0

consider an e-commerce website which daily updates their transaction data. I take interest in predicting the revenue an account generates over a certain time range, say until the fifth year after the first purchase. What I'd like to have is a method which daily updates the predicted revenue at the end of the fifth year. This requires in particular:

  • if there had been a purchase on the day before, the predicted value must increase
  • if there was not purchase on the day before, the predicted value should fall slightly, but never below the already realized revenue

Do you have any suggestion what kind of model could achieve this?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
MaHo
  • 391
  • 2
  • 11
  • You can use time series neural network for this purpose. look at below link:http://stats.stackexchange.com/questions/10162/how-to-apply-neural-network-to-time-series-forecasting – Panda Jan 06 '17 at 11:03
  • Although this particular question is not what I am looking for, there are some interesting links in this thread - thank you! I was already suspecting neural networks could help me out, but my knowledge in this area is limited. Time to do something about it ;-) – MaHo Jan 06 '17 at 11:49

0 Answers0