Evaluation of rice processing suitability for different varieties of indica rice

In order to screen high-quality indica rice varieties suitable for rice processing, and establish a model for evaluating the suitability of rice processing. Twenty indica rice varieties with large planting area, high yield or pleasant taste in China were selected as raw materials for rice production. Correlation analysis and descriptive analysis were used to investigate the relationship between rice appearance quality, nutritional quality, cooking quality, and gelatinization characteristics with textural characteristics and sensory quality of rice. Sixteen core evaluation indicators with a variation coefficient greater than 10% and a significant correlation with rice quality were selected. They were whiteness, width, aspect ratio, moisture content, protein content, amylose content, straight branch ratio, taste value, valley viscosity, disintegration value, final viscosity, retrogradation value, iodine blue value, water absorption, expansion rate, and dry matter of rice soup. Using the maximum-minimum normalized method, multiple quality indicators of rice were calculated into a one-dimensional evaluation indicator. The obtained one-dimensional data were used as the comprehensive rice quality score, which were significantly correlated with the sensory score and the fitting degree was 0.824. A processing suitability model was established based on the screened rice raw material quality indicators and the rice comprehensive quality score. By comparing the fitting quality of the four methods, namely, stepwise regression, partial least squares, support vector machine, and neural network models, the integrated rice quality was predicted. The model built using stepwise regression was simple, had good fitting effect, and could quickly predict the comprehensive quality score of rice. The model built using neural network had the highest fitting degree, with a model determination coefficient of R² of 0.999 8 and a root mean square error RMSE of 0.008. Using the neural network model, the comprehensive quality score of rice could be accurately predicted, and the suitability of rice processing of different indica rice varieties could be judged. Finally, five high-quality indica rice varieties suitable for rice production were identified through cluster analysis, and the comprehensive quality score was 2.253-2.678, namely Liangfaxiangyouzhan, Guangliangxiang No.2, Tianyouhuazhan, Guangliangxiangsi, and Xiangzaoxian 45. This study can provide a scientific basis for comprehensive quality evaluation and breeding selection of indica rice.