 Algorithm based on modified angle‐based outlier factor for open‐set classification of text documentsTomasz Walkowiak, Szymon Datko and Henryk Maciejewski Applied Stochastic Models in Business and Industry, 2018, vol. 34, issue 5, 718-729 Abstract: This paper presents a new method of open‐set classification of text documents, with respect to subject areas. Standard (closed‐set) approaches to text classification involve training classifiers on annotated text corpora, representing a fixed number of subject areas. Such classifiers assign a new document with unknown annotation to one of the trained classes, even if the new document is not related to any class. We propose a two‐step procedure for open‐set classification. We first use a closed‐set classifier to assign a new document to one of the known classes. Then, we evaluate the (dis)similarity between the document and the chosen class using a novel criterion of outlierness named interquartile ranged angle‐based outlierness factors, which we find effective in high‐dimensional data. Based on this, we can avoid spurious assignment of documents to unrelated subject classes. We demonstrate the feasibility of this procedure in the task of subject classification of a collection of Wikipedia documents. As compared to the standard closed‐set approach, our open‐set classifier realizes significantly better precision with only small decrease of the recall measure observed in recognition of the tested classes. Date: 2018 Downloads: (external link) Related works: Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text Persistent link: https://EconPapers.repec.org/RePEc:wly:apsmbi:v:34:y:2018:i:5:p:718-729 Access Statistics for this article More articles in Applied Stochastic Models in Business and Industry from John Wiley & Sons |
Is your work missing from RePEc? Here is how to contribute.
Questions or problems? Check the EconPapers FAQ or send mail to .
EconPapers is hosted by the
School of Business at Örebro University.