ABSTRACT
Machine learning technology is commonly used for the prediction of the compressive strength of geopolymer composites. This research is focused on using algorithms ensembled by heterogeneous regression methods with stacking. Modelling is done with variables such as fly ash, fine aggregate, coarse aggregate, sodium hydroxide (NaOH), Sodium Silicate (Na2SiO3), molarity, added water, ground granulated blast furnace slag (GGBS), superplasticizer, curing time, and curing temperature. A total of 376 data points were collected from the standard literature. Various statistical metrics, such as mean absolute error (MAE), correlation coefficient (R), and root mean square error (RMSE), are used to measure model results. The algorithm developed shows 90% efficiency. This accuracy of data suggests that the proposed stacked combination algorithm would help the construction industry in the prediction of the amount of constituent required for an expected compressive strength of any of the above-listed input data, as now one can curb the unnecessary ingredients and promote only required ingredients as per the suggested method. A strong association was found between machine learning models and experimental findings.
Disclosure statement
No potential conflict of interest was reported by the authors.