##自动构建评分卡
## 思想碰撞
> 仓库地址:https://github.com/ZhengRyan/autobmt > > 微信公众号文章:https://mp.weixin.qq.com/s/u8Nsp5M93WIGL2M0tU4U_g > > pipy包:https://pypi.org/project/autobmt/ > > 实验数据:链接: https://pan.baidu.com/s/1BRIHH9Wcwy2EZaO5xSgH9w?pwd=tdq5 提取码: tdq5
## 一、环境准备 可以不用单独创建虚拟环境,都是日常常用的python依赖包。需要创建虚拟环境,请参考”五、依赖包安装”
### autobmt 安装 pip install(pip安装)
`bash pip install autobmt # to install pip install -U autobmt # to upgrade `
Source code install(源码安装)
`bash python setup.py install `
## 二、使用教程 1、1行代码自动构建评分卡:请查看autobmt/examples/autobmt_lr_tutorial_code.py。里面有例子
2、1步1步拆解自动构建评分卡的步骤:请查看autobmt/examples/tutorial_code.ipynb。里面有详细步骤拆解例子
## 三、训练、自动选变量、自动单调最优分箱、自动构建模型、自动构建评分卡 1、Step 1: EDA,整体数据探索性数据分析
2、Step 2: 特征粗筛选
3、Step 3: 对粗筛选后的变量调用最优分箱
4、Step 4: 对最优分箱后的变量进行woe转换
5、Step 5: 对woe转换后的变量进行stepwise
6、Step 6: 用逻辑回归构建模型
7、Step 7: 构建评分卡
8、Step 8: 持久化模型,分箱点,woe值,评分卡结构
9、Step 9: 持久化建模中间结果到excel,方便复盘
## 四、保存的建模结果相关文件说明 1、all_data_eda.xlsx:整体数据的EDA情况
2、build_model_log_var_jpg文件夹,最终入模变量的分箱画图,在”build_model_log.xlsx”最后1个sheet也有记录
3、build_model_log.xlsx:构建整个模型的过程日志,记录有利复盘
4、fb.pkl、woetf.pkl、lrmodel.pkl、in_model_var.pkl:fb.pkl分箱文件,woetf.pkl转woe文件,lrmodel.pkl模型文件,入模变量文件
5、scorecard.pkl、scorecard.csv、scorecard.json:评分卡的pkl、csv、json格式。在”build_model_log.xlsx”的”scorecard_structure”sheet也有记录
6、var_bin_woe_format.csv、var_bin_woe_format.json、var_bin_woe.csv、var_bin_woe.json、var_split_point_format.csv、var_split_point_format.json、var_split_point.csv、var_split_point.json:分箱文件和转woe文件的csv、json格式
7、lr_auc_ks_psi.csv:模型的auc、ks、psi
8、lr_pred_to_report_data.csv:构建建模报告的数据
9、lr_test_input.csv:用于模型上线后,将次数据喂入模型,对比和lr_pred_to_report_data.csv结果是否一致。验证模型上线的正确性
## 五、依赖包安装(建议先创建虚拟环境,不创建虚拟环境也行,创建虚拟环境是为了不和其它项目有依赖包的冲突,不创建虚拟环境的话在基础python环境执行pip install即可) ####创建虚拟环境 conda create -y –force -n autobmt python=3.7.2 ####激活虚拟环境 conda activate autobmt
### 依赖包安装方式一,执行如下命令安装依赖的包 pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/
Metadata
Release files for autobmt 0.1.5
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| autobmt-0.1.5.tar.gz | 449.5 kB | Without provenance | Details for autobmt-0.1.5.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| autobmt-0.1.5-py2.py3-none-any.whl | Python 3, Python 2 | none | any | Details for autobmt-0.1.5-py2.py3-none-any.whl |
Total release size: 524.7 kB
Release files / autobmt-0.1.5.tar.gz
| Download URL | autobmt-0.1.5.tar.gz |
|---|---|
| Size | 449.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4f5132509997df7e4f5744c5ad60423f0329d16f40115e20e77a901cd539a04e
|
|
BLAKE2b-256 checksum How to use checksums |
93276b73ee3f71d814b55e0639ad9ca729fd5b3e78a0cd312ce59d58fe054347
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.7.2
|
Release files / autobmt-0.1.5-py2.py3-none-any.whl
| Download URL | autobmt-0.1.5-py2.py3-none-any.whl |
|---|---|
| Size | 75.2 kB |
| Tags | Python 2 Python 3 |
|
SHA-256 checksum How to use checksums |
419281f828cb4271af8556e9085fbaea1d49a117d04c213b3d5c0b84626c7130
|
|
BLAKE2b-256 checksum How to use checksums |
54b36dfa306cc459fb67c09aad9f52d5baa425538e6b41123a311164d731853a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.7.2
|
Security for autobmt 0.1.5
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.