The teaching experiment design takes Python’s machine learning integration package as the core, takes coal, rock and gas composite dynamic disaster prediction as the background, and is written in Python language, which facilitates students to realize the programming task of machine learning and mining data in limited teaching and experiment courses. The experiment contents include data set construction, Apis call, data set reading, data normalization processing, model training and export, sample set prediction, model accuracy test and other links. This teaching experiment involves interdisciplinary subjects and has strong practicability, which can improve students’ ability to use appropriate modern analysis technology and tools.