基于支持向量回归的颅内压时间系列无损估计方法

In this paper, a data mining framework has been proposed to estimate a desirable time series from a set of its related time series without knowing an explicit prior model relating them in our previous work. The mapping function only uses simple linear relation to depict the error and characteristics, the advantages of linear function method is simple, easy to describe is not completely describe. Actually, the actual situation may exist between them is more complicated nonlinear relation, we need to adopt complicated method to construct mapping function, thus to build the more accurate relationship between the eigenvalue and the error, that help to improve the accuracy of estimates. In this paper, we used the support vector regression analysis method to construct the nonlinear mapping function, the expectation of time series data mining framework for time series prediction can be more accurate result. Comparison of different mapping function: prediction of intracranial pressure obvious influences have adopted the method, based on support vector regression, the nonlinear mapping function is obviously superior to base on the Linear Least Squares of linear mapping function.