Soil Classification Modelling Using Machine Learning Methods

Classification of soil is a necessary aspect in geotechnical engineering purposes. Compared with traditional methods, smart and soft computing technology can classify different types of soil rapidly and effectively with high precise. A database of soil properties is collected and prepared based on the earlier researches and used for training and testing the machine learning classifier algorithms including Naïve Bayes and artificial neural network. The input detectable variables consist ۱۰۴ samples of soil mechanics features including cohesion, internal friction angle, and physical parameters such as water and dry density and used to design the Naïve Bayes and ANN models. The results of classification were considered for different soil typed such as clayey fine and coarse sandy components (GC, SC, GPGM, CL-ML, SC-SM, SM, CL). The developed network indicated that it can be considered as classifier network for soil classification. The results showed that only ۶ samples were not correctly identified among the total testing data (۳۴ samples) in Naïve Bayes model and ۷ samples were not correctly identifiedamong the total testing data (۳۴ samples) in artificial neural network model. Therefore, these networks can be used to enhance the accuracy and reduce the cost of projects