근적외 스펙트럼을 이용한 최적 다중선형희귀모델을 위한 알고리듬

Abstract

Near infrared(NIR) spectral data have been used for the noninvasive analysis of various biological samples. Nonetheless, absorption bands of NIR region are overlapped extensively. Even though 1st or 2nd derivative data are used, it is very difficult to select the proper wavelengths of spectral data, which give the best multiple linear regression(MLR) models for the analysis of constituents of biological samples. To find the best MLR models, all-possible combinations of available variables(in this case, wavelengths of spectral data) were derived by in-house programs written in MATLAB codes. All of the extensively generated regression models were compared in terms of standard error of calibration(SEC), R² and standard error of prediction(SEP) to find the best regression models. For the test of the developed program, aqueous solutions of BSA(bovine serum albumin) were prepared and analyzed. As a result, the best MLR models can be found using 1st and 2nd derivative spectra and SEP criteria.