Skip to main content

The Data Transformation and Machine Learning Accelerator

Project description

PARDON

Pardon is a Python library developed as a machine learning (ML) and data transformation accelerator, to rapidly prepare data as well as test and deploy supervised ML models, by massively reducing the amount of code required during development.

The library can be used to perform data cleansing and transformation, as well as model and feature selection, data visualisations, and prediction auditing. Data can be quickly prepared for use with ML algorithms, and all data transformations are recorded in a 'script'. This 'script' is then automatically applied to data prior to predictions, which ensures data consistency, transparency, and removes the need for multi-stage data engineering while reducing the chances of error.

The intention of this library is not to replace or automate the work of a data scientist, but to enable integration work to begin early by rapidly deploying a working machine learning model. This then allows the data science work to be completed in parallel, with the final model being deployed on completion.

The pardon library functionality includes: • Data transformation • Data cleansing • Feature selection • Feature engineering • Model selection • Class balancing • Prediction auditing • Principal Component Analysis • Data scaling • Data encoding • Sentiment analysis • Hyperparameter tuning • Model “explainability” • Model performance metrics • Model training, testing and validation • Identifying data outliers • Testing model output in a REST API • Data visualisation • Use unsupervised learning to cluster data

Installation

The easiest way to install and use pardon is to pip install using pip or the latest pardon package whl file.

pip install pardon
pip install pardon-1.0.0-py3-none-any.whl

Usage

import pardon

# create a new ML Data object. The minimum requirements are a data file or data stream, and the name of the column you're trying to predict.
paml = pardon.Pardon(data='\\your\\specified\\directory\\data_file.csv', target='column_name_to_predict')

# rapid ml will perform the following steps by default:
# -> 1. Perform the minimum required data cleansing and transformation to prepare your data for model training.
# -> 2. Trains multiple models using the data and selects the best performing model.
# -> 3. Where applicable, display a summary of the model performance.
# -> 4. Save the model as a .pkl file to the location specified. You must ensure the filename is a .pkl file.
# -> 5. Display the output scores of all the models tested.
paml.rapid_ml(model_fullpath='\\your\\specified\\directory\\your_model_name.pkl')

# you can run a test on the localhost to see the API output
pardon.TestAPI(model_fullpath='\\your\\specified\\directory\\your_model_name.pkl')

Contributing

For changes, please contact pardonaccelerator@gmail.com to discuss your amendments or requirements.

Please make sure to update tests as appropriate.

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distribution

pardon-2.0.5-py3-none-any.whl (125.9 kB view details)

Uploaded Python 3

File details

Details for the file pardon-2.0.5-py3-none-any.whl.

File metadata

  • Download URL: pardon-2.0.5-py3-none-any.whl
  • Upload date:
  • Size: 125.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.1 CPython/3.10.8

File hashes

Hashes for pardon-2.0.5-py3-none-any.whl
Algorithm Hash digest
SHA256 6a2913a8e384678376ca3c193f670aa2e491ac72d87df4d190f44107c22ba3e4
MD5 5f4753057c9b9ed0a55a69ec9d0f4494
BLAKE2b-256 c66b981a505d14f41129d9ed6241c2b97315e3f313f14c9f877031214cbcb4f0

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page