IntroductionThe essential oil of pennyroyal has a number of bioactive properties, including antiviral, antifungal, and anti-inflammatory effects. Accurate, rapid, and non-destructive assessment of its quality is therefore a major challenge for the pharmaceutical industry. Conventional methods, such as gas chromatography, although effective, are time-consuming, destructive, and require the use of costly and polluting reagents. In contrast, mid-infrared spectroscopy, a non-invasive analytical method, provides spectral data while preserving sample integrity.MethodsIn this work, Fourier transform infrared spectroscopy (FT-IR) was coupled with machine learning techniques to detect and quantify the adulteration of pennyroyal essential oil with spearmint essential oil at different levels. Principal component analysis (PCA) was used to explore the samples and better understand the structure of the data. Partial least squares regression (PLSR) was then used to predict the levels of adulteration. Variable selection strategies, such as Interval PLS (iPLS) in forward and reverse modes, were applied to improve model performance and interpretability.ResultsBy identifying the most informative spectral regions, these methods improved the accuracy and resilience of the predictive models. The detrend pre-processing method produced the most efficient model, with an RPD of 8.29, an RMSEt of 1.25, and an R2t coefficient of determination of 99.47%. Most of the variable selection methods performed remarkably well, whether with forward or reverse iPLS, and regardless of the type of pretreatment used. Cross-validation R2 values ranged between 99.0% and 99.95%, underlining their exceptional predictive capacity, with low RMSE values between 0.75 and 1.1.DiscussionThese findings demonstrate how infrared spectroscopy and machine learning may be successfully combined to quickly, sustainably, and accurately assess the quality of pennyroyal essential oil, offering a powerful alternative to traditional chromatographic techniques for industry-level quality control.