| Summary: | This study explored the first direct and inverse modelling of Carbon monoxide (CO) concentration by applying three different computational approaches namely, Least Square-boost (L-Boost), Hammerstein Weiner (HW), and Multi-variate regression (MVR) models for modelling CO using Sulphur dioxide (SO2), Nitrogen dioxide (NO2) and Ozone (O3). Two filters feature extraction methods were used in the input-combinations selection, which was classified into the direct modelling approach (C1) and inverse modelling approach (C2). Four different statistical metrics, including Nash-Sutcliffe efficiency, Person correlation coefficient, mean absolute error and mean absolute percentage error was used in evaluating the performance of the models. Based on the performance skills of the models, it is indicated that the direct modelling approach showed superior performance to the inverse approach. Furthermore, the non-linear models (HW and L-Boost) depict higher performance skills than the linear MVR model. Overall, the L-Boost-C1 models showed higher performance based on statistical metrics.
|
|---|