UFJF-MLTK: a framework for machine learning algorithms

Machine learning techniques have become increasingly common due to the extension of their application domains and because they can improve their performance when exposed to new data. Several methods have been proposed to address problems of the area, bringing the challenge of comparing different methods to find the one that best solves a problem. Frameworks and libraries focused on learning algorithms can reduce this effort. This paper describes the UFJF-MLTK, an object-oriented framework that helps to choose between different methods, in the development of new algorithms through the instantiation of a C++ class architecture that covers various types of learning algorithms and also helps in teaching the subject. We discuss the problems faced in the project architecture, the components of the framework, the algorithms that currently compose it, how it was documented and examples of its instantiation.