Improved model population analysis in near infrared spectroscopy

Abstract

Model population analysis has been widely used as an effective variable selection method in near infrared spectroscopic analysis. In this study, two model population analysis have been studied and improved i.e. bootstrapping soft shrinkage (BOSS) and interval variable iterative space shrinkage approach (iVISSA). The improved approach was (i) using the reproducible variables i.e. choosing the most consistent variables and applying iterative retained informative variables (IRIV), and (ii) using the uninformative variable elimination based on Monte Carlo (MC-UVE) for unstable variables. This study compares the proposed model with BOSS, iVISSA, and a hybrid model By using four different datasets. The results show that the proposed model outperformed BOSS, iVISSA, and VCPA-IRIV model in all the four datasets.