Bayesian nets and medical diagnosis. A different way to learn conditional probabilities

Abstract

Bayesian networks are a formal tool that allows to model processes characterized by uncertainty, which is typical of many real problems. A Bayesian network can establish a comprehensive model on a set of random variables and their relationships. This model can be used to estimate probabilities of certain variables of the network, which are called state variables, when other variables, named evidence variables, are fixed. The process of obtaining the probability distribution of the state variables, when the evidences are fixed, is named Bayesian probabilistic inference. In this work, after introducing the Bayesian networks, it is exposed how to use them in the classroom analysing medical diagnosis problems. This leads to a more meaningful learning of the concepts of independence and conditional probability, which are essential for the correct application of numerous probabilistic and statistical methods.

Keywords

• redes bayesianas
• teorema de bayes
• modelización matemática
• diagnóstico médico