sunnymarkliu / naive_bayes_meet_adaboost Goto Github PK

View Code? Open in Web Editor NEW

160.0 14.0 67.0 427 KB

:email: Implement Naive Bayes and Adaboost from scratch and use them to filter spam emails.

Python 100.00%

naive-bayes-classifier adaboost

naive_bayes_meet_adaboost's People

Contributors

Stargazers

Watchers

Forkers

qianguos hpplinux jeep123 ritali hanwei2008 popgreen xsongx incidunt terryking1992 yishengxiaoyao thomasding2016 alexyandy2016 chunchun1116 dinghe xiaofengzhiyu zengxia l421072474 gaozhuolun ethanray zhucanxiang drfilter fengchangfight shmathrabbit samueltt amingzym yjian180 eycab biteni iamblinking spongebobcatcher niwangyuan heilce cocoonyh mars2018 dodf diamondgloves lurenyi233 xeq-machine-learning ecore2018 joseph-zxy embolismsoil gmwu toccator makeadifference sky4star happinesschen githubbeinner pukingli c-peng avb43 mqm501 wruochao19 ranglang niyingqiu shennaichuan150 nwpusewang anneyinsj juliansteam sfpeace hzjkaka jade1998 cxncu001 culacola edwinlzw aaakz0347 happycoooo

naive_bayes_meet_adaboost's Issues

关于Adaboost部分的疑问

您好，请问Adaboost部分的代码是否有相关的理论依据（比如论文）？感觉其和西瓜书上的Adaboost算法描述差别有些大。

预测时计算P(W|Ci)P(Ci)，P之间相乘还是相加？

请问：
p1 = sum(testWordsMarkedArray * pWordsSpamicity) + np.log(pSpam)
这个公式不太明白，不因该是(testWordsMarkedArray * pWordsSpamicity) 里不为零的值相乘得到P(W|C)吗？
已解决，训练时 pWordsSpamicity 取了对数。

理论推导疑问

您好：

所以最终比较的是，P(W1|S)P(W2|S)....P(Wn|S)P(S)和P(W1|H)P(W2|H)....P(Wn|H)P(H)的大小
可能是我理解错了，但是我觉得比较公式应该是

P(W1|S)P(W2|S)....P(Wn|S)*P(S)^n和P(W1|H)P(W2|H)....P(Wn|H)*P(H)^n的大小

testWordsMarkedArray

请问testWordsMarkedArray指的是什么呢？
没有python基础看得有些吃力……不好意思哈

DS计算部分: alpha是不是写错了？
alpha = ps - ph
if alpha < 0: # 原先为spam，预测成ham
DS[testWordsCount != 0] = np.abs((DS[testWordsCount != 0] - np.exp(alpha)) / DS[testWordsCount != 0])
else: # 原先为ham，预测成spam
DS[testWordsCount != 0] = (DS[testWordsCount != 0] + np.exp(alpha)) / DS[testWordsCount != 0]

改成
alpha = ps - ph
if alpha > 0: # 原先为ham，预测成spam
DS[testWordsCount != 0] = np.abs((DS[testWordsCount != 0] - np.exp(alpha)) / DS[testWordsCount != 0])
else: # 原先为spam，预测成ham
DS[testWordsCount != 0] = (DS[testWordsCount != 0] + np.exp(alpha)) / DS[testWordsCount != 0]
错误率更低了。

sunnymarkliu / naive_bayes_meet_adaboost Goto Github PK

naive_bayes_meet_adaboost's People

Contributors

Stargazers

Watchers

Forkers

naive_bayes_meet_adaboost's Issues

AdaBoost

关于Adaboost部分的疑问

数据集

预测时计算P(W|Ci)P(Ci)，P之间相乘还是相加？

理论推导疑问

testWordsMarkedArray

Adaboost方法疑问

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent