Predicción de la contaminación atmosférica generada por las emisiones del CO2 en el Perú utilizando los métodos ARIMA y Redes Neuronales

Abstract

Carbon Dioxide is the main greenhouse gas (GHG) that leads to global warming and consequently, climate and environmental change. It brings negative effects to economic development, human life and the environment. It is extremely important to be able to accurately measure and predict the emission of carbon dioxide, since this way we can carry out a good sustainable policy for our environment. The main objective of this research is to find the best prediction model of carbon dioxide (CO2) emissions in Peru, through the comparative evaluation of the ARIMA and Artificial Neural Networks methods. The annual data of the emissions of (CO2) from the World Bank were used, which were analyzed by programming the free software R studio. To determine the best model, forecast errors such as: square root of the mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were used. The results reveal that the most appropriate model between these two methods for predicting (CO2) emissions in Peru is the neural network ANN (5-10-1), that is, the neural network with five lagged values as input connected by ten nodes in the hidden layer and a single output layer, which had higher precision with RMSE = 1125.82, MAE = 1040.68 and MAPE = 1.90 in the test phase compared to the best ARIMA model (0,1,10) that had an RMSE = 4223.73, MAE = 3143.40 and MAPE = 5.80 in the test phase. In conclusion, neural networks can be used to predict emissions of (CO2), which clearly showed that annual emissions of (CO2) in Peru will increase in the coming years. These real insights will be useful for policy makers to bring significant changes in major environmental areas in our country