University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program


Data mining of heterogeneous data with an ART(Adaptive Resonance Theory)-based neural network


Dr. Elena Sapozhnikova, Universitšt Konstanz, Germany

date & place

Wednesday, 26.11.2008, 16:15 h
Room C 252


As a result of increased data complexity in science and industry, there is often a variety of interrelated data sources which all can be used to describe the same problem. Classical data mining utilizing just one data set at a time is insufficient in such a case. It is especially important in biological applications as, for example, in functional genomics where a single data source can often reveal only a certain perspective of the underlying complex biological mechanism. By integrating evidence from multiple data sources including e.g. gene expression data, phenotype data, genomic sequence information and so on, it is possible to obtain more accurate predictions of unknown gene functions. In contrast to existing approaches, the novel data mining system should accomplish Hierarchical Multi-Classification (HMC) that is a classification task with instances labeled by multiple classes and with the classes organized in a hierarchy. Additionally, the system should perform knowledge extraction from different data sources and build a common rule base. This will provide the user with important information that can be inferred by combining heterogeneous data. The knowledge extraction is also essential for testing the results by biologists in order to determine their plausibility.