Objective The coal seam gas content is a core parameter for resource assessment and development, but current gas content prediction models generally suffer from issues such as insufficient accuracy and weak generalization ability, which hinder the exploration and development of coalbed methane.

Methods Based on the logging response characteristics of coal seam gas content in Southern Yanchuan gas field, the MIV (Mean Impact Value) method was utilized to optimize the logging parameters. The BP neural network and random forest algorithm were introduced to establish a high−precision coal seam gas content prediction model.

Results Compared to the traditional multiple linear regression model, both the BP neural network and random forest models achieved notably higher prediction accuracy, with the random forest model performing even better.

Conclusions The Random Forest model is more suitable for predicting the coalbed methane content in the study area. Based on the model's prediction, the distribution range of gas content in the gas field is 4.84~21.83 m³/t, with an average of 11.63 m³/t. Spatially, the gas content of coal seam increases gradually from southeast to northwest, and its variation law is generally consistent with the buried depth pattern of coal seam. Vertically, with the increase of buried depth, the gas content of coal seam increases gradually, but the dispersion degree of gas content distribution increases.