Integrating Machine Learning and Deep Learning for Compressive Strength Prediction

Prediction of compressive strength in concrete is very important to ensure the safety and durability of any structure. In this paper, we introduce a machine-learning hybrid approach whereby we combine the strengths of two different models: CatBoost, working extremely well with structured data, and a neural network to grasp the underlying complex patterns within the data. Next, these models are combined, and their prediction is fine-tuned by another layer of linear regression to increase overall accuracy. The hybrid model achieved an RMSE of 3.86 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 0.94, demonstrating its superior accuracy compared to the individual models. Herein, the proposed method has been applied to the standard concrete strength data set, and the results of hybrid performance give more reliable predictions than either model alone. It is a promising technique that can find its practical applications in construction, providing insight into concrete strength and enabling engineers to make better decisions.