Sustainable decision-making for contaminated site risk management: A decision tree model using machine learning algorithms

The presence of contaminated land is an inevitable legacy of industrial activity, and the management decisions governing reclamation of this land are key in minimizing environmental risk and allowing safe and effective land reuse. In this context, to predict the optimal remediation options for future decision-making processes in sus-tainable site management, thus enhancing information communication between stakeholders, 17 decision sensitivity parameters are analyzed in this study and their influence on the management patterns of contami-nated sites identified with three decision tree (DT) algorithms including C4.5 (successor of Iterative Dichotomiser 3/ID 3), CHAID (Chi-squared Automatic Interaction Detection), and CART (Classification and Regression Trees), which is the first attempt to use artificial intelligence technology to predict strategy-based decision-making for contaminated site management. Based on four performance metrics (accuracy, precision, recall ratio and F1 score), CART-based DT model shows the highest prediction accuracy at an average value of 78.57%, which indicates a relatively credible decision simulation to assist in more efficient contaminated site management. With regard to specific factors and influence mechanisms on contaminated site management, the results demonstrate 7 recognition rules corresponding to 6 driving factors which have the greatest influence on the decision-making process. Long-term monitoring time, the type of land reuse and ex-situ performance are the most important factors in determining field implementation of cleanup activities. The built decision tree model and induced decision rules, once well-trained, can be relied on for a sustainable site management strategy as data become available at a new site.

Automated summary

This study applies artificial intelligence technology to predict the optimal remediation options for contaminated site management. By analyzing decision sensitivity parameters and using decision tree algorithms, the study identifies the factors and mechanisms that have the greatest influence on the decision-making process.