Technology},
  publisher={Informa UK Limited},
  author={Lee, Chansuh and Lim, Kyungtae and Yim, Man-Sung},
  year={2025},
  month=May,
  pages={198–218}
}
More projects from KAIST MLP Lab →