A novel whale optimization algorithm optimized XGBoost regression for estimating bearing capacity of concrete piles

• Published in: Neural computing & applications 2023-02, Vol.35 (5), p.3825-3852
• Main Authors: Nguyen, Hieu, Cao, Minh-Tu, Tran, Xuan-Linh, Tran, Thu-Hien, Hoang, Nhat-Duc
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial Intelligence; Artificial neural networks; Bearing capacity; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Concrete piles; Data Mining and Knowledge Discovery; Image Processing and Computer Vision; Load tests; Model accuracy; Optimization; Optimization algorithms; Original Article; Probability and Statistics in Computer Science; Rank tests; Static loads
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-022-07896-w

This paper presents a hybrid model combining the extreme gradient boosting machine (XGBoost) and the whale optimization algorithm (WOA) to predict the bearing capacity of concrete piles. The XGBoost provides the ultimate prediction from a set of explanatory experiment variables. The WOA, which is configured to search for an optimal set of XGBoost parameters, helps increase the model’s accuracy and robustness. The hybrid method is constructed by a dataset of 472 samples collected from static load tests in Vietnam. The results indicate that the hybrid model consistently outperforms the default XGBoost model and deep neural network (DNN) regression. In an experiment of 20 runs, the proposed model has gained roughly 12, 11.7, 9, and 12% reductions in root mean square error compared to the DNN with 2, 3, 4, and 5 hidden layers, respectively. The Wilcoxon signed-rank tests confirm that the proposed model is highly suitable for concrete pile capacity prediction.