Forecasting future performance per customer

Deep Learning
1

Hi,

I’m trying to solve a problem similar to the Rossman challenge from Part 2.
In short, I have a list of customers with monthly stats such as revenue and other key figures.

I want to predict (yes/no) if each customer will exceed $10,000 in the next 3 months.

To do this I can think of two options

  1. Have only one row per customer with key stats such as previous months revenue etc and one label (did they reach $10,000 3 months after yes, no)
  2. Have one row per month per customer with the label per row being if the customer reached $10,000 3 months out

Which option would you recommend?

1 Like
2

I’m not exactly sure I follow #1, but I think #2 makes a decent amount of sense. I would also have other labels that give 2 months out, 1 month out etc. That just seems like a better way to answer the problem and you may see that your model is better at answering a different timeframe. Something to try at least. Good luck!

3

Thanks!

Option 1 would basically be one row per customer where I would include features for each month.

Option 2 on the other hand would have one row per month per customer

Would you suggest adding the label (passed 10k revenue) for each month out?