ANN-based Prototype for the Prediction of CO2 Pollution Levels: Eco-Logica

Abstract

Eco-Logica is the development of a prototype based on artificial neural networks that allows the prediction and visualization of CO2 pollution levels. The lack of control of CO2 pollution levels produces impacts that negatively affect people's quality of life. For the development of the prototype, a research methodology with a quantitative approach is applied, which aims to analyze data associated with the pollution produced by carbon dioxide. The prototype, using regression techniques, makes predictions assuming pollution as a target variable, which is time-dependent. Additionally, the neural network model is trained using datasets consulted from national government databases, whose information is freely accessible and usable. The information processed by the network allows us to build reports that are rendered graphically in the prototype, and thus monitor the pollution levels. To assess the quality of the predictions, the coefficient of determination, known as R-squared (R2), is used, resulting in a value of 0.87117. From this, it can be concluded that the proposed model adequately describes the data's variability. Furthermore, cross-validation is performed using the standard deviation of R-squared, yielding a value of 0.0042, which is a positive indication that the model is not overfitting.