A predictive analysis on medical data based on outlier detection method using non-reduct computation

Abstract

In this research, a new method to predict and diagnose medical dataset is discovered based on outlier mining method using Rough Sets Theory (RST). The RST is used to generate medical rules, while outliers are detected from the rules to diagnose the abnormal data. In detecting outliers, a computation of set of attributes or known as Non-Reduct is proposed by proposing two new formula of Indiscernibility Matrix Modula(iDMM D) and Indiscernibility Function Modulo (iDMFM D) based on RST. The results show that the proposed method is a fast detection method with lower detection rate. In conclusion, the computation of the Non-Reduct is expected to give medical knowledge that able to predict abnormality in dataset that could be used in medical analysis. © 2009 Springer.