Using RVA-full pattern fitting to develop rice viscosity fingerprints and improve type classification

The rapid visco analyser (RVA) method has been widely used to investigate starch viscosity and to quality control starch based cereals. Different rice types vary significantly in composition, which is reflected by their different pasting properties. However, the RVA has a reduced sensitivity when it is used to identify types. This study combined RVA profiles with multivariate data analysis methods to obtain more information from the RVA profiles. A total of 152 rice profiles were collected and analyzed by principal component analysis (PCA) and partial least squares-discriminate analysis (PLSDA). The results showed that there were two rice subspecies group types that could be distinguished. In order to optimize the established model, all the data were subjected to a regression analysis and the initial stage was identified as the point when viscosity differences between the groups had the highest regression coefficients. The discrimination accuracy improved when the extracted viscosity data was used. For further verify the applicability of the models, 60 of unknown samples were examined, and the classification accuracy was 100%. All the findings confirmed that this study offers a practical and reliable way to predict rice types, providing a foundation for further studies on origin discrimination and age prediction.