discovery-capability 0.16.4

Data Science to production accelerator

Introduction

Unfortunately, 85% of data science projects fail due to a lack of understanding of the real business problem. This is usually because of poor communication between data scientists and business teams, resulting in a disconnect between the two groups. Project Hadron has been built to bridge the gap between data scientists and data engineers. More specifically between machine learning business outcomes or use case and a product pipeline. It translates the work of data scientists into meaningful, production ready solutions that can be easily integrated into a DevOps, CI/CD pipeline.

Project Hadron addresses data selection, feature engineering and feature transformation as part of the critical preprocessing of a Machine Learning pipeline or System Data pipeline. At its core the code uses PyArrow as its canonical combining with Pandas as a directed specialist toolset. PyArrow complements Pandas by providing a more memory-efficient in-memory representation, enabling efficient data interchange between different systems, supporting distributed computing, and enhancing compatibility with other programming languages. When used together, Pandas and PyArrow form a powerful combination for handling diverse data processing tasks efficiently.

Data Selection and Feature Engineering

Data selection and feature engineering is the art/science of converting raw data to a form that optimises the success of the next step in a pipeline. This involves a skilled blend of domain expertise, intuition and mathematics. Data selection and feature engineering are the most essential part of building a useable machine learning or data project, constituting an average of 80% of the project’s time, even with hundreds of cutting-edge machine learning algorithms appearing.

Prof Domingos, the author of ‘The Master Algorithm’ says:

"At the end of the day, some machine learning projects succeed and some fail. What makes the
difference? Easily the most important factor is the features used."

Preprocessing

The term “data preprocessing” is commonly used in the field of data science and machine learning to refer ata selection and feature engineering as steps taken to clean, format, and organize raw data into a suitable format for Model Evaluation & Tunning

Main features

• Data Preparation

• Feature Selection

• Feature Engineering

• Feature Cataloguing

• Augmented Knowledge

• Synthetic Feature Build

Feature transformers

Project Hadron is a Python library with multiple transformers to engineer and select features to use across a synthetic build, statistics and machine learning.

• Missing data imputation

• Categorical encoding

• Variable Discretisation

• Outlier capping or removal

• Numerical transformation

• Redundant feature removal

• Synthetic variable creation

• Synthetic multivariate

• Synthetic model distributions

• Datetime features

• Time series

Project Hadron allows one to present optimal parameters associated with each transformer, allowing different engineering procedures to be applied to different variables and feature subsets.

Background

Born out of the frustration of time constraints and the inability to show business value within a business expectation, this project aims to provide a set of tools to quickly build production ready data science disciplines within a component based solution demonstrating coupling and cohesion between each disipline, providing a separation of concerns between components.

It also aims to improve the communication outputs needed by ML delivery to talk to Pre-Sales, Stakholders, Business SME’s, Data SME’s product coders and tooling engineers while still remaining within familiar code paradigms.

Getting Started

The discovery-transition-ds package is a set of python components that are focussed on Data Science. They are a concrete implementation of the Project Hadron abstract core. It is build to be very light weight in terms of package dependencies requiring nothing beyond what would be found in an basic Data Science environment. Its designed to be used easily within multiple python based interfaces such as Jupyter, IDE or terminal python.

Package Installation

The best way to install the component packages is directly from the Python Package Index repository using pip.

The component package is discovery-transition-ds and pip installed with:

python -m pip install discovery-transition-ds

if you want to upgrade your current version then using pip install upgrade with:

python -m pip install -U discovery-transition-ds

This will also install or update dependent third party packages. The dependencies are limited to python and related Data Science tooling such as pandas, numpy, scipy, scikit-learn and visual packages matplotlib and seaborn, and thus have a limited footprint and non-disruptive in a machine learning environment.

Get the Source Code

discovery-transition-ds is actively developed on GitHub, where the code is always available.

You can clone the public repository with:

$ git clone [e-mail]:project-hadron/discovery-transition-ds.git

Once you have a copy of the source, you can embed it in your own Python package, or install it into your site-packages easily running:

$ cd discovery-transition-ds
$ python -m pip install .

Release Process and Rules

Versions to be released after 3.5.27, the following rules will govern and describe how the discovery-transition-ds produces a new release.

To find the current version of discovery-transition-ds, from your terminal run:

$ python -c "import ds_discovery; print(ds_discovery.__version__)"

Major Releases

A major release will include breaking changes. When it is versioned, it will be versioned as vX.0.0. For example, if the previous release was v10.2.7 the next version will be v11.0.0.

Breaking changes are changes that break backwards compatibility with prior versions. If the project were to change an existing methods signature or alter a class or method name, that would only happen in a Major release. The majority of changes to the dependant core abstraction will result in a major release. Major releases may also include miscellaneous bug fixes that have significant implications.

Project Hadron is committed to providing a good user experience and as such, committed to preserving backwards compatibility as much as possible. Major releases will be infrequent and will need strong justifications before they are considered.

Minor Releases

A minor release will include addition methods, or noticeable changes to code in a backward-compatable manner and miscellaneous bug fixes. If the previous version released was v10.2.7 a minor release would be versioned as v10.3.0.

Minor releases will be backwards compatible with releases that have the same major version number. In other words, all versions that would start with v10. should be compatible with each other.

Patch Releases

A patch release include small and encapsulated code changes that do not directly effect a Major or Minor release, for example changing round(... to np.around(..., and bug fixes that were missed when the project released the previous version. If the previous version released v10.2.7 the hotfix release would be versioned as v10.2.8.

Reference

Python version

Python 3.7 or less is not supported. Although it is recommended to install discovery-transition-ds against the latest Python version or greater whenever possible.

Pandas version

Pandas 1.0.x and above are supported but It is highly recommended to use the latest 1.0.x release as the first major release of Pandas.

GitHub Project

discovery-transition-ds: https://github.com/project-hadron/discovery-transition-ds.

Change log

See CHANGELOG.

License

This project uses the following license: MIT License: https://opensource.org/license/mit/.

Authors

Gigas64 (@gigas64) created discovery-transition-ds.