Comparative study on predicting the compressive strength of oven-cured fly ash and slag-based geopolymer concrete using soft computing techniques

Abstract

Sustainable concrete is created using waste materials and contributes to environmental conservation. In this study, fly ash and slag were key ingredients for this sustainable concrete. The focus was on constructing a predictive model using various input variables. To achieve this, machine learning techniques were employed: random forest (RF), support vector machine (SVM), and long short-term memory (LSTM). These algorithms were used to predict the properties of structural concrete produced from fly ash and slag. The predictive model was precisely trained and tested using experimental data. The assessment of model performance involved a comparative analysis based on two metrics: the coefficient of determination score and root mean square error. The coefficient of determination values for the RF, SVM, and LSTM were 0,8439; 0,8668; and 0,8694; respectively. The root mean square error values were 4,0318; 3,9692; and 3,6921 for RF, SVM, and LSTM, respectively. The results of this study demonstrate that the LSTM model outperformed both the RF and SVM models. This suggests that the LSTM algorithm is particularly suitable for capturing complex patterns and relationships within the data, making it well-suited for predicting the properties of sustainable concrete based on fly ash and slag.