Score Prediction in Accreditation Process

Based on Hybrid Models integrating Machine Learning Regression, Rule-Based Accreditation Constraints, and Explainable AI (SHAP)

ML Regression

XGBoost, Random Forest, and LightGBM for predicting accreditation scores at criterion level

Rule-Based Constraints

Rubric-aligned predictions ensuring scores remain within valid accreditation boundaries

Explainable AI (SHAP)

SHAP values explaining how each indicator contributes to the predicted accreditation score

📊 Data Management

Upload and manage accreditation data

🤖 Model Training

Train and compare ML models

🎯 Prediction

Predict accreditation scores

💡 Explainability

SHAP explanations and feature importance

Research Framework

Hybrid Model Components

  • 1.ML Regression: XGBoost, Random Forest, LightGBM
  • 2.Rule-Based Constraints: Rubric validation, score boundaries
  • 3.Explainable AI: SHAP values for feature contribution

Evaluation Metrics

  • •MAE: Mean Absolute Error
  • •RMSE: Root Mean Squared Error
  • •R²: Coefficient of Determination
  • •Rubric Validity: Constraint compliance