Machine learning-enhanced Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectroscopy for discrimination of green Arabica and Robusta coffee beans

Arabica and Robusta are the two primary commercial coffee species, and reliable discrimination between their beans is critical for detecting adulteration. This investigation employed Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectroscopy as a rapid, non-destructive technique to classify 14 coffee bean samples (10 Arabica, 4 Robusta). Spectral data were preprocessed using Savitzky-Golay (SG) smoothing, multiplicative scatter correction (MSC), and standard normal variate (SNV) transformation. Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) were then applied for exploratory analysis and classification model development. Key results demonstrated that SG smoothing optimally preserved essential spectral features while reducing noise and enhancing the signal-to-noise ratio. The SG-PCA approach achieved superior classification performance, whereas MSC-PCA and SNV-PCA showed reduced effectiveness. The PCA-LDA algorithm proved highly effective, yielding comparable or better classification than PCA alone. Critically, SG preprocessing significantly enhanced the PCA-LDA model’s performance. This work establishes ATR-FTIR spectroscopy combined with machine learning (PCA-LDA) as a robust, specific method for discriminating Arabica and Robusta coffee beans. The approach demonstrates strong potential for rapid, cost-effective coffee authentication in commercial and regulatory settings. • Savitzky-Golay (S-G) smoothing effectively preserved essential spectral features of ATR-FTIR spectra • Principal Component Analysis (PCA) was effective for dimensionality reduction and feature extraction • SG-PCA approach achieved the highest classification performance • Linear Discriminant Analysis (LDA) were employed for exploratory analysis • SG-PCA-LDA highly effective for constructing classification models