ORIENTATE

ORIENTATE applies automated machine learning classification and systematic feature selection to develop and evaluate predictive models for clinical and preventive healthcare.

Key Features:

• Automated Model Generation: Generates multiple classification models from chosen features and target variables and evaluates them for comparison.
• Cross-Validation and Model Selection: Employs cross-validation techniques to assess model performance and identify optimal models.
• Custom Feature Selection Algorithm: Implements a custom algorithm that systematically searches combinations of predictors to identify high-performing feature sets.
• Performance Metrics: Optimizes and reports performance using metrics such as F1 score and ROC AUC.
• Comprehensive Reporting and Global Interpretation: Produces detailed reports with graphs and global interpretation methods for model explanation.
• Feature Relevance and Interaction Plots: Provides plots of feature relevance and interactions to support statistical inference and interpretability.

Scientific Applications:

• Clinical predictive modeling: Develops classification models to predict clinical outcomes, exemplified by predicting need for a second sedation in pediatric dental patients.
• Pediatric dental research: Applied to a dataset of healthy children and children with special health care needs (SHCN) who underwent deep sedation to identify predictive factors.
• Preventive healthcare and early prediction: Supports early prediction tasks in preventive care through data-driven risk stratification.
• Feature discovery and hypothesis generation: Identifies predictive factors—for example, number of teeth with pulpar treatments—as candidates for further clinical investigation.

Methodology:

Integrates automated feature selection with cross-validation-based model validation; a custom algorithm systematically explores predictor combinations and optimizes performance metrics (F1 score, ROC AUC); results are interpreted using global interpretation methods and feature interaction plots.