Optimized machine learning models for predicting the tensile strength of high-performance concrete
Divesh Ranjan Kumar, Pramod Kumar, Pradeep Thangavel, Warit Wipulanusat, Chanachai Thongchom, Pijush Samui
Abstract
This study presents an effective implementation of machine learning models for predicting high-performance concrete (HPC) tensile strengths. This prediction is carried out using 714 data points gathered from previous experimental studies, and 70% of the data are used to train the model. The remaining 30% of the data are divided into two subsets for testing and validation of the model. Four predictive algorithms, XGBoost, ADABoost, CATBoost, and GBM, are employed to predict the tensile strengths of HPC. The computational efficiency of the proposed models was assessed using several performance metrics. Additionally, a regression plot, triangle diagram, regression error characteristic (REC) curve, and accuracy matrix were drawn to evaluate and compare the performance of the proposed models. Uncertainty analysis was also performed for all the models to assess their reliability and robustness in predicting the tensile strengths of HPC. The XGBoost model has the highest performance in training, testing, and validation, followed by ADABoost, GBM, and CATBoost. The developed robust soft-computing-based prediction methodology can serve as a reliable alternative for predicting high-performance concrete tensile strength.