I’m a beginner and am planning to apply transfer learning to a time series regression problem where I’ve data consisting of number of application forms received over an year and I’ve to extrapolate that to another couple of months predicting the number of application forms received in those months.

Could you please point me out to some good state-of-the-art time series regression models that I can start with?

Are you using a univariate time-series data? If that’s the case it’s typically the case that ML models perform worse than statistical models like ARIMA when it comes to forecasting, but a carefully crafted combination of the two methods can yield SOTA results. Although I can’t think of any out of the box approaches in fastai that would let you do forecasting easily

Thanks for your inputs.
Yes, in the current problem I’m using a univariate data. I’ll go through the link you shared and will try figuring out how can I use that in my case.
Just curious to know what would you recommend if I have multivariate time-series data for a regression task.

Just curious to know what would you recommend if I have multivariate time-series data for a regression task.

That’s too hard to say imo. If variables other than time is the main driver, then you may end up in a situation similar to the rossman kaggle competition where you’re mostly dealing with tabular data with a time component. But if time is still the main component it’s hard to say because it’s not really an popular area of public research as far as I know. If I were to guess what might work, I’d say it will involve some stats based forecasting on the dependent variable, some tabular model and some meta-learner to combine everything together