ECID lite: uma aplicação para previsão de explosões solares

Abstract

Solar flares are phenomena that happen on the Sun, which consist of a sudden release of a large amount of energy. This large release of solar radiation causes electromagnetic disturbances on the earth, affecting communications, interfering with radio frequencies and power grid transmissions. A forecasting method can detect in advance or mitigate such consequences. This project sought to improve the Ensemble of Classifiers for Imbalanced Datasets (ECID) method, making it more efficient and flexible. For this, a study of the theoretical method and its implementation was carried out, identifying points of improvement. The method was improved by proposing a new architecture, refactoring and documenting the code base. Testing with the new version resulted in a 22% reduction in execution time and less code complexity, measured from the cyclomatic complexity before and after the changes. In this work, an API was created that provides the results of the method, in addition to the development of a proof of concept for the use of the API developed in the form of the ECID lite web application, which allows, in a simple way, the analysis of a set of forecasts of the ECID