An Application Study Of Support Vector Regression In The Statistical Forecasting Of Maternal Mortality Rate, Infant Mortality Rate And Child Mortality Rate Under Age 5 In Sichuan

An Application Study of Support Vector Regression in the Statistical Forecasting of Maternal Mortality Rate, Infant Mortality rate and Child Mortality Rate under Age 5 in SichuanObjectives: 1. To apply support vector regression(SVR) in the statistical forecasting of Maternal Mortality Rate(MMR),Infant Mortality Rate(IMR) and Child Mortality under age 5(U5MR) (hereinafter 'Three-Mortality') in Sichuan. In addition, the regularized radial basis function artificial neural network (RBFNN) will also be used to forecast the same data so as to compare the prediction performance of the two methods. 2. To provide the methodology reference for the government to put forward reasonable goal of 'Two NPAs' (shortened form of National Plan of Action for Women Development and National Plan of Action for Children Development). 'Three-Mortality' in Sichuan separately and the prediction results were compared by some statistical indexes.Results: The comparison results of the prediction performance between SVR and RBFNN show that SVR is much better than RBFNN for the data.Conclusions: 1. When dealing with small-sample data, SVM based on structural risk minimization principle is much better than RBFNN based on empirical risk minimization principle. Besides, SVM can reduce overfitting, thus it has better generalization performance than RBFNN. Therefore, SVM is better than RBFNN when predicting the data. 2. This research carries on a beneficial exploration in the methodology to put forward reasonable goal of 'Two NPAs'.