Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/riverliway/meow
For the Hack 4 Impact hackathon!
https://github.com/riverliway/meow
Last synced: 29 days ago
JSON representation
For the Hack 4 Impact hackathon!
- Host: GitHub
- URL: https://github.com/riverliway/meow
- Owner: riverliway
- Created: 2024-02-25T17:17:09.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2024-02-26T02:54:16.000Z (11 months ago)
- Last Synced: 2024-02-26T18:57:23.150Z (11 months ago)
- Language: Jupyter Notebook
- Size: 656 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Classify Financial Transactions
The HCB API provides the [following information](https://hcb.hackclub.com/docs/api/v3/schemas/Transaction) for each transaction:
`amount_cents` **integer** - the number of cents for this transaction
`memo` **string** - a sentence representing what the transaction is about
`date` **string** - the date for the transaction (format unknown)
`type` **invoice, donation, ach_transfer, check, transfer, bank_account_transaction, card_charge** - how the transaction was processed (convert to int-enum)
`organization.category` **hackathon, hack_club, nonprofit, event, high_school_hackathon, robotics_team, hardware_grant, hack_club_hq, outernet_guild, grant_recipient, salary, ai, hcb_internals** - what type of organization this is for (convert to int-enum)
`tags.label` **string** - the label for the current tags on this transaction (target for prediction)
In addition, there are several optional fields for each payment type. Each of them have a `memo` field.
## Architecture Ideas
First embed the memo and check to see if there is another embedded memo with a tag. If so, return that tag.
We can use weak supervision to run the untagged transactions through a generative LLM.
We can use strong supervision with the tagged transactions to train a custom model from scratch.
### Custom Model:
* Log10 transform the `amount_cents` so the model is more sensitive to the magnitude
* Embed the memo