0

I'm trying to solve a problem to train my skill. I've daily data (each day during 10 years,2010 to 2020, except saturday and sunday). I would like to predict like ten day after my data end.

Problem is my model are atrocious. I tried xgboost, svm and Prophet for now.

I think the problem is : I always worked on monthly data, and with monthly it's far easier to detect seasonality and so on. With daily data I'm kind of lost.

I thought : "Why not transform into monthly instead of daily ?" But how to predict only few days with monthly ? I don't think it's possible.

So if you have some tips to improve my work with daily data, I'll be very grateful. I'm quite new so any hel would be appreciated.

Thanks. :)

LaurieFalcon
  • 1
  • Are you familiar with the Efficient Market Hypothesis? It basically says that stock (and other) prices are unforecastable, because if there were some way of forecasting them, people would use that forecast to go short or long, instantaneously driving the price to the "true" value. – Stephan Kolassa Feb 27 '20 at 12:48
  • https://www.amazon.com/s?k=financial+machine+learning&ref=nb_sb_noss_1 – Sycorax Feb 27 '20 at 13:49

0 Answers0