Rapid identification of coffee species and origin using affordable multi-channel spectral sensor combined with machine learning.

The rapid identification of coffee species and origin is critical for ensuring quality control and authenticity in the coffee industry. This study explores the use of an affordable multi-channel spectral sensor, AS7265X (410-940 nm), combined with machine learning techniques to achieve accurate classification of coffee species and origins rapidly and non-destructively. Spectral data were collected in three LED configurations: the original 18 spectral bands and two additional configurations of the data into 24 and 30 spectral features using configured LED emitters. The coffee samples included two species, Arabica and Robusta, with four distinct origins from Indonesia: Arabica Flores (AF), Arabica Gayo (AG), Robusta Dampit (RD), and Robusta Temanggung (RT). Four machine learning algorithms viz. Linear Discriminant Analysis (LDA), Artificial Neural Networks (ANN), Random Forest (RF), and Support Vector Machines (SVM) were employed, with hyperparameter tuning executed through cross-validation techniques. Additionally, physicochemical analysis was performed randomly on each coffee bean sample and principal component analysis (PCA) was performed as an exploratory analysis of the data. Our findings demonstrate that coffee species identification achieved a perfect accuracy of 100 % using LDA on the 24 and 30 spectral features. For coffee origin identification, the highest validation accuracy of 0.917 was attained with LDA using the 24 raw spectral features. Additionally, data pretreatment methods were applied and their impact on classification performance was evaluated. Still, all of them did not provide any improvement to the classification performance. The results underscore the efficacy of the AS7265X sensor combined with LDA for reliable and rapid coffee species identification. Furthermore, this approach presents a promising, cost-effective solution for coffee origin identification, enhancing quality control processes in the coffee industry.