MA separation is crucial because it reduces the heat generation of high-level liquid waste generated after reprocessing spent nuclear fuel, thereby reducing the area required for disposal facilities. Its introduction in next-generation aqueous reprocessing is being considered. An extraction system capable of highly separating trivalent Am and Cm from trivalent rare earth elements in a high nitric acid concentration range is required; however, the construction of an extraction system with sufficient performance for commercial use has not yet been achieved. We are continuing to search for more selective and practical extractants. However, since actinide experiments generate radioactive waste and high-throughput experiments are also expensive, we have developed an efficient search scheme that utilizes machine learning. To create a predictive model, it is necessary to prepare a diverse range of data for training extraction. Then, we have decided to combine the data from the International Database on Extractant Ligands (IDEaL), developed by the OECD Nuclear Energy Agency. The database provides comprehensive information on extractants essential for actinide separation from spent nuclear fuels. Initiated by the Expert Group on Fuel Recycling, it aims to support researchers by compiling key data to optimize separation processes. Following the "State-of-the-Art Report on Nuclear Fuel Cycle Chemistry" (2018), the database now includes 439 extractants with detailed entries. Plans for beta testing and data verification are underway, making IDEaL a valuable tool in nuclear chemistry. We have developed a scheme for extracting data from IDEaL, performing data wrangling, and then augmenting the training data by combining it with existing extracted data. This involves extracting feature values from the molecular structure of the extractant and utilizing convolutional neural networks to estimate the extraction rate. The selection of the learning model and detailed parameter settings is currently underway, but we aim to create applications that utilize such databases and develop actinide separation chemistry. The entire process is designed to run on the AACE program, which is produced at the Institute of Science Tokyo. In the presentation, the current status of development, scheme, and some results of regression attempts will be shown.