# <p align="center"> <a href="https://predict-idlab.github.io/tsflex"><img alt="tsflex" src="https://raw.githubusercontent.com/predict-idlab/tsflex/main/docs/_static/logo.png" width="66%"></a></p>
[](https://pypi.org/project/tsflex/)
[](https://anaconda.org/conda-forge/tsflex)
[](https://img.shields.io/pypi/pyversions/tsflex)
[](https://codecov.io/gh/predict-idlab/tsflex)
[](https://github.com/predict-idlab/tsflex/actions/workflows/codeql.yml)
[](https://pepy.tech/project/tsflex)
[](http://makeapullrequest.com)
[](https://github.com/predict-idlab/tsflex/actions/workflows/deploy-docs.yml)
[](https://github.com/predict-idlab/tsflex/actions/workflows/test.yml)
[](https://discord.gg/k2d59GrxPX)
<!--  -->
> _tsflex_ is a toolkit for _**flex**ible **t**ime **s**eries_ [processing](https://predict-idlab.github.io/tsflex/processing) & [feature extraction](https://predict-idlab.github.io/tsflex/features), that is efficient and makes few assumptions about sequence data.
#### Useful links
- [Paper](https://www.sciencedirect.com/science/article/pii/S2352711021001904)
- [Documentation](https://predict-idlab.github.io/tsflex/)
- [Example (machine learning) notebooks](https://github.com/predict-idlab/tsflex/tree/main/examples)
#### Installation
| | command |
| :--------------------------------------------------- | :------------------------------------ |
| [**pip**](https://pypi.org/project/tsflex/) | `pip install tsflex` |
| [**conda**](https://anaconda.org/conda-forge/tsflex) | `conda install -c conda-forge tsflex` |
## Usage
_tsflex_ is built to be intuitive, so we encourage you to copy-paste this code and toy with some parameters!
### <a href="https://predict-idlab.github.io/tsflex/features/#getting-started">Feature extraction</a>
```python
import pandas as pd; import numpy as np; import scipy.stats as ss
from tsflex.features import MultipleFeatureDescriptors, FeatureCollection
from tsflex.utils.data import load_empatica_data
# 1. Load sequence-indexed data (in this case a time-index)
df_tmp, df_acc, df_ibi = load_empatica_data(['tmp', 'acc', 'ibi'])
# 2. Construct your feature extraction configuration
fc = FeatureCollection(
MultipleFeatureDescriptors(
functions=[np.min, np.mean, np.std, ss.skew, ss.kurtosis],
series_names=["TMP", "ACC_x", "ACC_y", "IBI"],
windows=["15min", "30min"],
strides="15min",
)
)
# 3. Extract features
fc.calculate(data=[df_tmp, df_acc, df_ibi], approve_sparsity=True)
```
Note that the feature extraction is performed on multivariate data with varying sample rates.
| signal | columns | sample rate |
|:-------|:-------|------------------:|
| df_tmp | ["TMP"]| 4Hz |
| df_acc | ["ACC_x", "ACC_y", "ACC_z" ]| 32Hz |
| df_ibi | ["IBI"]| irregularly sampled |
### <a href="https://predict-idlab.github.io/tsflex/processing/index.html#getting-started">Processing</a>
[Working example in our docs](https://predict-idlab.github.io/tsflex/processing/index.html#working-example)
## Why tsflex? ✨
- `Flexible`:
- handles multivariate/multimodal time series
- versatile function support
=> **integrates** with many packages for:
- processing (e.g., [scipy.signal](https://docs.scipy.org/doc/scipy/reference/tutorial/signal.html), [statsmodels.tsa](https://www.statsmodels.org/stable/tsa.html#time-series-filters))
- feature extraction (e.g., [numpy](https://numpy.org/doc/stable/reference/routines.html), [scipy.stats](https://docs.scipy.org/doc/scipy/reference/tutorial/stats.html), [antropy](https://raphaelvallat.com/antropy/build/html/api.html), [nolds](https://cschoel.github.io/nolds/nolds.html#algorithms), [seglearn](https://dmbee.github.io/seglearn/feature_functions.html)¹, [tsfresh](https://tsfresh.readthedocs.io/en/latest/text/list_of_features.html)¹, [tsfel](https://tsfel.readthedocs.io/en/latest/descriptions/feature_list.html)¹)
- feature extraction handles **multiple strides & window sizes**
- `Efficient`:<br>
- view-based operations for processing & feature extraction => extremely **low memory peak** & **fast execution time**<br>
- see: [feature extraction benchmark visualization](https://predict-idlab.github.io/tsflex/#benchmark)
- `Intuitive`:<br>
- maintains the sequence-index of the data
- feature extraction constructs interpretable output column names
- intuitive API
- `Few assumptions` about the sequence data:
- no assumptions about sampling rate
- able to deal with multivariate asynchronous data<br>i.e. data with small time-offsets between the modalities
- `Advanced functionalities`:
- apply [FeatureCollection.**reduce**](https://predict-idlab.github.io/tsflex/features/index.html#tsflex.features.FeatureCollection.reduce) after feature selection for faster inference
- use **function execution time logging** to discover processing and feature extraction bottlenecks
- embedded [SeriesPipeline](http://predict-idlab.github.io/tsflex/processing/#tsflex.processing.SeriesPipeline.serialize) & [FeatureCollection](https://predict-idlab.github.io/tsflex/features/index.html#tsflex.features.FeatureCollection.serialize) **serialization**
- time series [**chunking**](https://predict-idlab.github.io/tsflex/chunking/index.html)
¹ These integrations are shown in [integration-example notebooks](https://github.com/predict-idlab/tsflex/tree/main/examples).
## Future work 🔨
- scikit-learn integration for both processing and feature extraction<br>
**note**: is actively developed upon [sklearn integration](https://github.com/predict-idlab/tsflex/tree/sklearn_integration) branch.
- Support time series segmentation (exposing under the hood strided-rolling functionality) - [see this issue](https://github.com/predict-idlab/tsflex/issues/15)
- Support for multi-indexed dataframes
=> Also see the [enhancement issues](https://github.com/predict-idlab/tsflex/issues?q=is%3Aissue+is%3Aopen+label%3Aenhancement+)
## Contributing 👪
We are thrilled to see your contributions to further enhance `tsflex`.<br>
See [this guide](CONTRIBUTING.md) for more instructions on how to contribute.
## Referencing our package
If you use `tsflex` in a scientific publication, we would highly appreciate citing us as:
```bibtex
@article{vanderdonckt2021tsflex,
author = {Van Der Donckt, Jonas and Van Der Donckt, Jeroen and Deprost, Emiel and Van Hoecke, Sofie},
title = {tsflex: flexible time series processing \& feature extraction},
journal = {SoftwareX},
year = {2021},
url = {https://github.com/predict-idlab/tsflex},
publisher={Elsevier}
}
```
Link to the paper: https://www.sciencedirect.com/science/article/pii/S2352711021001904
---
<p align="center">
👤 <i>Jonas Van Der Donckt, Jeroen Van Der Donckt, Emiel Deprost</i>
</p>
Raw data
{
"_id": null,
"home_page": "https://github.com/predict-idlab/tsflex",
"name": "tsflex",
"maintainer": null,
"docs_url": null,
"requires_python": "<3.13,>=3.8",
"maintainer_email": null,
"keywords": "time-series, processing, feature-extraction, data-science, machine learning",
"author": "Jonas Van Der Donckt, Jeroen Van Der Donckt, Emiel Deprost",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/a0/fa/67d34578971b50b2465b5cb9870c0ab5beb27e39a7de66337066670ba721/tsflex-0.4.1.tar.gz",
"platform": null,
"description": "# <p align=\"center\"> <a href=\"https://predict-idlab.github.io/tsflex\"><img alt=\"tsflex\" src=\"https://raw.githubusercontent.com/predict-idlab/tsflex/main/docs/_static/logo.png\" width=\"66%\"></a></p>\n\n[](https://pypi.org/project/tsflex/)\n[](https://anaconda.org/conda-forge/tsflex)\n[](https://img.shields.io/pypi/pyversions/tsflex)\n[](https://codecov.io/gh/predict-idlab/tsflex)\n[](https://github.com/predict-idlab/tsflex/actions/workflows/codeql.yml)\n[](https://pepy.tech/project/tsflex)\n[](http://makeapullrequest.com)\n[](https://github.com/predict-idlab/tsflex/actions/workflows/deploy-docs.yml)\n[](https://github.com/predict-idlab/tsflex/actions/workflows/test.yml)\n[](https://discord.gg/k2d59GrxPX)\n\n\n<!--  -->\n\n> _tsflex_ is a toolkit for _**flex**ible **t**ime **s**eries_ [processing](https://predict-idlab.github.io/tsflex/processing) & [feature extraction](https://predict-idlab.github.io/tsflex/features), that is efficient and makes few assumptions about sequence data.\n\n#### Useful links\n\n- [Paper](https://www.sciencedirect.com/science/article/pii/S2352711021001904)\n- [Documentation](https://predict-idlab.github.io/tsflex/)\n- [Example (machine learning) notebooks](https://github.com/predict-idlab/tsflex/tree/main/examples)\n\n#### Installation\n\n| | command |\n| :--------------------------------------------------- | :------------------------------------ |\n| [**pip**](https://pypi.org/project/tsflex/) | `pip install tsflex` |\n| [**conda**](https://anaconda.org/conda-forge/tsflex) | `conda install -c conda-forge tsflex` |\n\n## Usage\n\n_tsflex_ is built to be intuitive, so we encourage you to copy-paste this code and toy with some parameters!\n\n### <a href=\"https://predict-idlab.github.io/tsflex/features/#getting-started\">Feature extraction</a>\n\n```python\nimport pandas as pd; import numpy as np; import scipy.stats as ss\nfrom tsflex.features import MultipleFeatureDescriptors, FeatureCollection\nfrom tsflex.utils.data import load_empatica_data\n\n# 1. Load sequence-indexed data (in this case a time-index)\ndf_tmp, df_acc, df_ibi = load_empatica_data(['tmp', 'acc', 'ibi'])\n\n# 2. Construct your feature extraction configuration\nfc = FeatureCollection(\n MultipleFeatureDescriptors(\n functions=[np.min, np.mean, np.std, ss.skew, ss.kurtosis],\n series_names=[\"TMP\", \"ACC_x\", \"ACC_y\", \"IBI\"],\n windows=[\"15min\", \"30min\"],\n strides=\"15min\",\n )\n)\n\n# 3. Extract features\nfc.calculate(data=[df_tmp, df_acc, df_ibi], approve_sparsity=True)\n```\n\nNote that the feature extraction is performed on multivariate data with varying sample rates.\n| signal | columns | sample rate |\n|:-------|:-------|------------------:|\n| df_tmp | [\"TMP\"]| 4Hz |\n| df_acc | [\"ACC_x\", \"ACC_y\", \"ACC_z\" ]| 32Hz |\n| df_ibi | [\"IBI\"]| irregularly sampled |\n\n### <a href=\"https://predict-idlab.github.io/tsflex/processing/index.html#getting-started\">Processing</a>\n\n[Working example in our docs](https://predict-idlab.github.io/tsflex/processing/index.html#working-example)\n\n## Why tsflex? \u2728\n\n- `Flexible`:\n - handles multivariate/multimodal time series\n - versatile function support\n => **integrates** with many packages for:\n - processing (e.g., [scipy.signal](https://docs.scipy.org/doc/scipy/reference/tutorial/signal.html), [statsmodels.tsa](https://www.statsmodels.org/stable/tsa.html#time-series-filters))\n - feature extraction (e.g., [numpy](https://numpy.org/doc/stable/reference/routines.html), [scipy.stats](https://docs.scipy.org/doc/scipy/reference/tutorial/stats.html), [antropy](https://raphaelvallat.com/antropy/build/html/api.html), [nolds](https://cschoel.github.io/nolds/nolds.html#algorithms), [seglearn](https://dmbee.github.io/seglearn/feature_functions.html)\u00b9, [tsfresh](https://tsfresh.readthedocs.io/en/latest/text/list_of_features.html)\u00b9, [tsfel](https://tsfel.readthedocs.io/en/latest/descriptions/feature_list.html)\u00b9)\n - feature extraction handles **multiple strides & window sizes**\n- `Efficient`:<br>\n - view-based operations for processing & feature extraction => extremely **low memory peak** & **fast execution time**<br>\n - see: [feature extraction benchmark visualization](https://predict-idlab.github.io/tsflex/#benchmark)\n- `Intuitive`:<br>\n - maintains the sequence-index of the data\n - feature extraction constructs interpretable output column names\n - intuitive API\n- `Few assumptions` about the sequence data:\n - no assumptions about sampling rate\n - able to deal with multivariate asynchronous data<br>i.e. data with small time-offsets between the modalities\n- `Advanced functionalities`:\n - apply [FeatureCollection.**reduce**](https://predict-idlab.github.io/tsflex/features/index.html#tsflex.features.FeatureCollection.reduce) after feature selection for faster inference\n - use **function execution time logging** to discover processing and feature extraction bottlenecks\n - embedded [SeriesPipeline](http://predict-idlab.github.io/tsflex/processing/#tsflex.processing.SeriesPipeline.serialize) & [FeatureCollection](https://predict-idlab.github.io/tsflex/features/index.html#tsflex.features.FeatureCollection.serialize) **serialization**\n - time series [**chunking**](https://predict-idlab.github.io/tsflex/chunking/index.html)\n\n\u00b9 These integrations are shown in [integration-example notebooks](https://github.com/predict-idlab/tsflex/tree/main/examples).\n\n## Future work \ud83d\udd28\n\n- scikit-learn integration for both processing and feature extraction<br>\n **note**: is actively developed upon [sklearn integration](https://github.com/predict-idlab/tsflex/tree/sklearn_integration) branch.\n- Support time series segmentation (exposing under the hood strided-rolling functionality) - [see this issue](https://github.com/predict-idlab/tsflex/issues/15)\n- Support for multi-indexed dataframes\n\n=> Also see the [enhancement issues](https://github.com/predict-idlab/tsflex/issues?q=is%3Aissue+is%3Aopen+label%3Aenhancement+)\n\n## Contributing \ud83d\udc6a\n\nWe are thrilled to see your contributions to further enhance `tsflex`.<br>\nSee [this guide](CONTRIBUTING.md) for more instructions on how to contribute.\n\n## Referencing our package\n\nIf you use `tsflex` in a scientific publication, we would highly appreciate citing us as:\n\n```bibtex\n@article{vanderdonckt2021tsflex,\n author = {Van Der Donckt, Jonas and Van Der Donckt, Jeroen and Deprost, Emiel and Van Hoecke, Sofie},\n title = {tsflex: flexible time series processing \\& feature extraction},\n journal = {SoftwareX},\n year = {2021},\n url = {https://github.com/predict-idlab/tsflex},\n publisher={Elsevier}\n}\n```\n\nLink to the paper: https://www.sciencedirect.com/science/article/pii/S2352711021001904\n\n---\n\n<p align=\"center\">\n\ud83d\udc64 <i>Jonas Van Der Donckt, Jeroen Van Der Donckt, Emiel Deprost</i>\n</p>\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Toolkit for flexible processing & feature extraction on time-series data",
"version": "0.4.1",
"project_urls": {
"Documentation": "https://predict-idlab.github.io/tsflex",
"Homepage": "https://github.com/predict-idlab/tsflex",
"Repository": "https://github.com/predict-idlab/tsflex"
},
"split_keywords": [
"time-series",
" processing",
" feature-extraction",
" data-science",
" machine learning"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "16c2ab19fb54b574712476e0f5ce755b6b2987c974a4f831396118d307ccdfed",
"md5": "3301975ed678fab5d8c2e6663b42b68d",
"sha256": "c6606ec13ca4757bad6582ab2ded0f3743041c4a495b0686187f0707aa2dd6c5"
},
"downloads": -1,
"filename": "tsflex-0.4.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "3301975ed678fab5d8c2e6663b42b68d",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": "<3.13,>=3.8",
"size": 66989,
"upload_time": "2024-09-06T09:26:32",
"upload_time_iso_8601": "2024-09-06T09:26:32.338244Z",
"url": "https://files.pythonhosted.org/packages/16/c2/ab19fb54b574712476e0f5ce755b6b2987c974a4f831396118d307ccdfed/tsflex-0.4.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "a0fa67d34578971b50b2465b5cb9870c0ab5beb27e39a7de66337066670ba721",
"md5": "37f0a99949427aa69c7fb9a130b9f183",
"sha256": "ce5d5db6725a5c5748e543836acdf2a5187e0777a100dc637765ad4d4fa67988"
},
"downloads": -1,
"filename": "tsflex-0.4.1.tar.gz",
"has_sig": false,
"md5_digest": "37f0a99949427aa69c7fb9a130b9f183",
"packagetype": "sdist",
"python_version": "source",
"requires_python": "<3.13,>=3.8",
"size": 59359,
"upload_time": "2024-09-06T09:26:33",
"upload_time_iso_8601": "2024-09-06T09:26:33.948858Z",
"url": "https://files.pythonhosted.org/packages/a0/fa/67d34578971b50b2465b5cb9870c0ab5beb27e39a7de66337066670ba721/tsflex-0.4.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-09-06 09:26:33",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "predict-idlab",
"github_project": "tsflex",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "tsflex"
}
JSON
