 This study proposes a new method for selecting parameters of the probability integral method, PIM, which is commonly used in China for predicting surface deformation in mining areas. This method combines a multiple regression model with an extreme learning machine, ELM. It was tested on the Huainan mining area and found to have high accuracy in predicting the parameters of PIM. The results show that the ELM can accurately select the parameters of PIM, which can then be used to predict surface deformation in mining areas. This article was authored by Shenzhen Qi, Lei Wang, and Xu Xiaomu.