从时变数据流计算熵与基尼指数的更新公式与算法
人工智能
2016-08-02 v6 机器学习
摘要
尽管对数据流挖掘的兴趣日益增长，但最成功的增量学习器（如 VFDT）仍使用周期性重计算来更新属性信息增益和基尼指数。本说明提供了简单的增量公式与算法，用于从时变数据流中计算熵和基尼指数。
引用
@article{arxiv.1403.6348,
title = {Updating Formulas and Algorithms for Computing Entropy and Gini Index from Time-Changing Data Streams},
author = {Blaz Sovdat},
journal= {arXiv preprint arXiv:1403.6348},
year = {2016}
}
备注
Added directions future work; more consistent notation; fixed the errors in the updating algorithms for entropy; fixed an error in the statement of theorem 5; added two references to related work; fixed a few typos