Optimizing concrete compressive strength prediction with a deep forest model: an advanced machine learning approach

Abstract

ABSTRACT Accurate prediction of concrete compressive strength is essential for ensuring the durability and safety of concrete structures. This study utilizes the Deep Forest (GC Forest) model to predict compressive strength based on nine key factors: Granulated Blast Furnace Slag, Pulverized Fuel Ash, Mixing Water, High-Range Plasticity Reducer, Crushed Stone, Sand, Curing Time, Portland Cement, and Compressive Strength. The Deep Forest model's performance was compared with 12 other machine learning models, including Linear Regression, k-Nearest Neighbors, Decision Tree, Support Vector Machine, Neural Network, Random Forest, and Gradient Boosting Machines. Model evaluations were conducted using the coefficient of determination (R2) on a dataset of the listed parameters. The Deep Forest model achieved the highest predictive accuracy, with an R2 value of 9.55, outperforming all other models. This result demonstrates the model's robustness, owing to its ensemble-based architecture that enhances generalization. The study emphasizes the effectiveness of advanced machine learning models like Deep Forest in optimizing concrete compressive strength predictions, contributing to improved structural design and quality control.