| 一种新的决策树归纳学习算法 |
| |
| 引用本文： | 洪家荣,丁明峰.一种新的决策树归纳学习算法[J].计算机学报,1995,18(6):470-474. |
| |
| 作者姓名： | 洪家荣 丁明峰 |
| |
| 作者单位： | 哈尔滨工业大学计算机科学与工程系 |
| |
| 摘 要： | 本文不示例学习的重要分枝－－决策树归纳学习进行了分析探讨，从示例学习最优化的角度分析了决策树归纳学习的优化原则，指出了以往的以ＩＤ３为代表的归纳学习算法所固有的缺陷，并提出了一种新的基于概率的决策树归纳学习算法ＰＩＤ，ＰＩＤ在扩展属性的选择上仍采用基于信息增益率的方法，但在树上的扩展过程中，采用属性聚类的方法进行树的支合并。ＰＩＤ得到的决策树在树的规模和分类精度上都优于ＩＤ３。
|
| 关 键 词： | 决策树 归纳学习 算法 知识表示 |
A NEW ALGORITHM OF DECISION TREE INDUCTION |
| |
| Hong Jiarong, Ding Mingfeng, Li Xingyuan and Wang Liwei.A NEW ALGORITHM OF DECISION TREE INDUCTION[J].Chinese Journal of Computers,1995,18(6):470-474. |
| |
| Authors: | Hong Jiarong Ding Mingfeng Li Xingyuan and Wang Liwei |
| |
| Abstract: | The algorithms for decision tree induction, are investigated, and their limitations are analyzed. A new probability-based algorithm of decision tree induction, PID, is presented. Similar to ID3, PID selects attributes according to information gain, but PID is able to merge branches by using probability-based clustering. The experimental results show that the recognition rate gained by PID is better than that by ID3. |
| |
| Keywords: | Learning from examples decision tree merge of branches |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
