Support vector machine based classification analysis of SARS spatial distribution

Abstract

Support Vector Machine has been applied for the research of disease symptom diagnosis and epidemic spread prediction, and its effective and precise classification function could be used for clustering analysis of epidemic spatial distribution. Several indices reflecting the environmental, social, economic and traffic conditions of provinces in mainland China synthetically were selected for the fundamental training data for classification analysis of SARS spatial distribution in this paper. The original index data were sifted by bivariate correlation analysis with SARS cases distributed in provinces, and the selected index data and SARS cases were normalized for faster convergence speed and higher predictive precision. The relationship between SARS occurrence and environmental, social, economic and traffic factors were searched eventually. The results show that SARS epidemic indicates the characteristic of spatial clustering and has strong relationships with those typical factors. During the process of SARS epidemic spread, the provinces with similarities in environmental, social, economic and traffic conditions indicate some certain consistencies more or less.