Risk evaluation using evolvable discriminate function

Abstract

This essay proposes a new approach to risk evaluation using disease mathematical modeling. The mathematical model is an algebraic equation of the available database attributes and is used to evaluate the patient condition. If its value is greater than zero it means that the patient is ill (or in risk condition), otherwise healthy. In practice risk evaluation has been a very difficult problem mainly due its sporadic behavior (suddenly, the patient has a stroke, etc as a condition aggravation) and its database representation. The database contains, under the label of risk patient data, information of the patient condition that sometimes is in risk condition and sometimes is not, introducing errors in the algorithm training. The study was applied to Atherosclerosis database from Discovery Challenge 2003 - ECML/PKDD 2003 workshop.