User guide#
skore helps at structuring and storing what matters in your data science
experiments.
When it comes to data science, many libraries are available to help you experiment.
pandas or polars are great tools to explore and transform your data. skrub is the
one tool that brings the necessary statefullness to those transformations required by
the machine learning pipeline
(refer to).
scikit-learn and other scikit-learn compatible libraries (e.g. xgboost,
lightgbm) provide a set of algorithms to ingest those transformed data and create
predictive models. scikit-learn
provides even more tools to diagnose and evaluate those models.
skore is the cherry on the top. All those libraries are thought to be generic to
accommodate a wide range of use cases. When it comes to your particular use case,
your experience is the key to success by choosing the appropriate building blocks from
those libraries.
skore intends to consume the data science pipeline created by assembling those
libraries components and provide structured artifacts that would store the
information that matters for your use case. It will reduce the amount of time to
navigate through the documentation and guide you towards the right methodological
information to answer your questions. skore will also reduce the amount of code
required to show the information that matters, removing boilerplate code, making your
project easier to understand and maintain in the long run. Finally, skore provides
a way to store all those structured artifacts in a structured project and thus help
you later on to retrieve the experiment results that you need.