Abstract: The chemical and mechanical sedimentation of porous carbonate reservoir blurs geometric shape and external structure of clastic reservoir sedimentary microfacies in space, and complicated the physical properties of reservoirs of different origins. Conventional sedimentary microfacies modeling methods are difficult to reproduce objectively the complex spatial distribution of different microfacies, thereby the accuracy of facies control attribute modeling could be reduced. Therefore, taking the Cretaceous Mishrif Formation bioclastic limestone in the Middle East H Oilfield as the research object, we established a three-dimensional sedimentary microfacies model in the study area by carrying out wave impedance inversion, porosity inversion, and permeability inversion, using machine learning methods. The phase-controlled attribute modeling was performed through the variogram analysis of different microfacies. The sedimentary microfacies model established using machine learning methods conforms to the changes in facies sequence of marine carbonate platforms, fully reflecting the spatial morphology of microfacies and the contact relationship among microfacies. The reservoir attribute model that established with the sedimentary microfacies as a constraint not only meets the requirement of probability consistency between simulation results and known data, but also reflects the spatial variation characteristics of the reservoir by phase zones.