Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Hi! I'm one core developer (and creator) of the library. Thanks for all the comments. I just wanted to highlight a couple of things that we think are quite cool about Darts:

* It makes using all sorts of forecasting models (from ARIMA to deep learning) easy, using fit() and predict(), similar to scikit-learn.

* It's easy to fit deep learning and other ML-based models on multiple time series, potentially on big datasets too. The time series can be multivariate.

* Darts is not only wrapping existing models. We also have our own implementations, for instance of TCN (Temporal Convolutional Networks), or adaptations N-BEATS (which we extended to handle multivariate series), DeepAR and others.

* Darts makes it very easy to include past and/or future covariates as inputs for the predictions.

* Some models offer probabilistic forecasts; sometimes with the possibility to configure your favourite likelihood function (e.g. Gaussian for continuous values or Poisson for discrete values).

* Everything uses the "TimeSeries" class, which makes the API consistent across tools and models, and make it harder to make mistakes. For instance it's easy to consume the output of one model by another model, and all models can be backtested the same way.



I love to see more time series models becoming available in an easy-to-use format. There's always been such a gap between what is possible and what is convenient to use, much moreso than with other kinds of models.

This was also one of the areas where R always had better options than Python, but that seems to be gradually changing as well.

Darts looks very thorough and user-friendly, it makes me really want to work on a forecasting project!

It might be very helpful to readers/users if you could add a section to your documentation comparing Darts to Tslearn [0] (edit, and Sktime [1]), which already has a lot of time series models with the Scikit-learn style interface.

It would also be helpful to have some kind of writeup that explains the TimeSeries data structure and why you use that, instead of just a Series/DataFrame.

Finally - you really shouldn't say "non-Facebook alternative", because your Prophet implementation is literally a wrapper around Facebook's Prophet library. If anything, I suggest moving the Prophet, Torch, and Pmdarima dependencies to setuptools "extras", so you don't force the users to depend on those projects.

[0]: https://tslearn.readthedocs.io/en/stable/index.html

[1]: https://news.ycombinator.com/item?id=28155384


Thanks for the feedback, I absolutely agree about the need for easy-to-use tools for dealing with time series. This is exactly the motivation that prompted us to work on Darts initially.

I like your suggestions of adding comparison to the few other libraries out there, as well as explaining the need for having our own TimeSeries data structure. We should try to do that sometime soon.

Concerning dependencies, we already have some dependencies as extras. "pip install darts" will install everything, but "pip install u8darts" will install only the core (without Prophet and pmdarima), or "pip install u8darts[torch]" only the core+pytorch models.


Regarding deps, I just saw that you have two setup.py files, and one of them has the `extras` listed. Good stuff!


> I suggest moving the Prophet, Torch, and Pmdarima dependencies to setuptools "extras",

Yep, the dependence on Prophet prevented me to quickly try darts on a Windows machine because of the pystan issue with prophet:

https://github.com/facebook/prophet/issues/732

May try it later on Linux.


You can try installing it using

    pip install 'u8darts[torch]'
This will give you everything besides Prophet and pmdarima.


Will you add support for anomaly detection or classification (something like predictive maintenance)?


In short: yes, at least anomaly detection. That's one of the main things on our to-do list.


Do you have any plans to implement some sort of model averaging or stacking? I believe it would bring great benefits to this landscape to have a working implementation of hierarchical stacking across various backends wrapped in a Python library.


We do have ensemble models. Something like

    model = NaiveEnsembleModel([model1, model2, ...])
    model.fit(my_series)
    prediction = model.predict()
Will return an average prediction. Look at RegressionEnsembleModel for an ensemble model which uses a regression model to learn how to combine the individual forecasts.

At the moment Darts doesn't have hierarchical reconciliation methods (if that's what you meant), but it's on the backlog :)


FYI, Hacker News has code formatting if you indent by 4 spaces:

    model = NaiveEnsembleModel([model1, model2, ...])
    model.fit(my_series)
    prediction = model.predict()


Oh, nice, I didn't know that. I've edited my reply, thanks!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: