https://github.com/Freemanzxp/GBDT_Simple_Tutorial
python实现GBDT的回归、二分类以及多分类，将算法流程详情进行展示解读并可视化，庖丁解牛地理解GBDT。Gradient Boosting Decision Trees regression, dichotomy and multi-classification are realized based on python, and the details of algorithm flow are displayed, interpreted and visualized to help readers better understand Gradient Boosting Decision Trees
https://github.com/Freemanzxp/GBDT_Simple_Tutorial
gbdt gnm gradient-boosting gradient-boosting-decision-trees machine-learning
Last synced: 6 months ago
JSON representation
python实现GBDT的回归、二分类以及多分类，将算法流程详情进行展示解读并可视化，庖丁解牛地理解GBDT。Gradient Boosting Decision Trees regression, dichotomy and multi-classification are realized based on python, and the details of algorithm flow are displayed, interpreted and visualized to help readers better understand Gradient Boosting Decision Trees
- Host: GitHub
- URL: https://github.com/Freemanzxp/GBDT_Simple_Tutorial
- Owner: Freemanzxp
- License: apache-2.0
- Created: 2019-04-04T15:41:00.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2019-06-15T06:33:32.000Z (almost 6 years ago)
- Last Synced: 2024-08-07T21:11:54.175Z (9 months ago)
- Topics: gbdt, gnm, gradient-boosting, gradient-boosting-decision-trees, machine-learning
- Language: Python
- Homepage:
- Size: 2.13 MB
- Stars: 714
- Watchers: 11
- Forks: 196
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# GBDT_Simple_Tutorial（梯度提升树简易教程）
## 简介
利用python实现GBDT算法的回归、二分类以及多分类，将算法流程详情进行展示解读并可视化，便于读者庖丁解牛地理解GBDT。
## 项目进度：
- [x] 回归
- [x] 二分类
- [x] 多分类
- [x] 可视化
***
**算法原理以及公式推导请前往blog：**[GBDT算法原理以及实例理解](https://blog.csdn.net/zpalyq110/article/details/79527653)
***
## 依赖环境
- 操作系统：Windows/Linux
- 编程语言：Python3
- Python库：pandas、PIL、pydotplus，
其中pydotplus库会自动调用Graphviz，所以需要去[Graphviz官网](https://graphviz.gitlab.io/_pages/Download/Download_windows.html)下载`graphviz的-2.38.msi`
，先安装，再将安装目录下的`bin`添加到系统环境变量，此时如果再报错可以重启计算机。详细过程不再描述，网上很多解答。
## 文件结构
- | - GBDT 主模块文件夹
- | --- gbdt.py 梯度提升算法主框架
- | --- decision_tree.py 单颗树生成，包括节点划分和叶子结点生成
- | --- loss_function.py 损失函数
- | --- tree_plot.py 树的可视化
- | - example.py 回归/二分类/多分类测试文件
## 运行指南
- 回归测试：
`python example.py --model = regression`
- 二分类测试：
`python example.py --model = binary_cf`
- 多分类测试：
`python example.py --model = multi_cf`
- 其他可配置参数：`lr` -- 学习率, `trees` -- 构建的决策树数量即迭代次数,
`depth` -- 决策树的深度, `count` -- 决策树节点分裂的最小数据数量,
`is_log` -- 是否打印树的生成过程, `is_plot` -- 是否可视化树的结构.
- 结果文件： 运行后会生成`results`文件夹,里面包含了每棵树的内部结构和生成日志
## 结果展示
仅展示最后所有树的集合，具体每棵树的详细信息望读者自行运行代码~
