A Machine Learning-Based Water Potability Prediction Model by Using Synthetic Minority Oversampling Technique and...

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Water quality degradation is a significant global issue.
• Accurate water quality prediction is crucial for environmental management.
• Existing machine learning models lack transparency.

Purpose of the Study:

• To compare machine learning algorithms for water quality classification.
• To enhance model transparency using explainable AI (XAI).
• To identify key features influencing water quality predictions.

Main Methods:

• Comparative analysis of Support Vector Machine (SVM), Decision Tree (DT), Random Forest, Gradient Boost, and Ada Boost algorithms.
• Dataset normalization using Z-score.
• Handling imbalanced data with Synthetic Minority Oversampling Technique (SMOTE).
• Feature importance analysis using Local Interpretable Model-agnostic Explanations (LIME).

Main Results:

• Random Forest and Gradient Boost models achieved the highest accuracy at 81%.
• Explainable AI (XAI) techniques, specifically LIME, were successfully applied.
• Key features contributing to water quality classification were identified.

Conclusions:

• Machine learning, particularly Random Forest and Gradient Boost, offers effective water quality classification.
• Integrating XAI improves model interpretability, addressing a critical limitation.
• Feature importance analysis provides insights for targeted water quality management strategies.