Modeling the Geotechnical Properties of Small Soil Samples of Oil Polluted Soil: The Power of Supervised Machine Learning Models

Abstract

ABSTRACT Increasing levels of oil spills in both the terrestrial and aquatic environments destroy aquatic life and crops, fill the air with dangerous fumes, and change the physical and chemical features of the soil. Numerous authors have researched oil spills, but the application of supervised machine-learning models has been incredibly under-explored. Therefore, this study compared four supervised machine learning algorithms (Naïve Bayes Classifier, Decision Tree, Random Forest, and Support Vector Machine) to model the geotechnical properties of oil-polluted soil samples. The soil samples were collected from Engenni Ahoada in Rivers State and Izombe, Imo State, Nigeria. The samples were tested for the following parameters according to the ASTM standard method of measurement: shear strength (SS), liquid limit (LL), plastic limit (PL), bulk density (BD), maximum dry density (MDD), optimum moisture content (OMC), California bearing ratio (CBR), and plasticity index (PI). These measurements and the kind of soil (clay or laterite) constituted the inputs for the machine learning algorithms for predicting pollutant type (crude or diesel). The data from the thirty-two (32) sample size was split into train (70%) and test (30%) sets. The models were trained using the train set, and the test sets were used to validate the models. The support vector machine outperformed the competition, with prediction accuracy in the train and test sets of 96% and 89%, respectively. The findings demonstrated that the plastic and liquid limits are the most crucial factors for determining the pollutants in a sample of polluted soil. Also, plastic limit values less than or equal to 22 are likely contaminated by crude oil, while those above 22 are contaminated by diesel. The results therefore recommend using a support vector machine learning model for predicting the pollutant type of polluted soil samples in the study area.