Towards a predictor for CO2 emission using regression analysis and an artificial neural network Sarinya Sala-ngam

Abstract

The purpose of this research is to demonstrate an approach for predicting carbon dioxide (CO2) emissions in Thailand which caused form sector of industry, transportation, economic and power generation. This research focuses on applying regression analysis and an artificial neural network with the multi-layer feedforward networks trained using the back-propagation algorithm. The causes of CO2 emissions in transport and economic sector such as increasing volume of vehicles, petroleum products consumption, CO2 emission, Thailand population and GDP were investigated and collected in this study. In order to design the predictive model of CO2 emissions, the data obtained were analyzed using regression analysis with 95% confidence interval. Also, an artificial neural network with neuron architecture: 4-12-1 (input-hidden-output), transfer function: positive linear, learning rate: 0.02 and 1000 Epoch training with MALAB was used to achieve designing the accurately predicting model for CO2 emissions. The results demonstrated that an artificial neural network has the ability to predict CO2 emissions more efficiently and accurately than the regression analysis method with 6 times considered by the value of the mean absolute deviation (MAD). Therefore, an artificial neural network with the multi-layer, back-propagation and feedforward network is the appropriate approach for generating the predictive model of CO2 emission.