信息论、熵、决策树

最新推荐文章于 2025-07-26 21:00:00 发布

原创最新推荐文章于 2025-07-26 21:00:00 发布 · 1.1k 阅读

0 ·

本内容遵循CC 4.0 BY-SA版权协议

python 同时被 3 个专栏收录

32 篇文章

订阅专栏

English

6 篇文章

订阅专栏

3 篇文章

订阅专栏

本文分享了作者在实际工作中使用决策树算法进行分类任务的经验，包括实现98.3%的准确率及在一分钟内完成一百万条数据的处理过程。文章还探讨了决策树的基本原理，并提及了几种常用的Python库。

http://www.borgelt.net/dtree.html

https://pypi.python.org/pypi/DecisionTree

http://scikit-learn.org/stable/modules/tree.html

http://orange.biolab.si/

For my job, I need to learn the Decision Tree for sths' classification. As we know, the AI, machine learning, especially the classification algorithms, comes it hey day now. Without these, it cost a lot of man-times for the classification jobs. It leads a new world that free not only the works but the professional works' hands.

I have experience of SVM, KNN and Naive Bayes. Now, i try the DT in another domain, and surprisingly, it did a great job. it make it 98.3% accuracy classifying the two category data with more than 40 features. And i randomly made one million lines of data, it make the tree in 700 secs. it did a great job really.

So I get involve with the basic theory of the DT. And I just find that those 4 libraries are basic on C4.5 and make an improvement. But never goes fast. I am seeking a more effective algorithms and let it be more useful.

Besides, in fact i am more likely to write python framework in the coming days. I think the weak AI algorithms nowadays comes its end, and it already successfully solves lots of problems.

I have a waste of my time last week, lots of distinguish communications with other companies. An I find that I should have a master career tomorrow.!~