Screening wavelengths with consistent and stable signals to realize calibration model transfer of near infrared spectra

Transfer learning based on dynamic time warping algorithms to improve qualitative analysis and quantitative prediction of rocks over multiple LIBS instruments

“…23–26 At the same time, some researchers have focused on calibration transfer without standard samples, as the standard samples are not available in all cases and their preservation and management can be a tedious task. Ni and others 27,28 proposed a wavelength screening method with stable and consistent signals and a wavelength screening method with scale invariant signals, and the regression model was built by selecting stable and consistent signals, and model sharing between the three instruments was achieved. To avoid the influence of environmental factors and test parameters on the spectra, Zhang 29 et al proposed a stability competitive adaptive reweighted sampling (SCARS) algorithm to select features and reduce the root mean square error of prediction (RMSEP) of the secondary instrument.…”

Application of stable consistency wavelength in optimizing gasoline RON near‐infrared analysis model transfer

“…The model transfer process was relatively simple, and only a few representative samples were selected for screening consistent wavelengths, so that the model transfer was easier to achieve. 15,16 Although the use of SWCSS algorithm reduced the number of wavelengths to a certain extent and simplified the model transfer process, this method may include highly correlated, uninformative, and unimportant wavelengths (variables) in the selection of consistent wavelengths, resulting in a decrease in the analysis ability of SWCSS-PLSR model for slave samples. Therefore, if the length optimization method of uninformative variable elimination (UVE), 17,18 competitive adaptive reweighted sampling (CARS), and successive projections algorithm (SPA) was used to eliminate the useless wavelengths in the SWCSS results, 19,20 it was beneficial to retain the important variables, and the ability to analyze the slave samples by using the remaining wavelengths to establish the master model will be enhanced.…”

“…Ni Lijun et al proposed a method (screening wavelengths with consistent and stable signals [SWCSS]) for model sharing between different near‐infrared spectrometers by using the difference spectrum analysis of master and slave spectrometers to select wavelengths. The model transfer process was relatively simple, and only a few representative samples were selected for screening consistent wavelengths, so that the model transfer was easier to achieve 15,16 . Although the use of SWCSS algorithm reduced the number of wavelengths to a certain extent and simplified the model transfer process, this method may include highly correlated, uninformative, and unimportant wavelengths (variables) in the selection of consistent wavelengths, resulting in a decrease in the analysis ability of SWCSS‐PLSR model for slave samples.…”

Calibration of Near Infrared Spectroscopy of Apples with Different Fruit Sizes to Improve Soluble Solids Content Model Performance

“…Antonucci et al [17] conducted a study on the internal quality of oranges by spectroscopic techniques and achieved good results in predicting their acidity and soluble solids by regression analysis using the PLS model, with correlation coefficients of 0.843 and 0.812 for soluble solids and acidity of oranges, respectively. Ni et al [18] performed the NIR spectral model transfer of different instruments by filtering the wavelength information of different NIR instruments. Two datasets of maize and scutellaria samples measured by different NIR instruments were used to test the performance of the method, where the overall prediction performance of the SWCSS-PLS model for the secondary measurement samples was much better than that of the full-wavelength PLS model.…”