Soils impregnated with oil undergo different transformations depending on the amount and composition of the pollutant. This refers to changes in both the structure of the soil and the component composition of oil. Fluctuating ratios of oil-soil mixtures, especially long-term exposure to the environment, have a significant synergistic effect on these changes. These changes are important to consider when developing effective technologies for the rehabilitation of oil-contaminated soils. In the presented paper, the results of the extraction cleaning method of soil samples with a high content of oil fractions, which were exposed to natural influences for a long time in natural conditions, are discussed. The dependence of the extraction efficiency of oil-contaminated soils on the percentage of extractants, their various ratios, temperature, and the duration of mechanical treatment was studied. Processing of experimental data by static methods formed the basis of a mathematical model of the process based on the algorithm of extreme gradient boosting. The obtained model with optimal hyperparameters was used for forecasting the training data. The totality of the obtained experimental data allowed us to establish the most significant factors affecting the efficiency of oil-contaminated soil cleaning.