The key to improving the quality of preschool education lies in high-quality preschool teachers, and high-quality preschool teachers in turn depend on high-quality teacher education. A pilot project to “strengthen the training of kindergarten instructors” was recommended by the Ministry of Education. This research studies the resident training mode implemented by pilot normal colleges and universities, uses big data technology to collect the required data, and uses neural networks to evaluate the talent training mode of preschool teachers, and finally completes the following work: 1) This article introduces the research progress of the talent training mode of preschool teachers at home and abroad; 2) Introduced are the RBF neural network's fundamental ideas and algorithmic stages, and the evaluation index system for the training of preschool teachers is established; and 3) Experiment to determine the best model parameters, then feed the test data into the model you trained, and compare the output against expert evaluations.