This study implements a set of machine learning algorithms to building material science, which predict the compressive strength of zeolite and diatomite substituting concrete mixes in sodium chloride solution. Particularly, Random Forest, Support Vector Machine, Extreme Gradient Boosting, Light Gradient Boosting, and Categorical Boosting algorithms are exploited and their optimal parameters are tuned. In the training and testing of these models, 28 day, 56 day, and 90 day compressive strength observations of 63 samples of 7 different concrete mixtures substituting Portland cement, zeolite, diatomite, zeolite + diatomite were used. Consequently, compressive strength experimentation results and machine learning predictions were compared through statistical methods such as RMSE, MAPE, and R2. Results denote that the prediction performance of machine learning is improving with tuned models. Particularly, RMSE, MAPE, R2 scores of Categorical Boosting are, respectively, 1.15, 1.45%, and 98.03% after parameter tuning design. The results denote that presented machine learning model can provide an advantage in the cost and duration of the compressive strength experiments.