Timeseria is an object-oriented time series processing library which aims at making it easy to manipulate time series data and to build statistical and machine learning models on top of it.
Unlike common numerical and data analysis frameworks, Timeseria does not make use of low level data structures as arrays and matrices to represent time series data. Instead, it builds up from well defined and reusable logical units (objects), which can be easily combined together in order to ensure an high level of consistency.
Thanks to this approach, Timeseria can address by design several issues that are often left as implementation details but that are actually a great source of headaches - as handling data losses, non-uniform sampling rates, differences between aggregated data and punctual observations, timezones, DST changes, and so on.
Timeseria comes with a built-in set of common operations (resampling, aggregation, moving averages, derivatives etc.) as well as extensible models (for reconstruction, forecasting and anomaly detection), and integrates a powerful plotting engine capable of plotting even millions of data points.
You can get started by reading the quickstart or the welcome notebooks, or have a look at the other example notebooks provided in the Timeseria-notebooks repository.
Also the reference documentation might be useful.
You can install Timeseria by just using the the PyPI package:
pip install timeseria
Alternatively, a Timeseria Docker image with a Jupyter Notebook server and all the requirements is ready to be played with on Docker Hub:
docker run -it -p8888:8888 -v$PWD:/notebooks sarusso/timeseria
You can also clone this repo, install the requirements from the requirements.txt file and add it to your PYTHONPATH.
To work in development mode, you can either run a Jupyter notebook:
./jupyter.sh
or run the unit tests:
./test.sh
Both commands will start a Docker container and mount the local codebase inside it as a volume to allow for live code changes. They will trigger a container build so that if any requirement is changed, it will be reflected in the container as well.
If you don't want to automatically trigger a container build, prepend a BUILD=False:
BUILD=False ./test.sh
You can also run only specific tests:
BUILD=False ./test.sh timeseria.tests.test_datastructures
To instead set a specific log level when testing (default is CRITICAL):
TIMESERIA_LOGLEVEL=DEBUG ./test.sh
Every push on the Timeseria codebase as well as all the pull requests are automatically tested with GitHub Actions: check all branch statuses. Check the previous paragraph for how to run the unit tests when in development mode.
Timeseria is licensed under the Apache License version 2.0, unless otherwise specified. See LICENSE for the full license text.