Integrating DEA and Machine Learning for Sustainable Resource Management in Arid Regions: A Comprehensive Framework for Groundwater Quality Assessment

This study introduces an innovative methodology integrating Data Envelopment Analysis (DEA), Principal Component Analysis (PCA), and Machine Learning (ML) to evaluate groundwater quality objectively and accurately. Mitigating the subjectivity of conventional Water Quality Indices (WQIs), this method incorporates statistically rigorous techniques to ensure reproducibility and facilitate cross-contextual applicability. The PCA was employed to reduce dimensionality, grouping 14 physicochemical parameters into meaningful components while retaining essential variability. These components informed a DEA framework, which generated Proxy Water Quality Indices (PWQIs) to rank 64 wells based on efficiency in meeting acceptable water quality standards. The analysis revealed significant spatial disparities in water quality across Algeria's Hodna Basin, linking poor quality to agricultural runoff and industrial pollution, while highlighting the mitigating role of hydrological features like the Soubella dam. Comparison with expert-based aggregation and conventional methods confirmed the robustness and enhanced discriminatory power of the PCA-DEA approach. This methodology provides a scalable, data-driven tool for water quality assessment, offering actionable insights for resource managers and policymakers.