Aggregation of multidimensional object data in association analysis

The paper analyzes peculiarities of preprocessing learning data, with are represented in object data bases constituted by multiple relational tables with ontology on metalevel. On the other side, exactly such learning data structures are peculiar to many novel challenging applications. The paper proposes a new approach and heuristic algorithm intended for ontology–centered transformation of heterogeneous raw learning data into homogeneous feature space, that is based on rational aggregation of the sets of the ontology notion instances. Such transformation results in less dimensional and more informative feature space. Experimental results are provided for explanation and examplebased validation of the proposed approach and corresponding algorithm.