Distributed collaborative knowledge elicitation

In this study, we develop an idea of knowledge elicitation realized over a collection of databases. The essence of such elicitation deals with a determination of common structure in databases. Depending upon a way in which databases are accessible abd can collaborate, we distinguish between a vertical and horizontal collaboration. In the first case, the databases deal with objects defined in the same attribute (feature) space. The horizontal collaboration takes place when dealing with the same objects but being defined in different attribute spaces and therefore forming separate databases. We develop a new clustering architecture supporting the mechanisms of collaboration. It is based on a standard FCM (Fuzzy C-Means) method. When it comes to the horizontal collaboration, the clustering algorithms interact by exchanging information about local partition matrices. In this sense, the required communication links are established at the level of information granules (more specifically, fuzzy sets forming the partition matrices) rather than patterns directly available in the databases. We discuss how this form of collaboration helps meet requirements of data confidentiality. In case of the horizontal collaboration, the method operates at the level of the prototypes formed for each individual database. Numeric examples are used to illustrate the method.