Desenvolvimento de um framework para utilização da regressão linear múltipla com algoritmos de otimização e interface gráfica

Abstract

The technological advances of the last decades have brought great opportunities for society. The rapid development of new technologies has allowed common users to acquire equipment with large computational processing capabilities. Increasingly, these resources are being used as a way to perform tasks previously seen as long and difficult to execute, by means of computer programs that automate functions previously performed manually. Based on these premises, this work developed a framework that enables the application of the multiple linear regression (MLR), very common in several branches of literature. It allows the optimization of free parameters using optimization metaheuristics, namely, particle swarm optimization, genetic algorithm and differential evolution. A user-friendly interface was developed and at the same time the software has all the necessary functions for customization of the algorithms and analysis of the results. The programming of the software was done in Python language, which has the necessary resources both for the programming of the algorithms and for the creation of the interface layout. As an example of application, a real database was used to estimate the number of hospitalizations for respiratory diseases in the city of São Paulo. At the end of the work, the effectiveness of the algorithms is evaluated and their performance is compared.