Edible oils may become contaminated with harmful substance residues during transportation, posing a serious threat to food safety and public health. This study utilized Fourier Transform Near-Infrared (FT-NIR) Spectroscopy to extract spectral data for kerosene content in soybean and corn oil. Three feature selection models, Competitive Adaptive Reweighted Sampling (CARS), Bootstrapping Soft Shrinkage (BOSS), and Iteratively Variable Subset Optimization (IVSO), were applied to Savitzky-Golay (SG) preprocessed data. Using the selected features, Partial Least Squares (PLS) regression models were developed. The CARS-optimized PLS model demonstrated superior generalization performance, achieving an RMSEP of 2.4520 mg·kg