Abstract
Characterization factors are used in life cycle assessments to translate the quantity of chemicals and other pollutants produced during a product's life cycle into the standard unit of an impact category, such as ecotoxicity. The chemical ecotoxicity (HC50) evaluation is an expedient method to determine the harmful effects of these chemicals on ecosystems. Ecotoxicity parameters for a wide variety of compounds are therefore a genuine and significant concern. The current work employed an optimized XGBoost model for real value prediction of HC50 values. Further the optimized prediction model is analysed using a series of model agnostic explainable learning approaches. The best optimized model achieved an acceptable R2 = 0.684 with mse = 0.587. Further, the best optimized model is subjected to various global and local explainable algorithms, resulting in turning the black box model into a white box model, resulting in interpretation of how the decisions are being made by the learned model.
| Original language | English |
|---|---|
| Article number | 103574 |
| Journal | Ain Shams Engineering Journal |
| Volume | 16 |
| Issue number | 9 |
| DOIs | |
| State | Published - Sep 2025 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 9 Industry, Innovation, and Infrastructure
-
SDG 12 Responsible Consumption and Production
Keywords
- 2D-ALE
- ALE
- Feature interaction
- HC50
- SHAP
- XGBoost
Fingerprint
Dive into the research topics of 'Understanding chemical ecotoxicity through interpretable ensemble learning'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver