Network science and explainable AI-based life cycle management of sustainability models

Area of Science:

• Environmental science and policy
• Data science and machine learning
• Sustainability studies

Background:

• Model-based assessments are crucial for understanding variable impacts on Sustainable Development Goals (SDGs).
• Machine learning (ML) offers data-driven solutions for sustainability planning and model development.
• Continuous model review and development are essential for effective decision support in dynamic environments.

Purpose of the Study:

• To propose and demonstrate a novel approach for identifying policy intervention points for SDGs using the Machine Learning Operations (MLOps) framework.
• To leverage Shapley values and network analysis for understanding variable contributions and identifying key drivers.
• To validate the proposed methods through a case study on the Hungarian water model.

Main Methods:

• Utilized the Machine Learning Operations (MLOps) life cycle framework for model development.
• Applied Shapley value to quantify individual direct and indirect contributions of variables.
• Employed network analysis to identify key drivers and potential policy intervention points.
• Conducted a case study using the Hungarian water model, focusing on SDG 6.4.1 and 6.4.2 indicators.

Main Results:

• Identified water reuse and water circularity as more effective intervention options compared to pricing or utilizing renewable water resources.
• Quantified the contributions of various variables to water efficiency and water stress indicators.
• Validated the utility of Shapley values and network analysis in identifying impactful policy levers.

Conclusions:

• The proposed MLOps-based approach effectively identifies key drivers and validates policy interventions for SDGs.
• Water reuse and circularity present promising strategies for enhancing water security and achieving SDG targets.
• This methodology provides a robust framework for data-driven sustainability planning and decision-making.