Predicting post operative Major Adverse Cardiovascular Events (MACE) with hybrid AI models that combines data driven AI with ontology-based reasoning

Abstract

This thesis presents a method for predicting Major Adverse Cardiovascular Events (MACE). The approach combines two worlds: data-driven machine learning and knowledge-based reasoning. The core model is a classifier that outputs three possible answers: “yes,” “no,” or “I don’t know.” Instead of forcing uncertain cases into a yes/no label, the model shows its uncertainty. This uncertainty is based on disagreement among the trees in a random forest. When the model is unsure, the system does not stop there. Instead, it switches to an ontology-guided method to provide a more in depth result.