Predicting the Strength of Concrete in an Intelligent Manner: A Research Grounded on Random Forest

Abstract

Concrete is one of the most extensively used materials in civil engineering due to its versatility, durability, and cost-effectiveness. As a fundamental construction component, its mechanical properties, especially compressive strength, play a crucial role in ensuring the structural integrity and safety of buildings and infrastructure. Accurate prediction of compressive strength is therefore essential in both the design and quality control phases of construction projects. Traditional prediction methods such as empirical formulas or lab testing are time-consuming and error-prone. This study uses a Random Forest (RF) regression model to improve prediction accuracy, utilizing a public dataset. Techniques including GridSearchCV, standardization, and cross-validation were implemented. The optimized RF model achieved an R² score of 0.88. Feature importance analysis revealed that cement, water, and age were the most influential factors. These studies underscore the potential of RF and similar ensemble learning techniques as effective tools for material property estimation. Future work may consider integrating more diverse datasets, exploring deep learning models, or embedding domain-specific constraints to further improve performance and applicability in real-world construction scenarios.