This research aims to provide a rapid, non-destructive solution for monitoring ionic liquid mass fraction in pharmaceutical production by proposing and validating a systematic chemometric modeling framework based on near-infrared (NIR) spectroscopy. The study first identifies 1350?1650 nm as the optimal analytical band through spectral mechanism analysis. Subsequently, based on the spectral data of 59 ionic liquid samples, the efficacy of various preprocessing and feature selection algorithms is systematically compared. Standard normal variate (SNV) is identified as the optimal preprocessing method, and competitive adaptive reweighted sampling (CARS) is determined to be the superior feature selection algorithm. The final SNV-CARS-PLSR quantitative model, integrated with partial least squares regression (PLSR), demonstrates excellent predictive performance: a root mean square error of calibration of 0.0689, a calibration determination coefficient of 0.9836, a root mean square error of prediction of 0.1008, and a prediction determination coefficient of 0.9666. To verify the universality of this methodological framework, it is further applied to a quantitative analysis of a glucose aqueous solution system, which also achieves outstanding prediction accuracy (0.9993). The systematic optimization strategy established in this study not only provides robust technical support for the real-time monitoring of ionic liquids but also offers a versatile methodological reference for the application of micro-NIR spectroscopy in the fine chemical industry.