Multi-scale characterization and artificial neural network-based modeling identify leachate composition as a primary factor of cooked rice palatability.

The eating quality of cooked rice is governed by complex interactions among compositional, structural, and sensory attributes, yet integrated prediction model encompassing these multi-scale parameters are limited. This study evaluated eight rice varieties to identify key determinants of quality and to develop predictive models for overall acceptability. Multi-scale analyses encompassed rice flour composition, leachate properties, instrumental texture, microstructural characterization, and sensory properties. Microstructural observations revealed that thicker coated layers, greater starch granule disruption, and higher porosity were associated with increased leachate amylopectin content, contributing to enhanced stickiness, glossiness, and overall acceptability. Artificial neural network models were developed to evaluate the predictive power of different variable groups. Models using leachate characteristics alone exhibited substantially higher accuracy than those based only on rice flour composition, while integrating leachate, textural, and compositional variables produced the best overall performance. These results demonstrate that leachate composition and structure are dominant predictors of cooked rice palatability and that combining multi-scale parameters markedly improves prediction. This approach provides a scientific basis for more precise assessment of rice quality and supports the development of data-driven tools for rice breeding and product optimization.