Large　span　reticulated　shell　structures　have　broad　application　prospects.　However,　due　to　their　complex　structures　and　material　properties,　accurately　predicting　the　performance　of　large　span　reticulated　shell　structures　has　always　been　a　challenge.　Based　on　machine　learning　methods,　this　paper　establishes　an　intelligent　prediction　model　for　the　performance　of　large　span　K6　reticulated　shell　structures.　First,　3240　verified　K6　reticulated　shell　finite　element　models　are　established　for　a　series　of　parameters　that　affect　the　ultimate　bearing　capacity,　thereby　generating　the　database　required　for　machine　learning　algorithms.　Secondly,　nine　machine　learning　algorithm　models　are　established　based　on　the　open　source　platform　Python,　and　the　generated　data　sets　are　trained　and　tested.　Finally,　the　Shap　algorithm　is　used　to　explain　the　model　prediction　results.　The　results　show　that　the　root　mean　square　errors　(RMSE)　of　the　artificial　neural　network,　XGBoost,　and　gradient　boosting　in　predicting　ultimate　bearing　capacity　are　1.31,　1.83,　and　2.30　respectively,　and　the　R-squared(R²)　are　all　greater　than　0.98,　indicating　that　the　above　three　prediction　results　are　very　accurate.　The　Artificial　Neural　Network　model　has　the　best　performance　and　accuracy　in　predicting　the　ultimate　bearing　capacity　of　large　span　K6　reticulated　shell　structures,　with　a　mean　absolute　percentage　error　(MAPE)　of　0.023,　an　R2　value　of　0.998　%,　and　97　%　of　the　sample　relative　errors　between　−　10　%　and　10　%.　time,　indicating　that　it　has　high　prediction　accuracy　and　generalization　ability,　and　can　well　capture　the　nonlinear　relationship　and　complex　characteristics　between　the　ultimate　bearing　capacity　and　input　parameters　in　large　span　K6　reticulated　shell　structures.　©　2024　Institution　of　Structural　Engineers