Pattern structures for analyzing complex data

Abstract

For data given by binary object-attribute datatables Formal Concept Analysis (FCA) provides with a means for both convenient computing hierarchies of object classes and dependencies between sets of attributes used for describing objects. In case of data more complex than binary to apply FCA techniques, one needs scaling (binarizing) data. Pattern structures propose a direct way of processing complex data such as strings, graphs, numerical intervals and other. As compared to scaling (binarization), this way is more efficient from the computational point of view and proposes much better vizualization of results. General definition of pattern structures and learning by means of them is given. Two particular cases, namely that of graph pattern structures and interval pattern structures are considered. Applications of these pattern structures in bioinformatics are discussed. © 2009 Springer-Verlag Berlin Heidelberg.