首页> 外文会议>International Conference on Artificial Intelligence and Pattern Recognition >A Study on Evaluation of Statistical categorization methods
【24h】

A Study on Evaluation of Statistical categorization methods

机译:统计分类方法评估研究

获取原文

摘要

Usually, macro average precision, macro average recall and macro f_1 are used to evaluate categorization technique. However, these measures are not perfect for evaluation. Experiments have shown that these measures are unstable on different datasets, which brings many issues for research. A new evaluation method of statistical categorization algorithm called new-macro-f_1 is proposed here according to the analysis of how the training dataset affects the classification result. The new measure remains stable on different datasets and given the performance of an algorithm on one collection, the performance e.g. precision on other collections could be estimated with the help of the new measure.
机译:通常,宏观平均精度,宏观平均召回和宏F_1用于评估分类技术。但是,这些措施并不完美评估。实验表明,这些措施在不同的数据集上不稳定,这为研究带来了许多问题。根据训练数据集如何影响分类结果的分析,提出了一种新的统计分类算法的新评估方法。新措施在不同的数据集中保持稳定,并在一个集合上鉴于算法的性能,性能为例如可以在新措施的帮助下估算其他收集的精度。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号