Abstract: The intelligent development of coal mining leads to decreasing annual coal mine safety accidents, yet safety production still requires constant vigilance. Accurate prediction of mine gas emission is vital to ensuring safe production and improving efficiency. Conventional prediction methods are deficient for their complex calculation and insufficient accuracy in dealing with high-dimensional data, unable to satisfy modern intelligent management of coal mines. Therefore, proposes a Principal Component Analysis-Hunter Prey Optimization-Extreme Learning Machine (PCA-HPO-ELM) model for gas emission prediction of intelligent mines: 1) 13 key influencing factors such as coal seam thickness and mining depth were selected, and Principal Component Analysis (PCA) was used to reduce the data from 13 dimensions to 4 dimensions. This not only reduced the dimension but also retained the main information, laying a foundation for model training; 2) Hunter Prey Optimization (HPO) algorithm was introduced to solve the randomness of the input weights and hidden layer threshold selection of the traditional Extreme Learning Machine (ELM) model, and the accurate prediction of gas emission is realized. PCA-HPO-ELM, PCA-PSO-ELM and PCA-ELM models were compared using the same data for the proposed models. Results show that the PCA-HPO-ELM model exhibited better iteration speed than the PCA-PSO-ELM model, and the determination coefficient R² of predicting mine gas emission was 0.993 76, higher than that of the other two (0.988 54 and 0.894 3, respectively), showing superiority; the model can be used for reference to improve the prediction accuracy and efficiency of intelligent mine gas emission.