Question about data markup and prediction of this field in time series

[pskucherov]

Hi!
I want to mark up price data so that it predicts not the price value itself, but by what percentage the price will change in the future relative to the current situation.

Initially, I thought of solving this problem using long term forecasting, but it is obviously not suitable for marked up data, since it is this markup that needs to be predicted and it will not be in the real data.

It looks like a classification problem, but it has a time series and time dependence, and this confuses me. And maybe I should use Imputation?

What do you think is the best way to solve this problem?

[wuhaixu2016]

Hi, I think the most reliable solution for your question is just predicting the future time series but further calculating the change percentage based on your time series prediction.

Note that if you model this problem as classification, you may need to handle 100 classes (depends on the fineness of the categorizes), which is hard for current techniques.