An intelligent multivariate approach for optimum forecasting of daily ozone concentration in large metropolitans with incomplete inputs

Previous studies show that it is quite necessary to accurately analyse and forecast ozone level especially in complex and large urban regions with incomplete inputs. Also, there is a need for more precise and efficient models to determine effective warning policies with respect to ozone concentration level in large cities. This study presents a flexible and adaptive approach to overcome the above issues. Moreover, an adaptive approach based on artificial neural network (ANN), adaptive neuro-fuzzy interference system (ANFIS) and conventional regression for forecasting of daily ozone levels is developed and discussed. The preferred model is selected via mean absolute percentage of error (MAPE). The proposed model is applied to one of the most polluted and populated cities in the world. Five pollutants and four meteorological variables are considered as inputs and ozone level is considered as output. The results show the flexibility of the proposed approach. The superiority and applicability of the proposed approach over previous models are also shown and discussed in this paper.