To address the issue of adulteration in grape seed oil, a quantitative analysis method for detecting adulteration of grape seed oil with sunflower seed oil was studied. Simulation experiments were conducted by applying portable near-infrared Raman spectroscopy combined with a least absolute shrinkage and selection operator-least squares support vector machine (LASSO-LS-SVM) algorithm to samples with varying adulteration levels. A total of 11 mixed oil samples with varying adulteration concentrations were prepared in this research. A portable 785 Raman spectrometer was utilized to acquire the Raman spectra of all mixed oil samples. A wavelet algorithm was adopted for the baseline correction, noise reduction, and normalization of the raw spectral data. A machine learning algorithm was used to extract the eigenvectors of the Raman spectra, and a quantitative analysis model was established. The LASSO algorithm reduced the dimensionality and extracted features of the spectral data, and the LSSVM algorithm enabled successful construction of a quantitative analysis model for adulteration. The coefficient of determination (R2) of the model for the test set is 0.984 6, and the RMSE is 0.024 9. The integration of the LASSO-LS-SVM algorithm enables the application of Raman spectroscopy for quantifying adulteration levels in grape seed oil. The proposed method demonstrates substantial application value and promising commercial potential for promoting the domestic grape seed oil market.