The dissertation was defended at the public meeting of the Dissertation Defence Council of Informatics Engineering in the SRA-I Hall of Vilnius Gediminas Technical University at 2 a.m. on 2 May 2024.
Oil spills on the ground cause significant damage to the geological environment, including groundwater, which becomes uninhabitable once contaminated. It is necessary to predict the scale of an oil spill and the consequences of contamination on the geological environment to minimise the damage of pollution. Specialists could use the prediction results to choose a strategy for its elimination. The literature provides many approaches for predicting an oil spill on the water. However, approaches are lacking for predicting oil spills in the geological environment. A detailed analysis of the prediction of oil spills on the geological environment and the water shows that scientists most often use machine learning algorithms and fuzzy logic to solve these problems. However, it is unclear what machine learning algorithms and fuzzy inference should apply to oil spill prediction in the geological environment. Moreover, in the real-world cases of oil spills in the geological environment, scientists and practitioners often face the challenge of a small dataset, which makes prediction difficult. The dissertation consists of an introduction, three main chapters, general conclusions, and a list of references. The first chapter provides a literature review and formulates the dissertation’s objectives. The second chapter proposes fuzzy inference and machine learning-based prediction with a small dataset of oil spills on a ground environment. It consists of two main parts: the first, where the fuzzy inference model for predicting oil spill contamination of the geological environment uses the fuzzification of two in-puts (the spilt oil product volume and the specific oil capacity), and defuzz-ification, applying a newly proposed procedure based on the area ratio of a fuzzy membership function, to predict whether an oil product will penetrate the ground layer; and the second, where the proposed approach of machine learning (Linear Regression, Decision Trees, SVR, Ensembles, and GPR) and fuzzy inference allow for the prediction of the consequences of oil spills into the groundwater using small datasets. The third chapter describes a two-part experiment with the proposed fuzzy inference model, machine learning algorithms, and an ANFIS-based model. The results of the experiment with the fuzzy inference model show that the proposed model is correct and does not contradict reality. The two calculated performance measures (MAE and RMSE) show that the proposed fuzzy inference model can predict the geological consequences of an oil spill with sufficient accuracy.
Doctoral dissertation readers can search via VILNIUS TECH Virtual Library.
