Providing access to underground energy resources such as oil, gas, and geothermal energy requires drilling through subterranean rock formations. Drilling fluids are commonly used to enable this drilling process by serving a variety of functions. Chief among these is circulating into the drilled wellbore to remove the produced rock cuttings. Other functions include exerting the hydrostatic pressure necessary to prevent the flow of underground fluids into the wellbore while drilling, minimizing the invasion of solids and undesired fluids into the wellbore rock, and ensuring the fluid remains flowable within the means of the pumps available on site. To assess the ability of a drilling fluid to serve these functions, it has to have certain rheological properties, which are conventionally measured using specialized equipment. Considering that the formulation and components of these drilling fluids can vary greatly for different scenarios, the process of preparing samples and measuring their rheological properties, which is usually performed on a daily basis on a drilling site, is unavoidably time-consuming, repetitive, and error-prone. Based on this, it is apparent that there is a need for a computational model that can accurately predict drilling fluids properties based on the proposed concentration of their components without the need for further laboratory testing. This study describes a novel methodology to train a machine learning model derived from over 6,878 drilling fluid formulations to successfully predict water-based drilling fluids properties with a resulting R2 of 91.07 ± 6.35%.