节点文献
特征权对贝叶斯分类器文本分类性能的影响
Influence of feature weight on text categorization performance of Bayesian classifier
【摘要】 在文本分类研究中,人们希望用特征权来改善文本分类效果。以最优分类器——贝叶斯分类器为基准分类器,研究了特征权对文本分类性能的可能影响。理论推导表明,就最优分类器而言,特征权不能有效提高文本分类效果。
【Abstract】 In the field of text categorization, researchers tend to use feature weights to promote the performance of text classifiers. Taking the optimal classifier-Bayesian classifier, as the benchmark, theoretical analysis was performed about the possible influence of feature weight on text categorization performance. Theoretical deduction proves that feature weight can not effectively improve the performance of text categorization if the text classifier is a Bayesian one.
【关键词】 文本分类;
文本表示;
特征权;
贝叶斯分类器;
分类器性能;
【Key words】 text categorization; text representation; feature weight; Bayesian classifier; classifier’s performance;
【Key words】 text categorization; text representation; feature weight; Bayesian classifier; classifier’s performance;
【基金】 国家自然科学基金资助项目(60473039);江苏省高校自然科学研究指导性项目(04KJD520037)
- 【文献出处】 计算机应用 ,Journal of Computer Applications , 编辑部邮箱 ,2008年12期
- 【分类号】TP391.1
- 【被引频次】4
- 【下载频次】233