An knowledge acquisition approach for colliery gas forecast expert system based on rough sets theory was proposed. This method firstly forecast samples were established between gas data and gas outburst intensity relationship; then algorithms of continuous attribute discretization, attribute reduction and rules extraction based on rough sets theory, and knowledge from lots of forecast samples was obtained automatically, and expert system knowledge database was constructed; Finally, based on reasoning machine, real-time data of gas forecast was realised. Example analysis results show that rough sets theory in coal mine forecast of gas expert system knowledge acquisition validity and practicability.