Uranium-plutonium mixed oxide (MOX) serves as a fuel in fast reactors, which are being developed as the next generation of nuclear reactors. The MOX fuel fabrication process is a complex process involving numerous steps with intricate parameters. Therefore, the optimal fabrication conditions are determined by highly skilled engineers and researchers who perform preliminary tests after reviewing the parameters based on past knowledge. However, access to the MOX fuel fabrication equipment and data collection during the fabrication process are limited due to the need to confine the MOX in an airtight container called a glove box. Therefore, determining the ideal fabrication conditions is a time- and cost-intensive process.
 Herein, a model was developed to predict the optimal fabrication process for MOX fuel by compiling a database of existing fuel fabrication data and using machine learning to analyze the relationship among feedstock powder type, fabrication parameters, and relative density. This model enables predictions such as changes in relative density when powders with different properties are used and fabricated under the same parameters (Fig. 1), and has higher prediction accuracy than models reported in other studies. The proposed model enables quick determination of various parameters for MOX fuel fabrication without preliminary testing and/or professional expertise.