search

robjhyndman / forecast

Forecasting Functions for Time Series and Linear Models
http://pkg.robjhyndman.com/forecast
1.13k stars 342 forks source link

Can I create a python implementation of the entire package ? #754

Closed Ic3fr0g closed 6 years ago

Ic3fr0g commented 6 years ago

What?

I want to create a python implementation of this package.

Why?

There is a lack of good python libraries that support time-series analysis and forecasting models. For example, prophet is limited by daily forecasting, statsmodels.tsa supports ARIMA models, pyflux supports GARCH models and scikit-learn does not have forecasting models. In short, there's no library out there that is quite like forecast for R, having metrics, models, seasonal-plots, etc. all in one place. Creating an implementation in python will help to compare statistical models with deep learning models in the same environment. Moreover, this would be a great project to include native ways to handle multivariate time-series forecasting, XGBoost Regression (à la forecastxgb) and possibly even LSTMs.

Would you accept a pull request for the same? Or would you rather it be a different repository?

robjhyndman commented 6 years ago

In principle I am ok with a python version of the package provided it remains GPL-licensed. However, we are currently rewriting the package from scratch (https://github.com/tidyverts/fable) and some of that work might be better ported than the older forecast package code.

Unfortunately, I don't have time to be actively involved. It would be best to be a different repo.

Ic3fr0g commented 6 years ago

Wow! I just spent an entire afternoon looking at tidyverts and all repos associated with it. This is fantastic but it's going to take me a while to get started.