Domain constraints improve risk prediction when outcome data is missing

arXiv:2312.03878v2 Announce Type: replace
Abstract: Machine learning models are often trained to predict the outcome resulting from a human decision. For example, if a doctor decides to test a patient for disease, will the patient test positive? A challenge is that historical decision-making determines whether the outcome is observed: we only observe test outcomes for patients doctors historically tested. Untested patients, for whom outcomes are unobserved, may differ from tested patients along observed and unobserved dimensions. We propose a Bayesian …

abstract arxiv challenge constraints cs.lg data decision disease doctor domain example human machine machine learning machine learning models making patient positive prediction risk test type will