[New models - placeholder] More new models · unit8co/darts#660

(0 comments) (0 reactions) (0 assignees)Python (6,832 stars) (762 forks)batch import

good first issuenew model

Description

In general we might be interested by implementing some of the models below. This list is not exhaustive and we welcome reference implementations of other/newer models.

AdaRNN: https://github.com/jindongwang/transferlearning/tree/master/code/deep/adarnn
TBATS: https://github.com/intive-DataScience/tbats
Neural Prophet: https://github.com/ourownstory/neural_prophet
Multivariate Singular Spectrum Analysis https://arxiv.org/pdf/2006.13448.pdf
Quantile Regression Forest: https://www.statsmodels.org/devel/generated/statsmodels.regression.quantile_regression.QuantReg.html
Orbit https://arxiv.org/pdf/2004.08492.pdf
ensemble.HistGradientBoostingRegressor, ensemble.GradientBoostingRegressor, linear_model.Ridge
DA-RNN https://arxiv.org/abs/1704.02971

Contributor guide

Tech stack: python
Domain: datamachine learning
Issue type: feature
Difficulty: 3
Estimated time: 1-2 days
Activity status: fresh
Clarity: needs investigation
Prerequisites: PythonTime series forecastingBasic machine learning
Newbie friendliness: 40
Research direction: This issue lists several candidate models (e.g., AdaRNN, NeuralProphet). A good starting point would be to select one model from the list, study its reference implementation and paper, and then implement it within the darts library. Check the existing model implementations in darts for patterns (e.g., how other models are defined). There is no linked PR or further discussion, so consider commenting on the issue to express interest and ask for any preferences before starting.