Explainable AI and machine learning-based analysis of municipal solid waste generation rate: a South African case...

Area of Science:

• Environmental Science
• Data Science
• Urban Planning

Background:

• Solid waste generation presents a growing global challenge requiring sustainable management strategies.
• Existing machine learning (ML) models for waste forecasting often overlook multi-factor complexity, multicollinearity, explainability, and regional variations.
• This research addresses the need for a more comprehensive and interpretable approach to waste generation analysis.

Purpose of the Study:

• To develop and validate a multi-stage ML framework for analyzing complex waste generation patterns.
• To identify key socio-economic, demographic, meteorological, and infrastructural drivers of waste generation.
• To provide actionable insights for municipal planners to optimize waste management policies and resource allocation.

Main Methods:

• Integration of Principal Component Analysis (PCA) for dimensionality reduction.
• Application of k-means clustering to identify distinct waste generation profiles.
• Utilisation of SHapley Additive exPlanations (SHAP) for model interpretability.
• Development of an Adaptive Neuro-Fuzzy Inference System (ANFIS) for predictive modeling.

Main Results:

• PCA effectively reduced data complexity, retaining 90.3% variance in 13 principal components.
• K-means clustering identified 3 distinct groups based on service access and infrastructure levels.
• SHAP analysis highlighted refuse removal access, relative humidity, population density, and household income as significant predictors.
• The ANFIS model, optimized with PCA and Grid Partitioning, achieved high predictive accuracy (R² = 0.8943).

Conclusions:

• The developed ML framework offers a robust and interpretable method for analyzing complex waste generation data.
• Key socio-economic and environmental factors significantly influence waste generation, necessitating tailored management strategies.
• The findings support data-driven decision-making for targeted policy formulation and resource optimization in municipal waste management.