This is an anomaly detection solution that helps user to quickly identify anomaly in time series data.
In a typical time series analysis use case, this solution provides an easy way to quick analyze a subset of data with anomaly to quickly evaluation a couple of models during exploratory data analysis. This give user a sense of time series data quality and whether there is anomaly pattern in the dataset.
A simple user interface built on top of Streamlit to provide a quick glance on the analysis result.
- Univariate Inter Quantile Ratio
- Multivariate Vector Auto Regression
- Multivariate Isolation Forest
- Multivariate Mahalanobis Distance
This application has been tested with python 3.9.
git clone https://github.com/yapweiyih/simple-anomaly-detection
cd simple-anomaly-detection
# Activate virtual environment first and install python package, this may take a while, so take a cup of coffee.
pip install -e .
pip install -r requirements.txt
# streamlit run src/uc_timeseries/streamlit_app.py -- --data_dir <path/to/data/>
streamlit run src/uc_timeseries/streamlit_app.py -- --data_dir refdata
A demo data with some anomaly points has been included to get you familarized with this tool.
Select desired model to run training and evaluation.
Visualize time index with anomaly deteceted and feature importance for tree based model.
See CONTRIBUTING for more information.
This library is licensed under the MIT-0 License. See the LICENSE file.