The classification performance of binomial logistic regression based on classical and Bayesian statistics for screening P-Thalassemia

Abstract

Statistics plays an important role in many areas especially in classification tasks. Logistic Regression Model is one popular technique to solve problems, in particular, medical problems. P-Thalassemia, a common genetic disorder, lends itself to is interesting for using MLR to classify types of P-Thalassemia. There are several types of Thalassemia in the world, especially Thailand. From many methods to construct mathematical models, there are two approaches to generate these models, namely Classical and Bayesian Statistics. According to different views of both approaches, using MLR based on both approaches was selected to classify types of P-Thalassemia. The results show that classification results of all models based on Bayesian Statistics yield a greater accuracy percentage than using Classical Statistics (an accuracy percentage of this data set was 99.2126). Both approaches give different results because of the source of parameter, the transformation processes and data types are affect the classification performance based on using MLR In the future, we will use the model most suitable for implementing Thalassemia Expert System. © 2011 AICIT.