Prognosis of Concrete Strength: The State of Art in Using Different Machine Learning Algorithms

Advancements in machine learning and their algorithm have been used in recent times to calculate the strength and robustness of concrete. The algorithm used for the prognosis of the concrete properties and their relationship with other constituents from the data has added ease to the prediction of strength. This study attempts to use advanced machine algorithms and categorization feasibility based on accuracy and implementation. This study is done by using gradient boosting, XGBoost, CATBoost, LightGBM, Random Forest Regressor, and AdaBoost Regressor to predict strength. This study categorizes the proportional analysis of advanced machine learning algorithms and earlier used algorithms for strength prediction and their effectiveness based on accuracy and the data features. Feature engineering shows the importance of feature variables and their relationships using a different algorithm and filters out the less important features. The insights of using such concepts bring numerous possibilities for reducing the errors for better predictions. This study can demonstrate different possibilities for making the infrastructure sustainable and predictable by studying the mechanical properties and other factors affecting the compressive strength of concrete using Machine Learning. The outcome of the different regressor models in this paper showcases that the performance of CATBOOST with MAE error of 2.69 is better compared to other algorithms utilized for prognosis of cement concrete.