Abstract: To further enhance the reliability and applicability of parameter estimation for surface movement prediction under ultra-thick loose layer mining, this study takes the Xinji mining area as the research object. Data from multiple working face observation stations were collected to systematically analyze the influence mechanisms of geological and mining factors on the parameters of the probability integral method. By comprehensively employing three algorithms—grey relational analysis, principal component analysis, and artificial neural networks—a regression model was established to solve and analyze the parameters for predicting surface movement. The results show that grey relational analysis performs well in eliminating abnormal data, principal component analysis effectively reduces dimensionality and simplifies the model structure, and the neural network algorithm achieves high accuracy in predicting certain parameters. Each of the three methods has its own advantages and can complement one another, collectively improving the accuracy and reliability of parameter estimation. This research provides a scientific basis and technical support for mining subsidence prediction and control under similar geological conditions.