Leveraging Artificial Intelligence in Public Procurement of Dual-Use Goods

Authors:

Abstract
This paper examines the use of artificial intelligence (AI) in public procurement of dual-use goods, emphasizing the need to automate the identification of risky procurements and enhance state oversight efficiency. The paper solved two main tasks: development of an algorithm for analysing public procurement based on the Common Procurement Vocabulary (CPV) codes and lot descriptions, and training an AI model to classify risky procurements.The analysis is based on a sample of over 72,000 public procurements by the defence and security sector of Ukraine (2001-2025), as well as data on dual-use goods. The Term Frequency-Inverse Document Frequency (TF-IDF) method combined with cosine similarity was used to process textual data, identifying lots likely related to dual-use goods. An expanded procurement sample, including additional characteristics of suppliers and lot parameters, was used to train Naive Bayesian, Random Forest, and Extreme Gradient Boosting (XGBoost) classifiers. XGBoost demonstrated the best accuracy. In the paper, we also developed an algorithm for using blockchain to store procurement data.

This publication has 11 references indexed in Scilit: