Advanced Machine Learning Approaches for Prediction of Compressive Strength of Sustainable Concrete

Predicting the compressive strength of concrete using SCMs is difficult since the relationships between the mix components are not linear. The fact that these parts interact with one another in complex ways is one explanation. In order to address this, five distinct kinds of math-based tools were evaluated. Random Forest, which aggregates findings, was one of these methods. Gradient boosting is an alternative method that iteratively improves upon previous mistakes. Next up is XGBoost, a popular choice for situations when speed is crucial. As an alternative to making a sudden leap, AdaBoost refines its predictions incrementally. The Multiple Linear Regression method, on the other hand, is exclusive to linear patterns. How was the data utilized? A database containing 1030 distinct recipe mixtures, with 80% taught and 20% tested. Performance wasn’t evaluated based on intuition but rather four metrics: R squared demonstrated fit, root mean square error highlighted the significance of large mistakes, mean absolute error quantified the average amount of errors, and percentage deviation demonstrated the extent to which estimations fell off. The reliability and consistency of the predictions were shown numerically. Out of all the techniques that were tested, boosting emerged as the clear winner, easily outperforming bagging and conventional linear models. With a test R squared value of 0.927, XGBoost achieved the best score. How does SCM concrete react under pressure? Tangled designs are much easier to handle with these tools. When the issue fights back fiercely, a steady, step-by-step improvement might emerge victorious.