https://github.com/heardacat/sklearn-predict
Run predictions inside the database. Aims to parse scikit-learn model objects and returns a formula that calculates predictions
https://github.com/heardacat/sklearn-predict
Last synced: 2 months ago
JSON representation
Run predictions inside the database. Aims to parse scikit-learn model objects and returns a formula that calculates predictions
- Host: GitHub
- URL: https://github.com/heardacat/sklearn-predict
- Owner: HeardACat
- Created: 2018-12-21T11:14:25.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-04-05T21:40:06.000Z (about 5 years ago)
- Last Synced: 2025-01-31T17:20:05.239Z (4 months ago)
- Language: Python
- Size: 31.3 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
sklearn-predict - Run predictions inside the database
=====================================================
**sklearn-predict** Run predictions inside the database.
It aims to transcompile [scikit-learn](https://scikit-learn.org) models to SQL equivalents in a
similar manner to [tidypredict](https://db.rstudio.com/tidypredict/).
Motivation
==========
Often when we're building Machine Learning models, the challenges are around
how can these models be built and deployed? How can we make these models smarter?
In many systems the complexity of the algorithms are often limited - this could
be due to the languages or libraries which are used, the latency or complexity
requirements which are in place, or other contraints.
In this package, we aim to provide a *minimal* set of tools to enable
machine learning deployment across wide variety of contexts. Whereas packages like
[sklearn-porter](https://github.com/nok/sklearn-porter) aim to support the final classifier, we aim to provide opinionated
feature engineering tools to support more complex workflows.
Goals
=====
To provide a pathway for Python users to deploy onto databases; through enabling simple
rule-based features and linear models
Non-Goals
=========
We do not aim to employ a range of complex mathematical transformations in our implementation.
For testing, we will leverage sqlite database as the limit of the allowable in-built
transformations for our machine learning pipelines.
The reasoning is that many more complex systems which we may aim to embed will likewise
only support a small subset of functions. We believe that these systems too should
be enabled to deployed more complex machine learning workflows in spite of such limitations
Testing
=======
```
pytest --capture=sys
pytest --cov=sklearn_predict
```
To do
=====
* Convert the required code to use SQlAlchemy ORM better (still undecided)
* Provide end to end examples