1
$\begingroup$

I have a dataset that looks like,

order datetime, customer id, product name, type of product, quantity sold

I want a model to forecast sales for each individual item. I'm thinking of using one of the following but would like some advice, literature, or other options.

  1. One time series forecast (ARIMA) for each product
  2. RNN
  3. HMM

What's the best method to predict sales of each individual item into the future?

$\endgroup$
2
$\begingroup$

Do you want to forecast sales by day over a series of days? That would be more like ARIMA. Do you want to forecast "How many widgets will we sell in the next month"? That would be more like a regression problem.

As an aside, if the retailer you're working with has a large assortment of products that changes over time (e.g. seasonal clothing) then you should also look into forecasting by product category and not an individual product. It will be much more accurate.

| improve this answer | |
$\endgroup$
  • $\begingroup$ The differentiation between ARIMA and Regression is helpful advice. Products don't change very often. For the ARIMA route, should I make 1 model per product name or is there some method to do multivariate, categorical ARIMA? The Regression route makes sense to me. I can make one model that uses the product name as a categorical variable. $\endgroup$ – 0111001101110000 Jul 18 '17 at 16:35
  • $\begingroup$ I have not heard of a multiple output ARIMA, that's actually kind of uncommon and normally something only neural networks do. Normally you have multiple inputs to one output. $\endgroup$ – CalZ Jul 18 '17 at 17:22

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.