Comparative Evaluation of Machine Learning Models for Predicting Compressive Strength of Concrete Made with Soap Factory Wastewater

Abstract

This study explores how machine learning (ML) models can predict the compressive strength of concrete when soap factory wastewater is used as a partial replacement for mixing water. Using a dataset with key mix parameters like cement content, water-to-cement ratio, aggregates, and curing age, five ML models—Gradient Boosting, Random Forest, Decision Tree, k-Nearest Neighbors (kNN), and Linear Regression—were tested. Among them, Gradient Boosting performed best, with the highest accuracy (R² = 0.868) and lowest error. Random Forest also showed strong results, while Decision Tree was moderately effective. In contrast, kNN and Linear Regression struggled to capture the complex relationships in the data. These findings highlight the potential of ML in optimizing concrete mix design and promoting sustainable construction practices.