Multivariate Calibration in Chemical Analysis

In chemometrics, multivariate calibration methods provide a convenient way to determine several components in a mixture within one experimental step, without the tedious operation of separation of these components. The method of calculation usually used is PLS method. Artificial neural network is also often used especially when the data set exhibits obvious nonlinearity. But it is prone to overfitting. Therefore, several types of techniques have been developed to prevent overfitting. At the same time, support vector regression, as a method suitable to treat nonlinear data without serious overfitting, can be used as a new method of computation in multivariate calibration. An example of using SVR in multivariate calibration will be described as follows. 

Amino acids are the structural units of protein, and some of them have been used as drugs or food additives, so the determination of amino acids is useful for biochemical research and for commercial product analysis. Among essential amino acids, there are three aromatic amino acids, phenylalanine, tyrosine and tryptophan, which exhibit fluorescence when they are excited by ultraviolet rays. So it is possible to determine them by the fluorescence spectroscopic method. The A ax of phenylalanine, tyrosine and tryptophan are 282 nm, 303 nm and 348 nm respectively, but their fluorescence spectra are partially overlapped. Since the separation operation of these three amino acids is tedious and 

As the methods of computation, the data of 23 samples are treated by support vector regression, artificial neural networks and PLS concurrently. The errors of leave-one-out (LOO) cross-validation test are listed in Table 13.2. 

Algorithm Root mean square error Mean absolute error 

From Table 13.2 it can be seen that the best result is obtained by SVR with Gaussian kernel. 

---