realseries

RealSeries is a comprehensive out-of-the-box Python toolkit for various tasks, including Anomaly Detection, Granger causality and Forecast with Uncertainty, of dealing with Time Series Datasets.

RealSeries has the following features:

• Unified APIs, detailed documentation, easy-to-follow examples and straightforward visualizations.
• All-levels of models, including simple thresholds, classification-based models, and deep (Bayesian) models.

RealSeries also considers the icing-on-the-cake functions, including:

• Bayesian Time series forecast/detect to get uncertainty.
• Model interpretations for the prediction/detection including Granger causality.

API Demo:

RealSeries uses the sklearn-style API and is as easy as

# train the SR-CNN detector
from realseries.models.sr_2 import SR_CNN
sr_cnn = SR_CNN(model_path)
sr_cnn.fit(train_data)
score = sr_cnn.detect(test_data, test_label)

Documentation

• Tutorials and API docs

Installation

RealSeries is still under development. Before the first stable release (1.0), you can install the RealSeries from source.

If you use RealSeries temporarily, you may clone the code and add it to your Python path:

git clone https://github.com/RealSeries/realseries.git
cd RealSeries # change current work directory to ./RealSeries
python
>>> import sys,os
>>> sys.path.append(os.getcwd()) # add ./RealSeries to sys.path
>>> import realseries
>>> from realseries.models.iforest import IForest

Alternatively, you can install it:

git clone https://github.com/RealSeries/realseries.git # clone
cd RealSeries
pip install .
python
>>> import realseries
>>> from realseries.models.iforest import IForest

:warning: RealSeries supports Python3 (=3.7.11) ONLY.

Dependencies

Realseries has several deep learning models like sr_cnn, lstm_encoder_decoder, which are implemented by pytorch. However, we do NOT install dependencies like pytorch automatically for you. If you want to use neural-net based models, see Installing Pytorch for installation Similarly, models depending on rrcf and luminol would NOT installed by default.

Required Dependencies:

• Python=3.7.11
• numpy>=1.13
• pandas>=0.25.3
• scikit-learn>=0.22
• scipy >=1.4.0
• pathlib >=1.0.1
• six >=1.13.0
• tqdm >=4.41.1
• setuptools==49.1
• statsmodels==0.10.2

Optional Dependencies:

:warning: We do NOT install all dependencies for you. Instead you should install dependencies by yourself for any models you want.

• pytorch (required for LSTM_AutoEncoder)
• rrcf>=0.3.2 (required for Random cut forest)
• luminol>=0.4 (required for luminol)
• more_itertools (required for LSTM_dynamic)
• mne (required for visualization examples)
• matplotlib (required for plotting examples)

Documentation

• Tutorials and API Docs
• Tutorials and API Docs Source Code (Need to build with Sphinx)
• Jupyter Notebooks
• Python Examples

Contribution and Contact

Most of the first version is done in RealAI and Tsinghua SAIL Group.

• So far, the core development members include
  • Wenbo Hu
  • Xianrui Zhang
  • Wenkai Li
  • Zhiheng Zhang
  • Peng Cui
  • Shuyu Cheng

If you have any questions, please leave issues.

Acknowkledgements

This project directly used many open-source libs:

• pytorch,
• luminol,
• sklearn,
• statsmodel.

Please leave an issue or send email to Wenbo Hu if your project wants to show in the list or does not want to be used.