A new ensemble classifier for multivariate medical data

Selecting an efficient classifier for medical data is considered as one of the most important part of today's computer aided diagnosis. The performance of single classifiers such as decision tree classifier can be increased by ensemble method. However, this approach relies on the data quality and missing values. In this paper, we propose a new ensemble classifier to overcome overfitting and biasness issues of traditional classifiers as applied to multivariate medical data with missing values. Medical professionals do not believe in filling the missing values by any of the existing statistical methods because each case is different in medical science. The proposed ensemble model was compared with the bagged tree classifier using Ggraph. The results of this study indicate that, the proposed ensemble classifier is able to achieve better accuracy of more than 96 percent without filling up missing values; and it does not suffer from over-fitting and biasness issues.