Fuzzy decision trees for dynamic data

Abstract

The fuzzy decision tree based approach is a very popular machine learning method that deals with imprecise and uncertain data. This approach offers a good way to handle static data. However, few works have been conducted on the use of this approach to deal with stream of data or temporal data when the training set is built incrementally time after time. To handle such kind of data brings out a number of problems for the algorithms used to construct such fuzzy decision trees. In this paper, a new approach is proposed to construct a fuzzy decision tree (FDT) when the training set is built incrementally and when training examples are provided temporally. A new measure of discrimination is defined in order to rank attributes during the process of construction of the FDT and to take into account aging of examples.