Brown rice authenticity evaluation by spark discharge-laser-induced breakdown spectroscopy

Rice is the most consumed food worldwide, therefore its designation of origin (PDO) is very useful. Laser-induced breakdown spectroscopy (LIBS) is an interesting analytical technique for PDO certification, since it provides fast multielemental analysis requiring minimal sample treatment. In this work LIBS spectral data from rice analysis were evaluated for PDO certification of Argentine brown rice. Samples from two PDOs were analyzed by LIBS coupled to spark discharge. The selection of spectral data was accomplished by extreme gradient boosting (XGBoost), an algorithm currently used in machine learning, but rarely applied in chemical issues. Emission lines of C, Ca, Fe, Mg and Na were selected, and the best performance of classification were obtained using k-nearest neighbor (k-NN) algorithm. The developed method provided 84% of accuracy, 100% of sensitivity and 78% of specificity in classification of test samples. Furthermore, it is simple, clean and can be easily applied for rice certification.