Assessing mineral profiles for rice flour fraud detection by principal component analysis based data fusion

The present work proposes to detect adulteration in rice flour using mineral profiles. Eighty-seven flour samples from two rice kinds (Indica and Japonica) plus thirty adulterated flour samples were analyzed by ICP OES. After obtaining the quantitative elemental fingerprint of the samples, PCA and LDA were applied. Binary and multiclass associations were considered to assess rice flour authenticity through fraud identification. Models based on element predictors showed accuracies ranging from 72 to 88% to distinguish adulterated and unadulterated samples. The fusion of the mineral features with the principal components (PCs) obtained from PCA provided classification rates of 100% in training samples, and 91-100% in test samples. The proposed method proved to be a useful tool for quality control in the rice industry since a perfect success rate was achieved for rice flour fraud detection.

Automated summary

This study aimed to detect adulteration in rice flour using mineral profiles, achieving high accuracy with models based on element predictors. By combining mineral features with principal components obtained from PCA, the method proved to be a useful tool for quality control in the rice industry, achieving a perfect success rate for rice flour fraud detection.