[ghego1]

Their findings are very much in line with my experience with fbprophet, it is usually the least effective lib to predict a time series in my tests.

[AndrewKemendo]

What would you suggest is the most effective time series analysis library?

[isoprophlex]

Temporal Fusion Transformers look very cool: accepts timeseries vectors but also categoricals and numerical features, outputs distribution using quintile regression.

https://arxiv.org/pdf/1912.09363v2.pdf

Available as pytorch implementation:

https://pytorch-forecasting.readthedocs.io/en/latest/index.h...

[chroem-]

SoTA in 2021 is the bespoke transformer model you implement yourself based on the idiosyncrasies in your data. Unless a lot of money is on the line, that is overkill for most situations though.

[SpaceManNabs]

What about be the feature vector for each of the x_i in the sequence? Would it just be the numerical value? Or could you also add a time embedding?