AI for Economic Diplomacy: Predicting the Global Impact of Sanctions Using Machine Learning

In this paper, we introduce a machine learning model for predicting the effectiveness of international economic sanctions based on structured data. From the Global Sanctions Data Base (GSDB-R4) and macroeconomic data from the World Bank, we construct models of sanction effectiveness that predict sanctions as effective or ineffective. Preprocessing data, feature engineering, and training machine learning models, such as logistic regression, random forest, gradient boosting models (XGBoost/LightGBM), and regularized regression methods, constitute the methodology. Precision, recall, F1-score, AUC-ROC, and confusion matrices serve as the evaluation measures. Results show that boosted and tree-based models outperform baseline models across the board in predicting sanction success. Feature importance analysis also shows significant economic and political drivers of sanction effectiveness and is an informative source of input for policymakers and diplomats. The contribution of this paper shows that artificial intelligence has promise as a decision tool in economic diplomacy and adds to an easy-to-reproduce model of future sanction strategy. Reproducibility is also ensured with the pipeline through the export of evaluation visualization plots in a variety of formats for policymaking and scholarly communication purposes. Key Words: economic sanctions, machine learning, international relations, GSDB, World Bank, economic diplomacy