The small scale reactor has diversified applications and broad prospects for development, and is currently a hot research reactor type. The core nuclear design of small scale reactors needs to consider the reasonable optimization of key physical parameters for multiple cycles. It is of great significance for the nuclear design of small scale reactors to conduct efficient and precise rapid calculation of core nuclear design parameters and their changes during the burnup process. This paper studies a rapid prediction method for key core nuclear design parameters of small scale reactors based on machine learning algorithms. Two mapping models between core states and effective increment factors are established using BP neural network and convolutional neural network. The comparison results show that the convolutional neural network prediction is more accurate. This model can be used for the selection and optimization of core design schemes for small scale reactors and has important engineering application value.