Metaheuristic algorithms to forecast future carbon dioxide emissions of Turkey

Abstract

This paper proposes the use of five different metaheuristic algorithms for forecasting carbon dioxide emissions (MtCO2) in Turkey for the years between 2019 and 2030. Historical economic indicators and construction permits in square meters of Turkey between 2002 and 2018 are used as independent variables in the forecasting equations, which take the form of two multiple linear regression models: a linear and a quadratic model. The proposed metaheuristic algorithms, including Artificial Bee Colony (ABC), Genetic Algorithm (GA), Simulated Annealing (SA), as well as hybrid versions of ABC with SA and GA with SA, are used to determine the coefficients of these regression models with reduced statistical error. The forecasting performance of the proposed methods is compared using multiple statistical methods, and the results indicate that the hybrid version of ABC with SA outperforms other methods in terms of statistical error for the linear equation model, while the hybrid version of GA with SA performs better for the quadratic equation model. Finally, four different scenarios are generated to forecast the future carbon dioxide emissions of Turkey. These scenarios reveal that if construction permits and the population is strictly managed while the economical wealth of Turkey keeps on improving, the CO2 emissions of Turkey may be less than in other possible cases.