Opened 13 years ago
Last modified 7 years ago
#1636 new enhancement
Weighted evaluation of data analysis models
| Reported by: | gkronber | Owned by: | gkronber |
|---|---|---|---|
| Priority: | medium | Milestone: | HeuristicLab 4.x Backlog |
| Component: | Problems.DataAnalysis | Version: | 3.3.5 |
| Keywords: | Cc: |
Description (last modified by gkronber)
It should be possible to indicate weights for observations. Benefits:
- More important data points can be weighted more strongly
- Outliers can be remove by setting the weight to zero
- If we know the variance for observations we can consider non-equal variances for observations
- We could more easily aggregate multiple observations and put a higher weight on it (the average has less variance).
Change History (3)
comment:1 Changed 11 years ago by gkronber
- Priority changed from medium to low
comment:2 Changed 7 years ago by gkronber
- Description modified (diff)
- Milestone changed from HeuristicLab 3.3.x Backlog to HeuristicLab 4.x Backlog
- Priority changed from low to medium
- Summary changed from Functionality for conditional evaluation of data analysis models to Weighted evaluation of data analysis models
comment:3 Changed 7 years ago by abeham
I think this would be useful in surrogate assisted optimization when you want to consider better solutions with higher weight than worse solutions.
Note: See
TracTickets for help on using
tickets.
This could also be implemented via weights for rows in the dataset.
Weights allow more fine grained control over evaluation. There is seemingly no other ticket for implementation of weighted evaluation of regression models. Therefore, I'm re-purposing this ticket.