Adaptive-IDT: algoritmo incremental para aprendizagem de árvores de decisão adaptativas

Abstract

The induction of decision trees is nowadays one of the most used methods for solving problems of classification, decision making and machine learning (MITCHELL, 1997). Using this method you can create a decision tree based on a training data set and then sort new examples based on their structure. Classification generates an output, called a class value, that acts in response to input values. However, there are several situations in which training data is constantly changing, such as in the problems of incremental learning, in which learning process considers the dynamism of the training set. For problems such as this, the adaptability presents a very adherent solution, since it allows a structure to self-modify in response to external stimuli, incorporating new information. This paper proposes an algorithm for induction of decision trees, called Adaptive-IDT, based on the union of statistical techniques with adaptive technology. Through its specification and implementation, it is expected that it may present an alternative solution to traditional machine learning methods, as well as providing a foundation for future work that focuses on the use of adaptability applied to classification and learning methods Machine.