Ensemble learning for predicting subsurface bearing layer depths in Tokyo

Abstract

In order to improve the accuracy of geotechnical investigations, this study developed an ensemble learning method for predicting the depth of the bearing layer in Tokyo. Due to the limitations of traditional geotechnical surveys and the need for detailed soil information, this study used machine learning to improve model performance. Different from traditional methods, this study used an integrated model composed of multiple learners for analysis. By analyzing 2422 geotechnical surveys in the Kanto area of Japan, an ensemble learning algorithm was used to predict the depth of the bearing layer. The results showed that the ensemble learning method had a higher model performance. In addition, to explore the accuracy of the data, the ensemble learning method was compared with the kriging method. The results showed that compared with traditional methods, the model performance of the integrated model is improved by about 20 %.highlighting the effectiveness of ensemble learning in dealing with spatial variability and limitations of geotechnical data. This study highlights the importance of integrating various datasets and ensemble learning algorithms to cope with complex geotechnical prediction challenges. Although there is still room for improvement in model performance, as the first study to predict the depth of the supporting layer, it may provide a deeper understanding of the underground conditions in cities such as Tokyo and pave the way for more sophisticated data-driven geotechnical engineering methods.

Keywords

• Artificial intelligence
• Bearing layer depth prediction
• Ensemble learning
• Geotechnical investigation
• Machine learning geotechnical engineering
• Spatial data analysis