Table 1 Two main problems addressed in Clinical Decision Support Systems (CDSSs): Given a patient, we denote the feature vector as x, label as y, and action as a; two problems include (1) a predictive modeling task f predicts the outcome \(\hat{y}\) aligning closely with a human physician’s judgment y; (2) a decision-making task that maximizes a specific reward function r(x, a)
From: Clinical decision making under uncertainty: a bootstrapped counterfactual inference approach
Predictive Modeling | \(f: x \rightarrow \hat{y}\) | \(\hat{y}\) and y be as close as possible |
Decision-Making | \(h: x \rightarrow a\) | a achieves maximum reward |