Tech Industry
Yesterday
641
Unemployment has been frustrating
World Conflicts
Yesterday
2154
Please accept Indians. I can’t go back. 🇺🇸🙏
Tech Industry
Yesterday
5317
Open AI GPT updates -- what's next? Any ideas?
Tech Industry
5h
1524
Serious question: why are Indian marriages arranged?
Tech Industry
Yesterday
1971
Getting sextorted
I like Facebook's prophet package for forecasting time series and I was excited to learn about neural prophet, however I tried my hands on a series data set and saw that they both performed poorly. Is there anyone here who understands the math behind them and open for a discussion/call where I can learn more and maybe debug this as well?
Lol! How about you help someone out and make a personal relationship, if this is a personal project.
That's wonderful ! I am happy for you but I can't afford you as I am just learning
Can't help... but would love to listen in
Sure. Will let you know if there is a session.
Read Hydmans books. Unless you have uni variate time series prophet performs poorly. Nueral prophet performs worse than prophet. You want a good benchmark try Amazons DeepAR. Is available on PyTorch Forecasting or MxNet GlounTS
Hyndman book is very good. I learned ses from it. Thanks for recommending deepar will check it out
In my own testing Nueral Prophet also randomly will crash with cryptic Python error messages. Also it supports no GPU acceleration so no point actually.
If this works on stocks then they would have been millionaires.
Thanks for sharing this. I understand what you are saying but the data I am using is hourly (temperature).
Is it a personal project or work?
Not even a project. Just tried it out for learning