Daniel Kudenko, Haym Hirsh
The problem of sequence categorization is to generalize from a corpus of labeled sequences procedures for accurately labeling future unlabeled sequences. The choice of representation of sequences can have a major impact on this task, and in the absence of background knowledge a good representation is often not known and straightforward representations are often far from optimal. We propose a feature generation method (called FGEN) that creates Boolean features that check for the presence or absence of heuristically selected collections of subsequences. We show empirically that the representation computed by FGEN improves the accuracy of two commonly used learning systems (C4.5 and Ripper) when the new features are added to existing representations of sequence data. We show the superiority ofFGEN across a range of tasks selected from three domains: DNA sequences, Unix command sequences, and English text.