Explainable Hybrid Ensemble Models for Agricultural Soil Erosion Mitigation
Students & Supervisors
Student Authors
Supervisors
Abstract
Soil erosion is a major environmental and agricultural challenge, leading to land degradation, reduced crop productivity, and long-term ecological imbalance. Traditional assessment techniques often rely on empirical or rule-based models, which are limited in scalability and adaptability to diverse environmental conditions. This study introduces a hybrid stacking ensemble framework for soil erosion risk prediction, formulated as a binary classification task (HighErosion vs. LowErosion). Eight baseline machine learning models such as Logistic Regression, Naïve Bayes, Decision Tree, Random Forest, K-Nearest Neighbors (KNN), AdaBoost, Gradient Boosting, and XGBoost were systematically evaluated against the proposed hybrid ensemble. The methodology integrates preprocessing, hyperparameter tuning, and interpretability techniques such as feature importance and SHAP analysis. Experimental results demonstrate that the hybrid stacking ensemble outperformed all baselines, achieving 98.07% accuracy, an F1-score of 0.981, and an AUC of 0.997. Key predictors included rainfall, slope, and soil type, while vegetation cover contributed secondary but significant effects. The findings highlight the potential of ensemble-based machine learning for developing decision-support tools that enable policymakers and farmers to implement proactive erosion control strategies, ensuring sustainable land management.
Keywords
Publication Details
- Type of Publication:
- Conference Name: 28th International Conference on Computer and Information Technology (ICCIT-2025)
- Date of Conference: 19/12/2025 - 19/12/2025
- Venue: Cox’s Bazar, Bangladesh
- Organizer: IEEE Bangladesh Section