Distance-Based Ensemble Online Classifier with Kernel Clustering

In this paper an on-line distance-based classifier is considered. The approach extends earlier proposed idea where a family of the online distance-based classifiers based on fuzzy C-means clustering followed by calculation of distances between cluster centroids and the incoming instance for which the class label is to be predicted, were suggested [8]. Now, instead of fuzzy C-means clustering we use kernel-based clustering method. The proposed algorithm works in rounds, where at each round a new instance is given and the algorithm makes a prediction. A portfolio of similarity or distance measures used to construct the ensemble of classifiers predicting the class of coming instances. The proposed approach is validated experimentally.