Machine　learning　techniques　were　used　to　predict　the　shear　resistance　of　angle　shear　connectors.　A　database　of　151　experimental　datasets　from　push-out　tests　was　constructed,　identifying　eight　critical　features.　Three　predictive　models　were　constructed:　Support　Vector　Regression　(SVR),　Artificial　Neural　Networks　(ANN),　and　Gradient　Boosting　Decision　Trees　(GBDT).　The　ANN　model　demonstrated　superior　predictive　accuracy　and　generalization.　The　SHapley　Additive　exPlanations　(SHAP)　method　identified　the　length　of　the　angle　shear　connector　as　the　most　influential　factor,　while　the　height　had　minimal　impact.　Partial　Dependence　Plots　and　Individual　Conditional　Expectation　(PDP-ICE)　visualized　feature　influences　on　predictions,　providing　insight　into　the　relationships　between　key　parameters　and　shear　resistance.　©　2025　IABSE　Symposium　Tokyo　2025:　Environmentally　Friendly　Technologies　and　Structures:　Focusing　on　Sustainable　Approaches　-　Report　All　rights　reserved.