Tweedie objective support for XGB model conversion
- Dominant language
- Python
- Stars
- 5.4k
- Forks
- 850
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 10
Description
## 🌱 Describe your Feature Request
In `/converters/xgboost/_tree_ensemble.py` you list the `reg:tweedie` objective in XGB under corner cases you don't support. This is actually quite a useful objective for retailers, since the predicted revenue of a customer is naturally modelled as the compound of a discrete conversion distribution with a continuous distribution of spend.
## Use cases
Modelling customer spend (on device), particularly where a large percentage of users will not convert.
## Describe alternatives you've considered
`reg:squarederror`, but it doesn't work as well
## Additional context
n/a
Contributor guide
Research direction
Start in /converters/xgboost/_tree_ensemble.py, where reg:tweedie is listed among unsupported corner cases. Trace how XGB objectives are converted and determine the validation needed for Tweedie models; done means models using this objective convert successfully without remaining unsupported-objective handling.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100