Evolutionary optimization of machine learning algorithm hyperparameters for strength prediction of high-performance concrete

High-performance concrete (HPC) is designed to be more efficient and shows a higher value of flowability, strength, and durability in comparison to conventional concrete. The strength property is the most critical parameter in concrete structure design it shows a high non-linear correlation with the...

Full description

Saved in:
Bibliographic Details
Published in:Asian journal of civil engineering. Building and housing Vol. 24; no. 8; pp. 3121 - 3143
Main Authors: Singh, Sourav, Patro, Sanjaya Kumar, Parhi, Suraj Kumar
Format: Journal Article
Language:English
Published: Cham Springer International Publishing 01.12.2023
Subjects:
Building Materials
Civil Engineering
Engineering
Sustainable Architecture/Green Buildings
Support vector regression
High-performance concrete
Genetic algorithm
Multi-layer perceptron
XGBoost
Machine learning
ISSN:1563-0854, 2522-011X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:High-performance concrete (HPC) is designed to be more efficient and shows a higher value of flowability, strength, and durability in comparison to conventional concrete. The strength property is the most critical parameter in concrete structure design it shows a high non-linear correlation with the mixed proportioned ingredients due to its heterogeneous characteristic. Laboratory methods of determining the strength cause loss of resources, time, and materials; hence, numerous attempts to predict the compressive strength of HPC from its combined constituents have been made. The research work focuses on predicting the strength utilizing different machine learning (ML) algorithms such as multi-layer perceptron, support vector regression, and XGBoost with random search and genetic algorithm as a hyperparameter optimization technique. ML algorithms were trained and tested with multination datasets using the cross-validation method. The extreme gradient boosting ensemble algorithm (XGBoost) with genetic algorithm optimization technique showed better accuracy owing to a higher value of R 2 , and lower values of RMSE, MAE, and MAPE. The genetic XGBoost algorithm performed better in comparison to previously developed models on multination datasets showing better efficacy. A graphical user interface is also developed by the transformation of the ensembled model by means of providing easy to use access.
ISSN:1563-0854
2522-011X
DOI:10.1007/s42107-023-00698-y