march_applications_21
march_applications_21 copied to clipboard
Vectorized implementations of 3/4 partner selection approaches
Coded mostly in numpy, pandas and scipy. I have focused on different strategies to improve the speed further.
- Avoid for-loops as much as possible
- Use broadcasting and matrix multiplication instead
- Experiment with advanced indexing and one hot encoded matrix multiplication
In this PR the following is included:
- The base structure for a module that can be used for partner selection.
- A stock downloader which is build on top of y-finance
- A notebook with a tutorial on how to use it.
Outlook:
- I have not yet finished the vectorised copula approach.
- I wanted to build a few dashboards outside of Python
- Unit testing is still behind
- An architecture to make it more usable with static methods for example.
If you have any questions. I'm looking forward to discuss them.