MACHINE LEARNING MODELS TO GENERATE A SUBSURFACE SOIL PROFILE: A CASE OF MAKATI CITY, PHILIPPINES

Abstract

Soils and rocks are natural geomaterials, and a variety of spatially varying factors influence their properties during their complex geological formation phase. As a result, geomaterials can have different properties at different points on a given site. It is advantageous to be aware of the soil profile of a target location in advance to avoid duplication of tests, determining the borehole depth, and sampling methodology. This can result in more economical borehole testing. The goal of this study is to apply Machine Learning Modeling Competition to generate the soil profile, in the case of this study, Makati City, Philippines. The models competing include Tree Model, Discriminant Model, Naïve Bayes Model, k-Nearest neighbor, and Artificial Neural Network. Among the models, k-Nearest Neighbor Model resulted in the highest accuracy rate, for validation. As an additional output, the generated data was transformed into a soil profile delineation that was represented by soils that are grouped into various classes.