Algorithm 6. The AI-Score ensemble description.

Input: SCB training dataset, SCB testing dataset. Output: prediction probabilities for each diagnosis class (Mild, Moderate, Severe) 1. Select a 5-fold split of the train SCB Dataset. 2. base_models = [“Ada Boost”, “Random Forests”,” XGBoost”] 3. meta_model = “CatBoost” 4. For each model in base_models: 5. Evaluate the model using 5-fold cross-validation. 6. Save all out-of-fold predictions. 7. Fit the model on the full training dataset and save. 8. Fit the meta-model on the out-of-fold predictions from the previous layer. 9. Evaluate the model on the SCB testing dataset. 10. For each class in the set of diagnoses do 11.  output prediction probability