This　study　developed　a　machine　learning　framework　to　optimize　MPCM-integrated　concrete　for　compressive　strength　and　slump.　A　comprehensive　database　of　157　experimental　datasets　was　established.　Three　models　(SVM,　BPNN,　ELM)　were　evaluated,　with　ELM　showing　superior　performance　(R-2=0.93　for　strength,　R-2=0.73　for　slump).　Feature　analysis　revealed　water　content　as　the　most　influential　factor,　followed　by　MPCM　dosage　and　sand　content.　Experimental　results　showed　adding　40-50　%　extra　water　improved　slump　but　reduced　strength　by　45　%.　Superplasticizer　effectiveness　plateaued　beyond　10　%　dosage.　Multi-objective　optimization　using　PSO　generated　practical　mix　designs　meeting　target　specifications　(30　similar　to　50　MPa　strength,　20　cm　slump).　Experimental　validation　confirmed　prediction　accuracy　with　less　than　5　%　deviation.　The　optimized　mixes　maximized　MPCM　content　while　minimizing　cement　usage.　This　data-driven　approach　provides　reliable　guidance　for　sustainable　concrete　design.　Future　research　will　incorporate　additional　parameters　like　thermal　properties　and　expand　the　dataset　for　broader　applicability.　The　method　offers　significant　potential　for　energy-efficient　construction　applications.