Application of Naive Bayes dichotomizer supported with expected risk and discriminant functions in clinical decisions — Case study

Abstract

In this paper, a case study on the application of Naïve Bayes dichotomizer in clinical decision supporting systems is described. The case study is about the diagnosis of the possibility of having Pervasive Developmental Disorder (PDD) in a child. The age group selected for the study is in between 2 and 3 years. Pervasive developmental Disorder is a neuro disorder that affects the social functioning, behavioural functioning and communication in a child. Conventional diagnosis is based on the scores obtained on checklists like DSM-IV Criteria. As Bayesian reasoning uses probability inferences, it is usually applied on decision making systems. Here for the study a Naive Bayes probabilistic dichotomizer was implemented. This dichotomizer calculates the most probable output depending on the inputs given to it, by applying the Bayes rule. Since the classifier is considering only two classes, the classifier is called as dichotomizer. The minimum expected risk and positive discriminant functions are also calculated, which again supports the decision of Naive Bayes dichotomizer. Implementation of Maximum A Priori Hypothesis and Maximum Likelihood Hypothesis are also discussing on the case study for a comparison. The main goal of this research work was to study the application of some probabilistic reasoning techniques in clinical decision supporting systems, where classification is more important. Based on the implementation of our case study, the findings shown that Naive Bayes dichitomizer supported with minimum expected risk and positive discriminant function, classifies correctly in clinical decision supporting systems.