Aim/purpose – The purpose of this paper is to introduce a new hybrid modelling to predict carbon dioxide emissions in order to make the correct decision to reduce air pollution in Iran. While there are not many data available for some variables, in this modelling, the goal is to make accurate predictions even with low data.
Design/methodology/approach – In the present paper, CO2 emissions in Iran in the period of 1980-2014 was predicted using three models of Auto-Regressive Distributed Lag (ARDL), Fuzzy Linear Regression (FLR) and hybrid model based on combination of ARDL and FLR models, and then the prediction accuracy of the models is compared.
Findings – Comparing prediction accuracy of models showed that the Fuzzy Auto- -Regressive Distributed Lag (FARDL) is more accurate than the initial patterns for predicting carbon dioxide emissions. Finally, the results showed that GDP and energy consumption has a positive and significant correlation with carbon dioxide emissions in short run. Also, the carbon dioxide emission indicated a low elasticity towards GDP and low energy consumption.
Research implications/limitations – When the number of data is low, the FARDL model provides a more accurate prediction than ARDL and FLR Models. FARDL’s combined model reduces the problems that exist in the ARDL and FLR models. One of the problems with the ARDL model is the need for many tests; the problem with the fuzzy regression model is also the high fuzzy distance length that makes decision making difficult. The FARDL model eliminates these constraints as much as possible.
Originality/value/contribution – This paper has been able to confirm that Fuzzy Auto- -Regressive Distributed Lag (FARDL) is more accurate than the initial patterns for predicting carbon dioxide emissions.
