The issue of adu

lteration in the high-end edible oil market is becoming increasingly severe，with a notable lack of effective rapid detection methods. In this study，

Xanthoceras sorbifolia

oil was used as a case study. Artificially adulterated samples with

Xanthoceras sorbifolia

oil content ranging from 0% to 100% at 5% increments were prepared，and their near-infrared（NIR） spectra were collected. Partial least squares discriminant analysis（PLS-DA） was employed to develop a method for identifying adulterated

Xanthoceras sorbifolia

oil. Several preprocessing techniques were evaluated and optimized to enhance the performance of the model. To address the issue of quantifying the adulteration ratio in

Xanthoceras sorbifolia

oil，a model was constructed using partial least squares regression（PLSR）. Model performance was further optimized by integrating different spectral preprocessing techniques and wavelength selection methods.The results indicate that the PLS-DA model，using convolutional first derivative preprocessing，achieved a 98% accuracy rate in identifying different types of adulterated oils. Furthermore，combining first derivative deconvolution preprocessing with competitive adaptive reweighted sampling（CARS） variable selection yielded NIR model RPD values of 5.90，40.00，5.20 and 4.90 for the adulterated oils in

Xanthoceras sorbifolia

oil，demonstrating excellent predictive performance. This study successfully conducted both qualitative and quantitative analyses of

Xanthoceras sorbifolia

oil samples adulterated with various types and proportions of other oils，offering an exploratory approach for rapid quality assessment of

Xanthoceras sorbifolia

oil adulteration.