A support vector regression model for the prediction of total polyaromatic hydrocarbons in soil: an artificial intelligent system for mapping environmental pollution

The significance of total polyaromatic hydrocarbons (TPAH) determination in assessing the carcinogenicity of environmental samples for measuring the level of environmental pollution cannot be overemphasized. Despite the environmental danger of TPAH, its laboratory quantification is laborious, which consumes appreciable time and other valuable resources. This research work develops a computational intelligence-based model for the first time, which directly estimates and quantifies the level of TPAH of any environmental solid samples using total petroleum hydrocarbons descriptor that can be easily determined experimentally. The hyperparameters of the developed support vector regression (SVR)-based model are optimized using manual search (MS) approach and genetic algorithm (GA) search approach with Gaussian and polynomial kernel functions. Experimental validation of the developed model was carried out using samples obtained from the marine sediments of Arabian Gulf Sea. The future generalization and predictive strength of the developed models were assessed using correlation coefficient (CC), root-mean-square error, mean absolute error and mean absolute percentage deviation (MAPD). GA-SVR-Gaussian performs better than MS-SVR and GA-SVR-poly with performance enhancement of 63.89% and 536.32%, respectively, on the basis of MAPD as a performance-measuring parameter, while MS-SVR model performs better than GA-SVR-poly with performance improvement of 288.25% using MAPD to evaluate the model performance. The estimation accuracy and generalization strength of the developed models indicate the potential of the models in measuring the level of environmental pollution of oil-spilled area without experimental stress, while experimental precision is preserved.