I need to predict cost for the next 4 weeks. Along with categorical variables (available for future as well), I have a numerical variable (a value from the day before prediction - I don't have future values for this variable), let's call it foo, that is telling me if I am spending too much or too little. Basically, it tells me if I need to increase or decrease the cost in the following days. I am not sure how to approach this. How to incorporate this variable in the model.

I thought that I can I maybe declare a categorical variable for future days that would have value 'grow' for the time that it is relevant (for example first 2 weeks), and for the rest value 'na' (last 2 weeks), since I don't know the trend of cost for them? But then how would I create this variable for the past...

Maybe I just need a multiplier, to raise my cost, or lower them depending on foo value from the day before.

Do you have any suggestions?
PS: I am planning to use LSTM, but I am open to other suggestions.

  • $\begingroup$ Why are you not just simply trying to predict the cost in the next month using regression? $\endgroup$
    – Taylrl
    Nov 19 '20 at 12:28
  • $\begingroup$ @Taylrl I will start with regression, but I also want to do multivariate LSTM and see if it will outperform the regression. Do you have any suggestions on how to use this foo variable? $\endgroup$
    – GileBrt
    Nov 19 '20 at 12:53
  • $\begingroup$ I would just include it as another categorical variable $\endgroup$
    – Taylrl
    Nov 19 '20 at 12:55
  • $\begingroup$ What is foo and why it is not available for the future? I hope it is not calculated based on cost. I mean a source of data leakage $\endgroup$
    – 10xAI
    Nov 19 '20 at 16:36

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.