Mathematical Model and Machine Learning Techniques to Predict the Compressive Strength of Groundnut Shell Ash Blended Sandcrete

Abstract

This study uses machine-learning (ML) methodologies to introduce predictive models for the compressive strength of sandcrete mixed with groundnut shell ash (GSA). The models were developed utilizing 140 datasets acquired from published articles. The datasets contained several input variables: aggregate-to-binder ratio, peanut shell ash concentration, and curing time. The output feature was the compressive strength of the sandcrete. Four mathematical and machine-learning models were used to predict the compressive strength of peanut shell ash-blended sandcrete. Based on analyses of several models, the boosted decision tree model outperformed others in predicting compressive strength. The sensitivity analysis outcomes of the boosted decision-tree model show that the aggregate-to-binder ratio was the most significant factor in determining compressive strength. Overall, the boosted decision-tree model achieved an R² of 0.965 during testing, indicating excellent predictive accuracy. Additionally, it was found that using 10% to 30% GSA as a cement substitute optimally enhances sandcrete strength. These findings contribute to the understanding of sustainable construction materials and support the practical application of GSA in construction.