A Study on the Prediction of Compressive Strength of Geo-Polymer Concrete through Stacking Regression and Random Forest

Abstract

Concrete plays a pivotal role in construction, and the emergence of innovative variations such as geo-polymer concrete holds significant promise for sustainability. This research focuses on predicting compressive strength in geo-polymer concrete, utilizing both Stacking regression and random forest algorithms. The models achieved notable R2 values of 0.72 and 0.89, indicating their ability to make accurate predictions. Compressive strength ranged from 20 MPa to 65 MPa, with an average of 40 MPa. Moreover, models with higher R2 values of 0.89 and 0.83 exhibited increased accuracy, showing differences between predicted and observed compressive strengths (highest: 90 MPa, lowest: 10 MPa, average: 45 MPa). These results highlight the effectiveness of machine learning techniques, particularly the random forest method, in forecasting the performance of geo-polymer concrete. Furthermore, they underscore the critical importance of meticulous model selection and evaluation in engineering applications to ensure dependable predictions. KEYWORDS: Machine learning, Concrete, Geo-polymer concrete, Compressive strength