Using　machine　learning　for　high　&　hybull;　throughput　screening　is　a　new　material　screening　method.　The　gas　adsorption　on　zeolite　molecular　sieves　was　studied　using　the　grand　canonical　Monte　Carlo　(GCMC)　simulation　and　machine　learning　methods.　The　GCMC　simulation　method　was　used　to　calculate　the　absolute　adsorption　capacity　of　12　types　of　electron　gases　on　240　varieties　of　silica　zeolite　molecular　sieves.　In　comparison,　the　Zeo++　program　was　employed　to　analyze　17　types　of　structural　characteristics　of　these　zeolite　molecular　sieves.　On　this　basis,　multiple　linear　regression　and　random　forest　regression　were　established　to　predict　the　adsorption　capacity　of　zeolite　molecu　&　hybull;　lar　sieves　for　electronic　gases.　Through　correlation　analysis　and　model　performance　evaluation,　the　impact　degree　of　different　structural　characteristics　on　gas　adsorption　capacity　was　revealed,　and　the　stability　and　prediction　accuracy　of　the　model　were　discussed.