Online Classifiers Based on Fuzzy C-means Clustering

In the online approach a classifier is, as usual, induced from the available training set. However, in addition, there is also some adaptation mechanism providing for a classifier evolution after the classification task has been initiated and started. In this paper two algorithms for online learning and classification are considered. These algorithms work in rounds, where at each round a new instance is given and the algorithm makes a prediction. After the true class of the instance is revealed, the learning algorithm updates its internal hypothesis. Both algorithms are 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. The proposed approach is validated experimentally. Experiment results show that both proposed classifiers can be considered as a useful extension of the existing range of online classifiers.