An Approach to Online Fuzzy Clustering Based on the Mahalanobis Distance Measure

The manuscript gives consideration to the problem of fuzzy clustering data streams. The offered approach incorporates the concepts of the probabilistic fuzzy clustering based on the specific sort of distance metrics. The main emphasis of the study was put on the application of Mahalanobis measures in the fuzzy clustering algorithms that let design classes of a hyperellipsoidal shape which can change the orientation of their axes in a feature space. The substantial hallmark of the presented fuzzy clustering scheme is its aptitude to group data in a sequential style on the assumption of the fact that groups have an arbitrary shape (which cannot typically be classified linearly) and to be mutually intersecting.