Continuous Dynamic Update of Fuzzy Random Forests

Fuzzy random forests are well-known machine learning classification mechanisms based on a collection of fuzzy decision trees. An advantage of using fuzzy rules is the possibility to manage uncertainty and to work with linguistic scales. Fuzzy random forests achieve a good classification performance in many problems, but their quality decreases when they face a classification problem with imbalanced data between classes. In some applications, e.g., in medical diagnosis, the classifier is used continuously to classify new instances. In that case, it is possible to collect new examples during the use of the classifier, which can later be taken into account to improve the set of fuzzy rules. In this work, we propose a new iterative method to update the set of trees in the fuzzy random forest by considering trees generated from small sets of new examples. Experiments have been done with a dataset of diabetic patients to predict the risk of developing diabetic retinopathy, and with a dataset about occupancy of an office room. With the proposed method, it has been possible to improve the results obtained when using only standard fuzzy random forests.