rvandewater/recipys

🥧ReciPys: easily define and execute preprocessing and feature engineering steps on Pandas dataframes.

Hi 👋, I am a PhD candidate in AI in healthcare at the Hasso Plattner Institute. I am a Computer Scientist and Data Scientist by education. I specialize in using predictive modelling methods to derive insights from healthcare data, thereby prolonging the life of patients. I am interested in applying the ml methods to improve healthcare.

TLDR:

• 🧠 AI in Health Researcher
• 🧑‍🎓 Ph.D. Candidate
• 🏛️ Hasso Plattner Institute
• 🏠 Berlin, Germany

recipys's Issues

Common tasks in `Step.transform()`

in rvandewater/YAIB#41 we raised the question whether common tasks in Step.transform() -- such as calling _check_ingredients() -- should be moved into the parent class. This is currently the case for Step.fit() with a dedicated do_fit() function that must be overridden by child classes.

Some steps are very slow

Steps like StepHistorical and StepImputeFill can take a long time. We may want to think about ways to speed those up. For example, I have in the past been able to speed up pandas ffill by inserting stopping rows between patients and then running ffill on the ungrouped dataframe.

Data augmentation `step` for "Last x hours"

Implement a data augmentation step that extracts averages over last x hours. Not mandatory (at the moment), but nice to have.

Add section about how to use tool to README

Would be nice to have a section in the README.md version showcasing how to use this library and how to integrate it in a scikit-learn pipeline flow