Abstract
Fly ash (FA) is a residual from thermal industries that has been effectively utilized in the production of FA-based geopolymer concrete (FGPC). To avoid time-consuming and costly experimental procedures, soft computing techniques, namely, random forest regression (RFR) and gene expression programming (GEP), are used in this study to develop an empirical model for the prediction of compressive strength of FGPC. A widespread, reliable, and consistent database of compressive strength of FGPC is set up via a comprehensive literature review. +e database consists of 298 compressive strength data points. The influential parameters that are considered as input variables for modelling are curing temperature (T), curing time (t), age of the specimen (A), the molarity of NaOH solution (M), percent SiO2 solids to water ratio (% S/W) in sodium silicate (Na2SiO3) solution, percent volume of total aggregate (% A(G)), fine aggregate to the total aggregate ratio (F/A G), sodium oxide (Na2O) to water ratio (N/W) in Na2SiO3 solution, alkali or activator to the FA ratio (A(L)/F-A), Na2SiO3 to NaOH ratio (N-s/N-o), percent plasticizer (% P), and extra water added as percent FA (E-W %). RFR is an ensemble algorithm and gives outburst performance as compared to GEP. However, GEP proposed an empirical expression that can be used to estimate the compressive strength of FGPC. The accuracy and performance of both models are evaluated via statistical error checks, and external validation is considered. The proposed GEP equation is used for sensitivity analysis and parametric study and then compared with nonlinear and linear regression expressions.