Chemometric Methods In Molecular Spectroscopy Quantitative Analysis Of Atomic Spectral Classification

Chemometrics applied mathematics and statistical analysis technique in chemistry and chemical analysis, In the 1970s an independent chemical branch formed, at present mainly have Multiple Linear Regression and Support Vector Machine, Principal Component Analysis and Partial Least Squares and Artificial Neural Network, etc..Among them, PLS has a small forecast squared residuals of the modle, thus has the high prediction stability, the parameters of the modle and the actual problems of experience laws are consistent. Through the compression reduced-order of original data, PLS could improve the model's performance, and be more appropriate for handling the problems of variables samples but less number.The back-propagation (back propagation neural network, BPN) algorithm is a kind of artificial neural networks ,and is used to establish to describe the nonlinear relation with weights between target and variable. BPN is a special kind of nonlinear regression analysis with less independent variables as input of the network, and predict unknown variables. BPN can worked with PLS and PCA algorithm et al to establish new arithmetic. The Support Vector Machine (SVM) method is based on statistical learning theory of VC dimension theory and structure risk minimization principle, by choosing different kernel functions to find the optimal planar space in order to avoid the loss of information ,and obtain more accurate results.In recent years,the Surface Enhanced Raman Scattering (SERS) spectroscopy have been widely used in many fields, such as life science, catalytic chemistry and environmental chemistry, etc. And the chemometrics methods can be used with SERS to establish the model not only for qualitative analysis but also for quantitative analysis.For this purpose, our research work includes the following aspects: 1. The elementary principles and features of PLS,BPN,SVM method, Raman spectra and SERS spectra.2. PLS-BPN methods are used in Quantitate Analysis the Raman Spectrogram of mixed amino acid solutions of cystine and cysteine.And try to take advantage of chemometrics methods to determine glycine, cysteine, methionine,histidine four components mixed amino acids aqueous solution of the Raman spectra data.3. Prepared silver sol as SERS substrate to quantitative analysis of the dopamine by BPN-PLS methods.4. By SVM methods,set models of energy level,isotope shift,etc. to classify the unknown electronic configuration levels of heavy metal.Compaired with several other chemometrics methods, Support Vector Machines (SVM) have more accurate prediction ability.