Abstract: For the quantitative analysis of lube oil base oil components, three oil components, mineral oil (KN4010), hydrocarbon-based synthetic oil (PAO40), and synthetic ester (PriEco 3000) were selected as quantitative analysis objects, and then the mid-infrared spectral data of lube oil base oil samples formulated in different ratios were collected. The synergy interval partial least squares-binary grey wolf optimization algorithm (SiPLS-BGWO) combination optimization method was used to screen the characteristic wavenumbers in the full range to eliminate redundant invalid information and reduce the search space dimension. By optimizing the selection of characteristic wavenumbers, the SiPLS-BGWO approach not only enhanced the prediction accuracy but also demonstrated its ability to address challenges associated with overlapping spectral features in complex mixtures. The test results showed that the combined optimization model’s error indexes were significantly improved for the content prediction of mineral oil, hydrocarbon-based synthetic oil, and polyol ester. The RMSE (root mean square error) was reduced by up to 60.58% compared to using all spectral wavenumbers, and the fit indexes’ R2 values were higher than 99%. The significant reduction in RMSE underscored the method’s capability to identify and eliminate irrelevant or noisy spectral information, ensuring that the predictive model focused only on relevant features. In addition, the SiPLS-BGWO method had reduced the number of characteristic wavenumbers to less than 40, significantly reducing the operational burden and effectively improving the accuracy and applicability of the quantitative analysis model for multi-matter components. The ability to reduce the number of characteristic wavenumbers to below 40 demonstrated the algorithm’s efficiency in dimensionality reduction while retaining essential predictive information. The results affirmed that the SiPLS-BGWO model was a powerful tool for predictive modeling, providing a balance between accuracy and efficiency in the quantitative analysis of multi-component systems. And a novel framework for bridging the gap between spectral data complexity and actionable chemical insights, setting a precedent for future developments in the field.