0
$\begingroup$

I'm trying to make a forecasting model for goods prices in an economy (trying to forecast inflation).

Dataset: has 300 goods prices % monthly variations for last 6 years. And also added $n$ macroeconomic variables time series.

Desired model:

  • Input: Last 12 variation of 300 goods prices + 12 variations of $n$ macroeconomic variables
  • Output: Next variation for 300 goods prices (only one step).

I was thinking about using a LSTM model, but I don't know if the dimensionality will be a problem or if there are better models or some kind of recommended data pre processing.

$\endgroup$
1
  • $\begingroup$ I have no experience in economic timeseries data. In business GBDTs can work quite well when it comes to tasks, such as sales forecasts, which share some similarities. This might be an option, too. However, I suggest to do some desk research on relevant papers to see what's being used there. $\endgroup$
    – Sammy
    Jul 28 at 16:20
-1
$\begingroup$

Have you tried XGBoost?

It used to have very good result in multi dimensional data including times series.

It was mainly used to retail forecasting, and it could be applied to your case:

https://towardsdatascience.com/machine-learning-for-store-demand-forecasting-and-inventory-optimization-part-1-xgboost-vs-9952d8303b48

$\endgroup$
1
  • $\begingroup$ Why downvoting? XGBoost has been efficient in many high dimensional fields, including good prices. I'd like to understand what is wrong with this proposal. $\endgroup$ Jul 30 at 7:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.