Probabilistic graphical models for the medical industry developed using enhanced learning algorithms

Abstract

The earliest artificial intelligence approaches to medical diagnosis were based on Bayesian and decision-theoretic schemes. Difficulties in obtaining and representing quantities of numbers and both the computational and representational complexity of probabilistic schemes caused a long-lasting departure from these approaches [1]. In numerous domains, such as genetics, clinical diagnoses, direct marketing, finance, and online business, data sets arise with thousands of variables and a low ratio of cases to variables. Such data present dimensional difficulties for classifying a target variable [2] and identifying critical predictor variables [3]. Hence, determining influence (i.e., the causal relationships between the target variable and predictor variables) is a great challenge.