Abstract: As industrial automation accelerates across diverse sectors, its socioeconomic repercussions remain complex and uncertain, particularly on employment, income distribution, and macroeconomic stability. This study proposes a data-driven framework that integrates labour-market microdata, industry performance metrics, and task-level automation-risk indices to forecast changes in key socioeconomic indicators. Five state-of-the-art machine learning algorithms—Random Forest, XGBoost, CatBoost, LightGBM, and TabNet—were trained and compared. These models capture complex, nonlinear interactions in socioeconomic data. Among them, Random Forest demonstrated the best predictive performance, achieving the lowest RMSE (1560.74) and highest R² (0.9690), significantly outperforming other models. Interpretable ML techniques, such as SHAP values and counterfactual simulations, are employed to identify the most influential predictors of automation-related socioeconomic change. The results offer a fine-grained, scenario-based understanding of future labour market trends, reinforcing the value of machine learning—especially Random Forest—as a robust forecasting tool for evidence-based policy in an era of rapid technological transformation.

Keywords: Industrial Automation, Socioeconomic Impact, Forecasting, Random Forest, XGBoost, CatBoost, LightGBM, TabNet, Data-Driven Policy

Pages: 59-74

WSEAS Transactions on