Introducing fuzzy decision stumps in boosting through the notion of neighbourhood

This study presents a novel approach to the weak classifier selection based on the GentleBoost framework. The authors include explicitly the notion of neighbourhood in one of the most common weak learner in boosting, the decision stumps. The availability of neighbouring points adds a new parameter to the decision stump, the feature set (i.e. neighbourhood), and turns the single branch selection of the decision stump into a fuzzy decision that weights the contribution of each branch using a neighbourhood-based confidence measure. The confidence measure of the fuzzy stumps uses neighbouring samples to increase the robustness to local data perturbations. The appropriate definition of the neighbourhood in the data set allows the application of the fuzzy stumps framework in a wide range of problems. The authors address two types of scenarios to show their advantages: (i) time-based neighbourhoods and (ii) space-based neighbourhoods. In both scenarios the properties of the fuzzy stumps are evaluated experimentally, considering computer-generated data sets and real classification problems, such as human activity recognition and object detection.