Decision Making System Based on Bayesian Network for an Agent Diagnosing Child Care Diseases

摘要

In some cases a pediatrician seeks help from super specialist so as to diagnose the problem accurately. In a Mutli-agent environment, an agent called Intelligent Pediatric Agent (IPA) is imitating the behavior of a pediatrician. The aim is to design a decision making framework for this agent so that it can select a Super Specialist Agent (SSA) among several agents for consultation. A Bayesian Network (BN) based decision making system has been designed with the help of a pediatrician. The prototype system first selects a probable disease, out of 11; and then suggests one super specialist out of 5 super specialists. To verify the results produced by BN, a questionnaire containing 15 different cases was distributed to 21 pediatricians. Their responses are compared with the output of the system using KS test. The result suggests that 91.83% pediatricians agree with the result produced by the system. So, we can conclude that BN provides an appropriate framework to imitate the behavior of a pediatrician during selection of an appropriate specialist.