miguelgfierro
commented
1 year ago - [x] Follow the structure of https://github.com/miguelgfierro/project_template
- [x] Create a MLOps procedure with tests and GitHub actions. Every time a PR is raised, all the tests need to pass. Follow https://github.com/miguelgfierro/project_template
- [x] Do not upload binary files to github, like https://github.com/IgnacioChirinos/CHATBOTS-RECO_PROJECT/blob/main/RecoRelease%231/Data%20Collection_Creation/Superstore.csv
- [x] Use notebooks to show the workflow and python files for the libraries https://github.com/IgnacioChirinos/CHATBOTS-RECO_PROJECT/blob/main/RecoRelease%231/Data%20Collection_Creation/Create%20Supermarket_data.py
- [x] Add the link to the dataset source
- [x] Instead of using a for loop, use numpy to generate the random matrices
- [x] Instead of hardcoding the data https://github.com/IgnacioChirinos/CHATBOTS-RECO_PROJECT/blob/main/RecoRelease%231/Data%20Collection_Creation/Create%20all%20data.py#L4, use neutral path or a downloader like maybe_download
- [x] Add some documentation and explanation to the notebooks.
- [x] Compute the ranking metrics, maybe compare your algo with other algos.
- [x] In the interactions, you go from purchase 1, wish list 3, view 2, add to cart 0. Change the order so purchase is the strongest signal. From less to more: view, wish list, add to cart, purchase. Do not start with 0, start from 1, and if you want to can try different weights, for example, purchase 4 or 10.
- [ ] The tag is typically a number, not a string. Here is an example: https://pypi.org/project/lightgbm/#history In this tutorial you can see how to rename a tag https://github.com/miguelgfierro/codebase/wiki/Git-tutorial
carlo8pastor