New strategy of sample set division in spectroscopy analysis——SWNW

The division of sample sets is a primary issue in analyzing the composition of complex solutions based on spectroscopy. For a limited number of samples, the selection of a calibration set for the built model with sufficient robustness and universality is the main goal of this research. Therefore, we proposed the single-weight nationwide (SWNW) sample set division strategy. To verify the practicality of this sample set division, we designed a spectral data collection experiment of isolated blood samples, with red blood cells counting as the measured component. The method is compared with the KS algorithm and the SPXY algorithm. The evaluation indicators are the correlation coefficient and the root mean square error, obtained by partial least square modeling. The experimental results in the modeling analysis show that the model built by the calibration set, selected by the SWNW algorithm is more stable than the KS algorithm and the SPXY algorithm. The correlation coefficient of the established model is 2.5% higher than that of the SPXY algorithm, and the root mean square error is reduced by 13.7%, which are significant improvements. The analysis results of the prediction set also show the robustness of the SWNW algorithm.