Combination of multiple model population analysis and mid-infrared technology for the estimation of copper content in Tegillarca granosa

Abstract The aim of this study was to use mid-infrared spectroscopy coupled with multiple model population analysis based on Monte Carlo-uninformative variable elimination for rapidly estimating the copper content of Tegillarca granosa . Copper-specific wavelengths were first extracted from the whole spectra, and subsequently, a least square-support vector machine was used to develop the prediction models. Compared with the prediction model based on full wavelengths, models that used 100 multiple MC-UVE selected wavelengths without and with bin operation showed comparable performances with Rp (root mean square error of Prediction) of 0.97 (14.60 mg/kg) and 0.94 (20.85 mg/kg) versus 0.96 (17.27 mg/kg), as well as ratio of percent deviation (number of wavelength) of 2.77 (407) and 1.84 (45) versus 2.32 (1762). The obtained results demonstrated that the mid-infrared technique could be used for estimating copper content in T. granosa . In addition, the proposed multiple model population analysis can eliminate uninformative, weakly informative and interfering wavelengths effectively, that substantially reduced the model complexity and computation time.