some understanding of《Improved Use of Continuous Attributes in C4.5》

Here are formulas provided in
“Improved Use of Continuous Attributes in C4.5”
1996,Journal of Artificial Intelligence Research 4 (1996)77-90

Info(D)=−∑j=1Cp(D,j)⋅log2(p(D,j))Info(D)=-\sum_{j=1}^{C}p(D,j)·log_2(p(D,j))Info(D)=−∑j=1Cp(D,j)⋅log2(p(D,j))

Gain(D,T)=Info(D)−∑i=1k∣Di∣∣D∣⋅Info(Di)Gain(D,T)=Info(D)-\sum_{i=1}^{k}\frac{|D_i|}{|D|}·Info(D_i)Gain(D,T)=Info(D)−∑i=1k∣D∣∣Di∣⋅Info(Di)

Split(D,T)=−∑i=1k∣Di∣∣D∣⋅log2(∣Di∣∣D∣)Split(D,T)=-\sum_{i=1}^{k}\frac{|D_i|}{|D|}·log_2(\frac{|D_i|}{|D|})Split(D,T)=−∑i=1k∣D∣∣Di∣⋅log2(∣D∣∣Di∣)

The followding are my understandings:
------------------first change-----------------------------
then,
Gain_Ratio=Gain(D,T)Split(D,T)Gain\_Ratio=\frac{Gain(D,T)}{Split(D,T)}Gain_Ratio=Split(D,T)Gain(D,T)

Then ,my understanding of the "first change"is
Gain_Ratio_adjusted=Gain(D,T)−log2(N−1)DSplit(D,T)Gain\_Ratio\_adjusted=\frac{Gain(D,T)-\frac{log_2(N-1)}{D}}{Split(D,T)}Gain_Ratio_adjusted=Split(D,T)Gain(D,T)−Dlog2(N−1)
is this right?
Many Thanks~
--------------------second change---------------------------
Relevant part of “second change” in this article is:
"This seems to be an unnecessary complication,so the threshold t is chosen instead to maximize gain.Once the threshold is chosen,however,the final selection of the attribute to be used for the test is still made on the basis of the gain ratio criterion using the adjusted gain
"
My understanding is:

1st step:
choose threshold t according to Gain(D,T)maxGain(D,T)_{max}Gain(D,T)max,
Not Gain_RatiomaxGain\_Ratio_{max}Gain_Ratiomax
Not (Gain(D,T)−log2(N−1)/∣D∣)max(Gain(D,T)-log_2(N-1)/|D|)_{max}(Gain(D,T)−log2(N−1)/∣D∣)max
2nd step:
the criterion to choose best feature is according to:
Gain_Ratio(discretefeature)=Gain(D,T)Split(D,T)Gain\_Ratio(discrete\ feature)=\frac{Gain(D,T)}{Split(D,T)}Gain_Ratio(discrete feature)=Split(D,T)Gain(D,T)
Gain_Ratio_adjusted(continuousfeature)=Gain(D,T)−log2(N−1)DSplit(D,T)Gain\_Ratio\_adjusted(continuous\ feature)=\frac{Gain(D,T)-\frac{log_2(N-1)}{D}}{Split(D,T)}Gain_Ratio_adjusted(continuous feature)=Split(D,T)Gain(D,T)−Dlog2(N−1)
Finally,just choose the feature whose Gain Ratio or Gain Ratio(adjusted) is the largest.

is this understanding right?
Many thanks~

some understanding of《Improved Use of Continuous Attributes in C4.5》相关推荐

《Improved Crowd Counting Method Based onScale-Adaptive Convolutional Neural Network》论文笔记
<Improved Crowd Counting Method Based onScale-Adaptive Convolutional Neural Network>论文笔记论文地址 ...
《Improved Techniques for Training GANs》-论文阅读笔记
<Improved Techniques for Training GANs>-论文阅读笔记文章目录 <Improved Techniques for Training GANs& ...
【论文学习笔记】《An Overview of Voice Conversion and Its Challenges》
<An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning&g ...
nlp论文——《Efficient Estimation of Word Representations in Vector Space》（向量空间中词表示的有效估计）
目录 <Efficient Estimation of Word Representations in Vector Space> 第一课时:论文导读 (1)语言模型 (2)词向量简介-- ...
（四十五：2021.08.05）《利用深度学习对ecg信号进行分割》
<Deep Learning for ECG Segmentation><利用深度学习对ecg信号进行分割> 讲在前面摘要 1. 介绍 2. 算法 2.1 预处理 2.2 神 ...
深度学习论文阅读图像分类篇（三）：VGGNet《Very Deep Convolutional Networks for Large-Scale Image Recognition》
深度学习论文阅读图像分类篇(三):VGGNet<Very Deep Convolutional Networks for Large-Scale Image Recognition> Ab ...
《用户至上：用户研究方法与实践》用户体验入门
本节书摘来自华章出版社<用户至上:用户研究方法与实践>一书中的第1章,第1节,作者凯茜·巴克斯特(Kathy Baxter)[美] 凯瑟琳·卡里(Catherine Courage)凯莉· ...
深度学习论文阅读目标检测篇（三）：Faster R-CNN《 Towards Real-Time Object Detection with Region Proposal Networks》
深度学习论文阅读目标检测篇(三):Faster R-CNN< Towards Real-Time Object Detection with Region Proposal Networks&g ...
《异常检测——从经典算法到深度学习》15 通过无监督和主动学习进行实用的白盒异常检测
<异常检测--从经典算法到深度学习> 0 概论 1 基于隔离森林的异常检测算法 2 基于LOF的异常检测算法 3 基于One-Class SVM的异常检测算法 4 基于高斯概率密度异常检测 ...

some understanding of《Improved Use of Continuous Attributes in C4.5》

some understanding of《Improved Use of Continuous Attributes in C4.5》相关推荐

最新文章

热门文章