Year: 2025
Type: conference-paper
Abstract: To enhance the predictive accuracy of machine learning models for undrained shear strength, this study explores the feasibility of constructing ensemble models using Genetic Algorithms (GA). Initially, in response to the challenges of data scarcity and high data collection costs in the field of soil engineering, the Variational Autoencoder (VAE) method was employed for data augmentation. Subsequently, the weights of several different machine learning models (including Artificial Neural Networks, Support Vector Machines, Gaussian Process Regression, Random Forest, and XGBoost) were adjusted and optimized using Genetic Algorithms to develop an ensemble model. During the experimental process, a set of random weights was initialized and iteratively evolved, continuously evaluating and selecting the best-performing weight combinations. The selection, crossover, and mutation in each generation aimed to reduce prediction errors and enhance the overall performance of the model. The experimental results demonstrate that the ensemble modeling approach, optimized by Genetic Algorithms, significantly improved the accuracy of predictions for undrained shear strength. This study not only highlights the substantial advantages of Genetic Algorithms in the integration of machine learning models but also provides an innovative and effective solution for predicting undrained shear strength in the field of soil engineering, showcasing the powerful potential of Genetic Algorithms in addressing complex problems in soil engineering.
Author: Tengyun Zhang
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4414396270