On strategies for imbalanced text classification using SVM: A comparative study

• Published in: Decision Support Systems Vol. 48; no. 1; pp. 191 - 201
• Main Authors: Sun, Aixin, Lim, Ee-Peng, Liu, Ying
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01-12-2009; Elsevier; Elsevier Sequoia S.A
• Subjects: Accuracy; Applied sciences; Artificial intelligence; Classification; Comparative analysis; Computer science; control theory; systems; Data processing. List processing. Character string processing; Decision theory. Utility theory; Exact sciences and technology; Imbalanced text classification; Instance weighting; Memory organisation. Data processing; Operational research and scientific management; Operational research. Management science; Resampling; Software; Speech and sound recognition and synthesis. Linguistics; Studies; Support Vector Machines; SVM; Taxonomy; Text categorization; SVM; Resampling; Instance weighting; Support Vector Machines; Imbalanced text classification; Performance evaluation; Electronic discussion group; Statistical analysis; Threshold detection; Taxonomy; Unbalanced conditions; Information retrieval; Text; Interest; Vector support machine; Natural language; Recall; Content analysis
• ISSN: 0167-9236; 1873-5797
• DOI: 10.1016/j.dss.2009.07.011

Many real-world text classification tasks involve imbalanced training examples. The strategies proposed to address the imbalanced classification (e.g., resampling, instance weighting), however, have not been systematically evaluated in the text domain. In this paper, we conduct a comparative study on the effectiveness of these strategies in the context of imbalanced text classification using Support Vector Machines (SVM) classifier. SVM is the interest in this study for its good classification accuracy reported in many text classification tasks. We propose a taxonomy to organize all proposed strategies following the training and the test phases in text classification tasks. Based on the taxonomy, we survey the methods proposed to address the imbalanced classification. Among them, 10 commonly-used methods were evaluated in our experiments on three benchmark datasets, i.e., Reuters-21578, 20-Newsgroups, and WebKB. Using the area under the Precision–Recall Curve as the performance measure, our experimental results showed that the best decision surface was often learned by the standard SVM, not coupled with any of the proposed strategies. We believe such a negative finding will benefit both researchers and application developers in the area by focusing more on thresholding strategies.