Feature Generation for Sequence Categorization

Daniel Kudenko; Haym Hirsh

Feature Generation for Sequence Categorization

Computer Science

Research output: Contribution to conference › Paper › peer-review

Abstract

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 of FGEN across a range of tasks selected from three domains: DNA sequences, Unix command sequences, and English text.

Original language	Undefined/Unknown
Pages	733-738
Publication status	Published - 1998

Access to Document

http://portal.acm.org/citation.cfm?id=295798

Cite this

@conference{19f14a630b4a46c8a0f7e85fd83331a7,

title = "Feature Generation for Sequence Categorization",

abstract = "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 of FGEN across a range of tasks selected from three domains: DNA sequences, Unix command sequences, and English text.",

author = "Daniel Kudenko and Haym Hirsh",

year = "1998",

language = "Undefined/Unknown",

pages = "733--738",

}

TY - CONF

T1 - Feature Generation for Sequence Categorization

AU - Kudenko, Daniel

AU - Hirsh, Haym

PY - 1998

Y1 - 1998

N2 - 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 of FGEN across a range of tasks selected from three domains: DNA sequences, Unix command sequences, and English text.

AB - 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 of FGEN across a range of tasks selected from three domains: DNA sequences, Unix command sequences, and English text.

M3 - Paper

SP - 733

EP - 738

ER -