Vegetable oils contain traces of heavy metals that can cause irreversible damage to human health. The present study employed near-infrared spectroscopy and variable selection in conjunction with partial least squares (PLS) for the rapid determination of Cd content in peanut oil. Firstly, the spectral data of peanut oil test samples were preprocessed by different preprocessing methods, and the best preprocessing method was selected according to the results obtained by the PLS regression model. Then, PLS regression models were established to determine Cd content in peanut oil by variable iterative space shrinkage approach (VISSA), competitive adaptive reweighted sampling (CARS), multiple feature spaces ensemble strategy with least absolute shrinkage and selection operator (MFE-LASSO), and bootstrap soft shrinkage (BOSS), respectively. The results show that all four feature optimization algorithms could improve the prediction accuracy of the model. Among them, the CARS-PLS model had high prediction accuracy. Its prediction coefficient of determination (R2P) was 0.9666, the root mean square error of prediction (RMSEP) was 2.8207 mg/kg, and the relative prediction deviation (RPD) was 5.4705, respectively. In summary, near-infrared spectroscopy combined with chemometrics could be used for rapid quantitative detection of Cd in peanut oil.

来源平台：INFRARED PHYSICS & TECHNOLOGY