Current manuscript presents a study on the use of 48 experimental data points containing parameters of concrete production technological process and its properties, such as strength, density, and bending strength. It was revealed that temporal characteristics, specifically - compressive strength at the age of 3, 7, 28 days, R3, R7, and R28, are significantly correlated with each other, indicating that only one characteristic, such as R28 or Rfl 28, is sufficient for prediction. The absence of multiple correlations between parameters and properties suggests that linear regression analysis may not be accurate. Therefore, the use of Machine Learning is optimal; specifically Random Forest method is preferable due to ease of use and minimum hyperparameters for tuning. Low prediction errors (∼1–11%) for 30% of the test data, as determined by the cross-validation method, confirm a relationship between the experimental parameters and the concrete properties. The most important parameters for achieving high values of compressive and bending strengths, R28 and Rfl 28, were identified, namely: air-entraining additives, granite crushed stone consisting of a mixture of fractions 5–20 mm, crushed stone derived from gravel of high strength grains of large fractions 10–20 mm. To obtain explanatory model, another Machine Learning method, that was used, called Decision Tree. The model showed that a high amount of crushed stone 10–20 mm from gravel, more than 212 (kg per 1 m3 of concrete mix), leads to a higher number of strong grains with smooth, rounded surface, thereby, reducing the bending strength of concrete. However, a large concentration of crushed stone mix fractions of 5–20 mm from granite, more than 537 (kg per 1 m3 of concrete mix), leads to the maximum roughness, which makes a significant contribution to the increased strength of concrete due to the adhesion of the matrix and aggregates to each other.