Alkali-activated　materials　are　significant　functional　materials　widely　utilized　in　industrial　and　technological　sectors.　They　efficiently　handle　solid　waste,　reduce　cement　consumption,　and　play　a　crucial　role　in　sustainable　construction.　However,　due　to　their　complex　reaction　mechanisms　and　differences　from　traditional　cementitious　materials,　further　research　is　necessary　to　explore　optimal　proportioning　methods　that　fully　harness　their　advantages.　In　this　study,　machine　learning　techniques　were　employed　to　rapidly　and　effectively　design　alkali-activated　material　mixtures.　With　a　total　of　47　samples,　27　were　used　for　training　and　20　for　testing.　The　BP　neural　network　model　was　optimized,　and　analysis　using　the　random　forest　algorithm　revealed　that　the　CaO　content　contributes　the　most　to　compressive　strength.　This　research　provides　novel　insights　and　practical　guidance　for　alkali-activated　material　proportioning,　reducing　experimental　costs,　and　expanding　the　application　of　machine　learning　in　the　engineering　field.