基于支持向量机的模型传递方法研究

Model Transfer Method Based on Support Vector Machine

模型传递是以数学方法通过在2台不同仪器之间寻求一种变换关系来增强光谱仪数据通用性、可比性的一种基本途径。由于实际测量数据具有非线性特征,加上校正样本集合的有限性,使得解决小样本条件下非线性关系的模型传递问题显得尤为重要。文章在概述支持向量机基本原理的基础上,探讨了支持向量机方法在光谱仪的模型传递问题中的应用,提出了基于支持向量机的分段直接校正方法,最后采用计算机模拟的方式对该方法进行了举例说明,并和人工神经网络方法进行了相应的比较。

The model transfer is a basic method to build up universal and comparable performance of spectrometer data by seeking a mathematical transformation relation among different spectrometers. Because of nonlinear effect and small calibration sample set in fact, it is important to solve the problem of model transfer under the condition of nonlinear effect in evidence and small sample set. This paper summarizes support vector machines theory, puts forward the method of model transfer based on support vector machine and pieeewise direct standardization, and makes use of computer simulation method, giving a example to explain the method and compare it with artificial neural network in the end.