Title :
Feature selection based on random forest and application in correlation analysis of symptom and disease
Author :
Hu Xue-qin ; Cui Meng ; Chen Bing
Author_Institution :
Inst. of Inf. on Traditional Chinese Med., China Acad. of Chinese Med. Sci., Beijing, China
Abstract :
A set of simple, rational diagnosis mode is the effective premise for intelligent diagnosis model. In this paper, selected the important symptoms of "five endogenous pathogens (FEP)" and measured these symptoms\´ contribution degree to FEP were main contents of this paper. Focused on the disease characteristics of "FEP", we introduced the method of random forest (RF), and used it to build feature selection evaluation criteria, then proved the effectiveness of this method. On this basis, the article also explored the effective way to build an intelligent diagnosis model for "FEP". Comparative experiment shown that RF model was superior in the diagnosis performance than the multi-classification support vector machine (SVM) classifier, and proven it to be an effective and high-performance "FEP" diagnosis model.
Keywords :
diseases; feature extraction; medical computing; patient diagnosis; pattern classification; FEP disease characteristics; SVM classifier comparison; disease-symptom correlation analysis; feature selection; five endogenous pathogens; intelligent diagnosis model; multiclassification support vector machine; random forest; rational diagnosis; Clinical diagnosis; Cognitive science; Diseases; Learning systems; Machine intelligence; Medical diagnostic imaging; Pathogens; Radio frequency; Support vector machine classification; Support vector machines;
Conference_Titel :
IT in Medicine & Education, 2009. ITIME '09. IEEE International Symposium on
Conference_Location :
Jinan
Print_ISBN :
978-1-4244-3928-7
Electronic_ISBN :
978-1-4244-3930-0
DOI :
10.1109/ITIME.2009.5236450
