Prediction for Hepatitis E Incidence Using Support Vector Machine

Hepatitis E is an acute viral hepatitis caused by hepatitis E virus, which has become a public health problem threatening people's health. Study on incidence of hepatitis E is effective in prevention and control of hepatitis E. We take the incidence of hepatitis E in Shandong, China, as a case. We studied the periodicity of hepatitis E incidence, and proposed a method to obtain the exact period of hepatitis E, in order to improve prediction performance. Then, we adopt support vector machine (SVM) to predict the incidence of hepatitis E. To make full use of correlation among data, we propose three modeling methods for SVM, including horizontal modeling, vertical modeling, and cross modeling. We take periodicity into account for prediction. To verify the effectiveness of our proposed method, we did a comparative experiment with ARIMA, which is the most commonly used method for predicting hepatitis E. Experiments show that the correlation in and between periods is helpful to improve the prediction accuracy. Especially, our proposed CM-SVM method has good performance and stability for hepatitis E prediction.