Hybrid discrimination method for samples classification in medicine

As indispensable solutions in classification problems, discrimination for samples has been employed in medicine, and it is performed subjectively by physicians at present, which hinders the diagnosis and treatment in medicine. In this paper, a hybrid discrimination method (HDM) in medicine is proposed, which consists of two phases, including attribute selection phase and discriminant phase. In attribute selection phase, critical attributes are selected from the original features by linear correlation and C5.0 decision tree. In discriminant phase, samples are discriminated by discriminant analysis. This discrimination method is evaluated through five datasets of chronic hepatitis B, cardiac Single Proton Emission Computed Tomography (SPECT) images, Lung Cancer, Hepatitis survival and Iris plant for demonstrating its viability and applications. Finally, this proposed method has obtained the critical clinical lab indicators and discriminants related to three syndromes in CHB dataset, and it also performs well than some typical classification methods in the other four datasets for its broader applications.