COMBINING DIFFERENT INFERENCE METHODS FOR MEDICAL DECISION SUPPORT SYSTEMS

Abstract

In this study, synthetic data with 100, 1000 and 2000 records have been produced to reflect the probabilities on the ALARM network. In this study, a medical diagnosis system called as DRCAD is presented. DRCAD system is more innovative and interesting than the classical diagnosis support systems. In other words, DRCAD collects possible diagnosis of the patients from two sub modules. Each of these two sub modules gives the possible diagnosis from symptoms in a specific confidence degree. The proposal of two sub modules are combined linearly and diagnosis decisions are presented as a list. Each of the sub modules consist of Bayesian inference and rule-based inference models respectively. As a result, the methods in which the conclusions are combined in a linear manner are 5% more successful than the Rule Based Method when applied individually and 30% more successful than the cases where the Bayesian Network Based Method is utilized.