A direct algorithm of possibilistic clustering with partial supervision

Fuzzy clustering plays an important role in intelligent systems design and the respective methods constitute a part of the areas of automation and robotics. This paper describes a modification of a direct algorithm of possibilistic clustering that takes into account the information coming from the labeled objects. The clustering method based on the concept of allotment among fuzzy clusters is the basis of the new algorithm. The paper provides the description of basic ideas of the method and the plan of the basic version of a direct possibilistic-clustering algorithm. A plan of modification of the direct possibilistic-clustering algorithm in the presence of information from labeled objects is proposed. An illustrative example of the method's application to the Sneath and Sokal's two-dimensional data in comparison with the Gaussian-clustering method is carried out. Preliminary conclusions are formulated.