Industrial pollution control based on artificial intelligence: A synergistic model using social network analysis and machine learning

Abstract

This research examines the impact of articulated robots (ARs), the Environmental Policy Stringency Index (EPSI), and foreign direct investment (FDI) on industrial air pollution—measured by PM2.5 levels—across 12 developed countries from 1993 to 2023. Employing Social Network Analysis (SNA) for variable selection, Granger causality tests for temporal validation, and machine learning (ML) for predictive modeling, this work captures the complex, nonlinear dynamics of pollution outcomes. The results yield three key insights. First, the EPSI consistently mitigates PM2.5 emissions, lending support to the Porter Hypothesis, which posits that stringent environmental regulations can drive innovation while reducing pollution. Second, FDI demonstrates a consistent negative effect on PM2.5 emissions, primarily by facilitating the transfer of cleaner technologies and advanced management practices. This pollution-reducing impact is particularly evident in contexts with robust regulatory frameworks, indicating that foreign investment can support environmental improvement when aligned with effective institutional oversight. Third, AR (articulated robots) consistently exhibits a negative impact on PM2.5 emissions by enhancing operational precision, improving energy efficiency, and minimizing resource waste in industrial processes. The integration of robotic automation contributes to cleaner production practices, particularly when supported by clean energy adoption and effective environmental regulations. To enhance the accuracy of PM2.5 predictions, six ML models were tested: ARIMA, SARIMA, LightGBM, XGBoost, LSTM, and GRU. Among these, the integration of SNA with LSTM achieved the highest predictive accuracy, outperforming traditional models in capturing the complex, long-term dynamics of pollution. This synergistic approach not only underscores the pivotal roles of the EPSI, FDI, and AR in pollution mitigation and but also offers a practical framework for incorporating advanced technologies into industrial pollution control approaches.