This page describes the main principles that drive the development of Probatus as well as the general directions, in which the development of the package will be heading.
Probatus has started as a side project of Data Scientists at ING Bank.
Later, we have decided to open-source it, in order to share the tools and enable collaboration with the Data Science community.
We mainly focus on analyzing the following aspects of building classification models:
- Model input: the quality of the dataset and how to prepare it for modelling,
- Model performance: the quality of the model and stability of the results.
- Model interpretation: understanding the model decision making,
Our main goals are:
- Continue maintaining the tools that we have built, and make sure that they are well documented and tested
- Continuously extend functionality available in the package
- Build a community of users, which use the package in day-to-day work and learn from each other, while contributing to Probatus
The main principles that drive development of Probatus are the following
- Usefulness - any tool that we build should be useful for a broad range of users,
- Simplicity - simple to understand and analyze steps over state-of-the-art,
- Usability - the developed functionality must be have good documentation, consistent API and work flawlessly with scikit-learn compatible models,
- Reliability - the code that is available for the users should be well tested and reliable, and bugs should be fixed as soon as they are detected.
We are open to new ideas, so if you can think of a feature that fits the vision, make an issue and help us further develop this package.