Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Common time series dataset to facilitate model training and evaluation #83

Open
ben-jy opened this issue Jan 17, 2025 · 2 comments
Open
Labels
core Related to core functions of onTime enhancement New feature or request

Comments

@ben-jy
Copy link
Collaborator

ben-jy commented Jan 17, 2025

When evaluating different models, datasets and their parameters may slightly differ. It would be beneficial to have a common dataset class that allows to define the important parameters for constructing a dataset, i.e. :

  • the covariate components (if needed, past, future, static, etc.)
  • the target components (components we must predict)
  • the split (train-val-test)
  • normalization
  • ...
@ben-jy ben-jy added enhancement New feature or request core Related to core functions of onTime labels Jan 17, 2025
@ben-jy
Copy link
Collaborator Author

ben-jy commented Feb 17, 2025

we already have a “BenchmarkDataset” class that seems to cover most of the points listed above. Why would we need another dataset class?

ben-jy added a commit that referenced this issue Feb 19, 2025

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
…columns. Although, models still make prediction on all input components. It should be solved in issue #83. Fix parameter order in BenchmarkDataset constructor, and add helper methods to retrieve input only columns. Update documentation accordingly.
@ben-jy
Copy link
Collaborator Author

ben-jy commented Feb 26, 2025

we already have a “BenchmarkDataset” class that seems to cover most of the points listed above. Why would we need another dataset class?

Some model implementations, such as Moirai e.g., can receive specific parameters for their fit/predict methods that might improve quality of prediction, such as covariate, static features, etc.
We must find a way to give these parameters in a common way across models, that is why this issue was created. With a new time series inheriting from onTime TimeSeries object, we could give this information, and model could check on them.
Still, maybe we can find another way to perform this.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
core Related to core functions of onTime enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

1 participant