0
$\begingroup$
What I have

A time series dataset of time stamps (hourly resolution), some covariates (like store foot-traffic) and items sold.

What to forecast

Number of items sold for next 24 hours, i.e. 24 forecasts.

Requirement

Have to use generative AI for modelling and benchmark against non-generative methods.

My question

I have used models like TFT, N-beats etc. for time-series forecasting using Gluon, Darts, but are they considered generative? If not, any tutorial or sample code with generative time series forecasting (preferably implemented with tensorflow, python) will be helpful. Or else, if I do not want to dig into the implementation detail, just is there any readily available python framework with simple interface to get forecasts using generative algorithms? I am looking for something akin to Huggingface models for natural language understanding.

Improve this question
$\endgroup$
3
  • $\begingroup$ Using generative stuff for forecasts is not that common of an idea (definitely not common enough for simple interfaces to exist). This is most likely because, well, generative models are used to generate stuff, while predictive models are used to make forecasts. $\endgroup$ Commented Jul 21, 2023 at 8:01
  • 1
    $\begingroup$ agree with @liakoyras that it is not a common thing, at least not that I have heard of. Why must it be a generative model? Forecasting is not something generative model is designed for, so I doubt you will get better result even if there is a way. $\endgroup$ Commented Jul 27, 2023 at 10:14
  • $\begingroup$ "Why must it be a generative model?" I am assuming either some CEO somewhere decided to join the bandwagon without knowing shit, or an academic researcher pushing boundaries. My money is on the first. $\endgroup$ Commented Jul 27, 2023 at 11:41

0

Reset to default

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.