Near infrared spectroscopic variable selection by novel swarm intelligence algorithm for rapid quantification of high order edible blend oil
作者：Xihui Bian*, Rongling Zhang, Peng Liu, Yang Xiang, Shuyu Wang, Xiaoyao Tan
關(guān)鍵字：Edible blend oil; Spectral analysis; Variable selection; Multivariate calibration; Whale optimization algorithm
論文來源：期刊
具體來源：Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2023, 284, 121788
發(fā)表時(shí)間：2023年
The quantification of single oil in high order edible blend oil is a challenging task. In this research, a novel swarm intelligence algorithm, discretized whale optimization algorithm (WOA), was first developed for reducing irrelevant variables and improving prediction accuracy of hexanary edible blend oil samples. The WOA is inspired by hunting strategy of humpback whales, which mainly includes three behaviors, i.e., encircling prey, bubble-net attacking and searching for prey. In discretized WOA, positions of whales were updated and then discretized by arctangent function. The whale population performance, iteration number and whale number of WOA were investigated. To validate the performance of selected variables, partial least squares (PLS) was used to build model and predict single oil contents in hexanary blend oil. Results show that WOA-PLS can provide the best prediction ability compared with full-spectrum PLS, continuous wavelet transform-PLS (CWT-PLS), uninformative variable elimination-PLS (UVE-PLS), Monte Carlo uninformative variable elimination-PLS (MCUVE-PLS) and randomization test-PLS (RT-PLS). Furthermore, if CWT is combined with discretized WOA, more parsimonious and efficient model can be obtained.