# THOR: Time-Varying High-dimensional Ordinal Regression
THOR is a new autoML tool for temporal tabular datasets and time series. It handles high dimensional datasets with distribution shifts better than other tools. Inspired by the Numerai competiton, THOR has evolved from a specific tool for Numerai competition into a general ML pipeline which has many applications in finance and healthcare.
## Gradient Boosting Decision Trees
Customised LightGBM-based Gradient Boosting Decision Trees models for temporal tabular datasets.
## Deep Learning Models
A novel deep learning model for temporal tabular datasets, which complements well with the above GBDT-based models.
## TimeSeries Hybrid
A new method which combines classical and machine learning techniques for feature engineering and sequence modelling. A hybrid approach which demonstrate robust performances for high dimensional time-series.
## Dynamic Hyperopt
Apply dynamic hyper-parameter optimisation methods under incremental learning framework for temporal tabular and multi-variate timeseries datasets.
## Portfolio Optimisation
A new method to combine predictions from machine learning model using well-known theories from finance. Using the best research results from both finance and reinforcement learning, the method can maximise the portfolio return (or minimise the given loss function) within required risk metrics.
## Trend Follower
An enhanced implementation of trend following strategies with improved robustness and lower risks than the standard implementation of moving averages. Offer a better way to replicate trend following strategies and build new strategies that are less correlated to existing ones.
### Docker
As this packages used various machine learning and CUDA libaries for GPU support, we recommend to use docker to manage the dependencies.
The image is now uploaded on [Docker Hub](https://hub.docker.com/repository/docker/thomaswong2023/thor-public/general).
```bash
docker pull thomaswong2023/thor-public
docker run --gpus device=all -it -d --rm --name thor-public-example thomaswong2023/thor:public bash
```
### PyPI
This project is also on [PyPI](https://pypi.org/project/thor-public/).
Install the package with the following command. Dependencies are not installed with the package
```bash
pip install thor-public -r requirements.txt
```
## Citation
If you are using this package in your scientific work, we would appreciate citations to the following preprint on arxiv.
[Robust machine learning pipelines for trading market-neutral stock portfolios](https://arxiv.org/abs/2301.00790 )
Bibtex entry:
```
@misc{https://doi.org/10.48550/arxiv.2301.00790,
doi = {10.48550/ARXIV.2301.00790},
url = {https://arxiv.org/abs/2301.00790},
author = {Wong, Thomas and Barahona, Mauricio},
keywords = {Computational Finance (q-fin.CP), Computational Engineering, Finance, and Science (cs.CE), Machine Learning (cs.LG), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Robust machine learning pipelines for trading market-neutral stock portfolios},
publisher = {arXiv},
year = {2023},
copyright = {Creative Commons Attribution 4.0 International}
}
```
Raw data
{
"_id": null,
"home_page": "https://github.com/ThomasWong2022/thor-public",
"name": "thorml",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": "",
"keywords": "autoML",
"author": "Thomas Wong",
"author_email": "mw4315@ic.ac.uk",
"download_url": "https://files.pythonhosted.org/packages/b0/f5/c118dc5f65906ea5ded766dc9b88320aaba220305963cd041da45dab69e5/thorml-0.1.1.1.tar.gz",
"platform": null,
"description": "# THOR: Time-Varying High-dimensional Ordinal Regression \n\nTHOR is a new autoML tool for temporal tabular datasets and time series. It handles high dimensional datasets with distribution shifts better than other tools. Inspired by the Numerai competiton, THOR has evolved from a specific tool for Numerai competition into a general ML pipeline which has many applications in finance and healthcare. \n\n\n## Gradient Boosting Decision Trees\n\nCustomised LightGBM-based Gradient Boosting Decision Trees models for temporal tabular datasets.\n\n## Deep Learning Models \n\nA novel deep learning model for temporal tabular datasets, which complements well with the above GBDT-based models.\n\n## TimeSeries Hybrid \n\nA new method which combines classical and machine learning techniques for feature engineering and sequence modelling. A hybrid approach which demonstrate robust performances for high dimensional time-series.\n\n## Dynamic Hyperopt \n\nApply dynamic hyper-parameter optimisation methods under incremental learning framework for temporal tabular and multi-variate timeseries datasets. \n\n## Portfolio Optimisation\n\nA new method to combine predictions from machine learning model using well-known theories from finance. Using the best research results from both finance and reinforcement learning, the method can maximise the portfolio return (or minimise the given loss function) within required risk metrics.\n\n## Trend Follower\n\nAn enhanced implementation of trend following strategies with improved robustness and lower risks than the standard implementation of moving averages. Offer a better way to replicate trend following strategies and build new strategies that are less correlated to existing ones. \n\n\n### Docker \n\nAs this packages used various machine learning and CUDA libaries for GPU support, we recommend to use docker to manage the dependencies. \n\nThe image is now uploaded on [Docker Hub](https://hub.docker.com/repository/docker/thomaswong2023/thor-public/general).\n\n```bash\ndocker pull thomaswong2023/thor-public\ndocker run --gpus device=all -it -d --rm --name thor-public-example thomaswong2023/thor:public bash\n\n```\n\n\n\n### PyPI \n\nThis project is also on [PyPI](https://pypi.org/project/thor-public/).\n\nInstall the package with the following command. Dependencies are not installed with the package \n\n```bash\npip install thor-public -r requirements.txt\n\n```\n\n\n## Citation\nIf you are using this package in your scientific work, we would appreciate citations to the following preprint on arxiv.\n\n[Robust machine learning pipelines for trading market-neutral stock portfolios](https://arxiv.org/abs/2301.00790 )\n\nBibtex entry:\n```\n@misc{https://doi.org/10.48550/arxiv.2301.00790,\n doi = {10.48550/ARXIV.2301.00790},\n \n url = {https://arxiv.org/abs/2301.00790},\n \n author = {Wong, Thomas and Barahona, Mauricio},\n \n keywords = {Computational Finance (q-fin.CP), Computational Engineering, Finance, and Science (cs.CE), Machine Learning (cs.LG), FOS: Economics and business, FOS: Economics and business, FOS: Computer and information sciences, FOS: Computer and information sciences},\n \n title = {Robust machine learning pipelines for trading market-neutral stock portfolios},\n \n publisher = {arXiv},\n \n year = {2023},\n \n copyright = {Creative Commons Attribution 4.0 International}\n}\n```\n\n\n\n\n\n\n",
"bugtrack_url": null,
"license": "",
"summary": "AutoML tools for Tabular Datasets",
"version": "0.1.1.1",
"split_keywords": [
"automl"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "0696368fac39f155e09cb2f3694515e2c107543a06c59462026ad31d2f1a1d08",
"md5": "4ab9542489565d096aaa6b7a37a1f2f1",
"sha256": "f44a5da8002eea2617dcb6dc68767e8ef6d294aff24e9d005e1efbf2b6bc62b9"
},
"downloads": -1,
"filename": "thorml-0.1.1.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "4ab9542489565d096aaa6b7a37a1f2f1",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.7",
"size": 36706,
"upload_time": "2023-03-23T08:31:58",
"upload_time_iso_8601": "2023-03-23T08:31:58.013709Z",
"url": "https://files.pythonhosted.org/packages/06/96/368fac39f155e09cb2f3694515e2c107543a06c59462026ad31d2f1a1d08/thorml-0.1.1.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "b0f5c118dc5f65906ea5ded766dc9b88320aaba220305963cd041da45dab69e5",
"md5": "a6f77857344dad51f5fd4eebe4e8bb44",
"sha256": "ce4d4f64e6d7ed4e133dc06252aca6f6ce64403fe80ba980346e326c036bb944"
},
"downloads": -1,
"filename": "thorml-0.1.1.1.tar.gz",
"has_sig": false,
"md5_digest": "a6f77857344dad51f5fd4eebe4e8bb44",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 32072,
"upload_time": "2023-03-23T08:31:59",
"upload_time_iso_8601": "2023-03-23T08:31:59.959214Z",
"url": "https://files.pythonhosted.org/packages/b0/f5/c118dc5f65906ea5ded766dc9b88320aaba220305963cd041da45dab69e5/thorml-0.1.1.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-03-23 08:31:59",
"github": true,
"gitlab": false,
"bitbucket": false,
"github_user": "ThomasWong2022",
"github_project": "thor-public",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [],
"tox": true,
"lcname": "thorml"
}
JSON
