Curve
---
Curve is an open-source tool to help label anomalies on time-series data. The labeled data (also known as the ground truth) is necessary for evaluating time-series anomaly detection methods. Otherwise, one can not easily choose a detection method, or say method A is better than method B. The labeled data can also be used as the training set if one wants to develop supervised learning methods for detection.
Curve is designed to support plugin, so one can equip Curve with customized and powerful functions to help label effectively. For example, a plugin to identify anomalies which are similar to the one you labeled, so you don't have to search them through all the data.
Curve is originally developed by Baidu and Tsinghua NetMan Lab. [click for preview](http://curve.baidu.com/web/index.html)
<img src="https://raw.githubusercontent.com/baidu/Curve/master/doc/pic/index.png">
## Getting Started
### Run and stop
Simply use control.sh to start or stop Curve.
```bash
./control.sh start
./control.sh stop
```
Server will blind 8080 by default, you can change it in `./api/uwsgi.ini`.
> The first start will take a while because of the compilation.
> If you pull updates from github, Rebuild will be triggered during start or reload.
### Data format
You can load a CSV file into Curve. The CSV should have the following format
* First column is the timestamp
* Second column is the value
* Third column (optional) is the label. 0 for normal and 1 for abnormal.
The header of CSV is optinal, like `timestamp,value,label`.
Some examples of valid CSV
* With a header and the label column
|timestamp|value|label|
|---|---|---|
|1476460800|2566.35|0|
|1476460860|2704.65|0|
|1476460920|2700.05|0|
* Without the header
|1476460800|2566.35|0|
|---|---|---|
|1476460860|2704.65|0|
|1476460920|2700.05|0|
* Without the header and the label colum
|1476460800|2566.35|
|---|---|
|1476460860|2704.65|
|1476460920|2700.05|
* Timestamp in human-readable format
|20161015000000|2566.35|
|---|---|
|20161015000100|2704.65|
|20161015000200|2700.05|
## Additional
### Recommend environments
#### For PC
Darwin(Mac OSX) or Linux(Ubuntu, CentOS, Arch, etc.) is Recommended
* Dependency:
* Python 2.7.3+/3.1.2+, if python is not owned by current user, virtualenv is required
* Node.js 4.7.0+
* gcc, pip and npm path is correctly set
> Control Scripts for Windows is under development
#### For VPS like EC2
**Minimal**
* Server: 1 CPU, 512MB RAM, 5GB Storage
* System: Ubuntu10.04LTS or CentOS5.5
> Swap is required during build
**Recommend**
* Server: 1 CPU, 1GB RAM, 10GB Storage
* System: Ubuntu16.04LTS or CentOS7
### Backend Unit Test
```bash
cd api && pytest
```
### Plugin dir
```api/curve/v1/plugins```
### GitHub oauth
GitHub Oauth is supported, please put a configuration file into ```api/curve/auth/github_oauth.json``` like this:
```json
{
"id": "your github application Client ID",
"secret": "your application Client Secret"
}
```
> [Doc:Creating-An-Github-Oauth-App](https://developer.github.com/apps/building-oauth-apps/creating-an-oauth-app/)
没有合适的资源?快使用搜索试试~ 我知道了~
Python-Curve一个用于时间序列数据异常检测的综合实验平台

共128个文件
py:55个
js:32个
json:10个

需积分: 50 59 下载量 185 浏览量
2019-08-11
04:40:28
上传
评论 6
收藏 749KB ZIP 举报
温馨提示
Curve:时序数据异常标记工具。Curve是由百度和清华大学联合推出的一款开源工具,用于帮助开发者标记时序数据中的异常。标签数据(也就是真实有效值)对于评估时序数据异常检测方法非常有必要。否则,我们无法轻松选择好检测方法,或者确定模型A好于模型B。Curve能让开发者在上面使用强大的自定义函数,高效标记数据。
资源推荐
资源详情
资源评论





















收起资源包目录





































































































共 128 条
- 1
- 2
资源评论


weixin_39840924
- 粉丝: 496
上传资源 快速赚钱
我的内容管理 展开
我的资源 快来上传第一个资源
我的收益
登录查看自己的收益我的积分 登录查看自己的积分
我的C币 登录后查看C币余额
我的收藏
我的下载
下载帮助


最新资源
- 高阶逻辑定理证明:第15届国际会议论文集
- (源码)基于CC++编程语言的简易操作系统.zip
- (源码)基于意图识别的假肢控制系统.zip
- (源码)基于ARM CortexM处理器的迷宫游戏开发.zip
- (源码)基于编程语言的Smart Utility Vehicle.zip
- 基于 MAX78000 与 SSD 目标检测网络的猫咪识别喂食器:借助单片机 CNN 加速器实现神经网络计算
- (源码)基于Python和DGL的图计算实验框架MyPaGraph.zip
- 从零开始设计并训练神经网络,助你透彻理解它
- (源码)基于Python的JSON数据图形化展示系统.zip
- (源码)基于Arduino的传感器读取系统.zip
- 电气工程手册:计算机与数字设备精华
- (源码)基于Arduino框架的IoT环境监控系统.zip
- (源码)基于Python的模拟村庄发展项目-村庄模拟器.zip
- (源码)基于Keil C51编程语言的MCS52单片机打地鼠游戏.zip
- 基于基于常用 CNN 神经网络实现超 30 万条手写数学符号识别
- (源码)基于C++的太阳能飞机控制系统.zip
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈



安全验证
文档复制为VIP权益,开通VIP直接复制
