Evaluation of ANN and Tree-Based Techniques for Predicting the Compressive Strength of Granite Powder Reinforced Concrete

Abstract

Machine learning is widely used in engineering applications. In civil engineering, soft computing has been investigated for the prediction of concrete mechanical properties. In this study, the performance of soft computing techniques in predicting the compressive strength of granite powder reinforced concrete is investigated. A total of 108 observations were collected from the existing scientific literature. Different soft computing techniques considered in this book chapter are Artificial Neural Network, Random Forest, Reduced Error Pruning Tree and Random Tree. According to the findings, the RF-based modelling approach performs best for the estimation of compressive strength of granite powder concrete. REP Tree-based model indicated the lowest values of CC which are 0.9598 and 0.9214 for the training and testing stage, respectively. The curing days (CD) appear to be the most significant input variable for determining the compressive strength (N/mm2) of granite powder concrete using this data set, according to sensitivity analysis.