I'm wondering is there any time series research area about the change of forecasting with added items.

For example, in the prototype below, when we check each recommended item, the forecast about sales will be updated.

I know how to do time series forecasting. But for forecasting with added items, I'm looking for more advanced research methods.

Do you know which research direction I can go?


  • $\begingroup$ I am not aware of any published research in this direction and would be very interested. The problem is that this gets complicated very quickly. You need all sorts of timestamped data (when something was recommended, whether it was selected, whether it was subsequently bought, which forecast horizon are we looking at?). $\endgroup$ – Stephan Kolassa Jun 9 '20 at 7:33
  • $\begingroup$ Thank you @StephanKolassa! So I'm planning to forecast next 1 week at most. The training data will be the latest 2 weeks historical data. $\endgroup$ – Cherry Wu Jun 9 '20 at 18:16
  • $\begingroup$ How does this differ from say ARIMA forecasts with regressors? (Obviously there are many approaches to regression in time series data, I picked one). $\endgroup$ – user54285 Jun 11 '20 at 21:53
  • $\begingroup$ @user54285 Sounds like a good inspiration. So in order to do this, for example, if I want to forecast the total sales, and predictors used in the forecasting regressor can be sales $$ or purchase amount of selected products? Is that on the right track? $\endgroup$ – Cherry Wu Jun 12 '20 at 5:47

Your Answer

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

Browse other questions tagged or ask your own question.