Incremental tuning of fuzzy decision trees

Handling stream data or temporal data is a difficult task and brings out a lot of problems to classical learning algorithms as the decision tree construction algorithms. In that context, incremental algorithms have been proposed but they often lie on the frequent reconstruction of the decision tree when this one provides a high number of misclassified examples. In this paper, we proposed a new algorithm to incrementally tune a fuzzy decision tree (FDT) that limit the number of reconstructions of the tree. That algorithm takes benefit of the fuzzy classification provided by a FDT to introduce a local tuning of the internal nodes of the FDT and avoid a complete reconstruction of the tree.