Prediction of CO2 Emissions Using an Artificial Neural Network: The Case of the Sugar Industry

In this paper, the back-propagation artificial neural networks (ANN) model is presented to predict expenditure of carbon (CO2) emission. The model was built based on the input variables that affect to expenditure of CO2 include the amount of bagasse, wood and marine fuel oil used in boiler machine. The objective of this paper is to monitor the CO2 emission based on the fuel used for operating the boiler machine. The data used for testing the models were obtained from Sugar Industry. It splits up into 90% of training data and 10% of testing data. The model experiment was conducted using trial and error approach to find the optimal parameters of ANN model. The result shows that the architecture of ANN model have optimal parameter on training cycle 50, learning rate 0.1, momentum 0.1, and 19 hidden nodes. The validity of the trained ANN is evaluated by using Root Mean Square Error (RMSE) with error value as 0.055. It indicates that the smallest error provides more accurate results on prediction and even can contribute to the industrial practice, especially helping the executive manager to make an effective decision for business operation by considering the expenditure of CO2 monitoring.