Feature Extraction for Machine Learning: Logic-Probabilistic Approach

Vladimir Gorodetsky and Vladimir Samoylov; JMLR W&P 10:55-65, 2010.

Abstract

The paper analyzes peculiarities of preprocessing of learning data represented in object data bases constituted by multiple relational tables with ontology on top of it. Exactly such learning data structures are peculiar to many novel challenging applications. The paper proposes a new technology supported by a number of novel algorithms intended for ontology-centered transformation of heterogeneous possibly poor structured learning data into homogeneous informative binary feature space based on (1) aggregation of the ontology notion instances and their attribute domains and subsequent probabilistic cause-consequence analysis aimed at extraction more informative features. The proposed technology is fully implemented and validated on several case studies.