Description and classification of complex structured objects by applying similarity measures

In this paper, we propose and investigate several similarity measures on complex structured objects. The objects are understood as examples of a target relation, and they are expressed in a first-order logic language. We also propose and experimentally verify an algorithm for description and classification of objects. The algorithm transforms data expressed in the first-order logic language into a decision table expressed in an attribute-value language. The table is constructed by applying a similarity measure as well as some notions of rough sets. Decision rules induced from such a table are treated as a description of objects. The rules can also be applied for classification of new unseen objects.