Improving detection of pollution fee declarations for environmental policy compliance through metaheuristic-optimized ensemble learning

Purpose – Given that the governments mandate industries to declare and pay fees for soil and groundwater contamination, relying on self-reporting creates risks of underreporting through fraudulent documentation. This study aims to address fraudulent pollution fee declarations by developing an advanced artificial intelligence (AI) detection model that enhances compliance with environmental policies. Design/methodology/approach – This study integrates the Synthetic Minority Oversampling Technique (SMOTE) and a forensic-based investigation (FBI) metaheuristic algorithm with ensemble machine learning to detect fraudulent declarations effectively. The model is optimized for class imbalance, ensuring strong performance across key metrics, including accuracy, precision, specificity, F1 score and area under the curve (AUC). Findings – The proposed model improves the detection of fraudulent pollution fee declarations and enhances the identification of tax evasion cases. Results indicate that combining data class imbalance techniques with model hyperparameter optimization significantly enhances the model's ability to distinguish between fraudulent and legitimate reports. Practical implications – This study enhances fraud detection in pollution fee declarations, ensuring that financial resources are allocated appropriately to remediation efforts. Reducing tax evasion and improving regulatory oversight support environmental sustainability, strengthen public health protections and promote fairer compliance practices, ultimately leading to more effective environmental policies and enforcement. Originality/value – This research presents a novel approach to environmental compliance monitoring using SMOTE-based ensemble learning optimized by the FBI algorithm, offering a scalable and adaptable solution for global regulatory frameworks. This methodological advancement enhances data-driven decision-making, improves fraud detection accuracy and streamlines compliance inspections, significantly outperforming traditional monitoring techniques.