Asian Basic and Applied Research Journal

A high concentration of particulate matter (PM) pollutants is a menace to the environment and proactive steps are required to minimize the concentrations of PM in our environment. This work was carried out to develop mathematical model for prediction of PM pollutants concentrations in Sarajevo city, Bosnia – Hergovina. Linear regression model (LRM), nonlinear polynomial regression model (NPRM) and nonlinear exponential regression model (NERM) were developed using the data generated in Sarajevo city with the aid of an in-built solver tool in Minitab Software version 2017. The developed models were subjected to error evaluation functions analysis in order to determine how they deviated from the experimental data. The error evaluation functions used were average relative error (ARE), sum of error square (ERRSQ), Marquardt’s percent standard deviation (MPSD), hybrid fractional error function (HYBRID), and root means square error (RMSE). Others include sum of absolute errors (EABS), standard error of estimate (SEE) and mean relative percentage error (MRPE). The applicability of the developed models was also investigated using statistical tools which include Pearson product moment correlation coefficient (r), coefficient of determination (R²), coefficient of non-determination (K²), student’s t – test(t - test), equality of variance test (F - test) and chi – square test (x²). The results revealed that based on the error evaluation functions analysis, NERM has the least deviation from experimental data when compared with LRM and NPRM. The statistical analysis also showed that NERM represented the experimental data better when compared with LRM and NPRM hence LRM and NPRM were jettisoned and NERM was adopted for the navigation of the experimental data. It was concluded that the NERM should be used to predict the PM pollutants concentrations in Sarajevo city.