Desenvolvimento de um framework de redes neurais com interface gráfica

Abstract

Driven by advances in device hardware and equipment such as Internet of Things technologies, the amount of information generated and the interaction with computers have grown strongly in recent decades. As a result, interest in AM became increasingly greater. Therefore, the space for the development of tools that facilitate the application of artificial neural networks becomes increasingly larger. However, existing tools often do not have a pleasant and simple to use interface, especially for less experienced users. For that, this work aims to present the development of a graphical tool called framework for modeling the artificial neural network Multilayer Perceptron, developed in the programming language python. This software aims to provide the user with an intuitive experience, with easy extraction of information from compatible databases such as Microsoft Excel (.xlsx), Matlab (.mat), Text file (.txt) and Values Separated by Commas (.csv). The data once entered may be pre-treated using the normalization, seasonally adjusted and distributed coding methods. In this way, users with little experience in artificial neural networks can develop trained models and visualize the results graphically.