Lets says I have 100 people that like to buy item x. They ask me to send them a message every time I have x available to sell.

Of the 100 people that like to buy item x: 1) Some people will pay more than others. 2) Others buy more frequently than others. 3) Buyer A will never buy item x if he learns that I first tried to sell it to buyer B. But buyer C, has no problem if buyer B saw it first.

I have a sequencing problem where I'm trying to find the optimal path to sell item x in order to maximize my earning per item.

The initial example only discusses item x, but I really have thousands of items that are only slightly different, but different buyers have slight preferences to these tiny difference.

I regularly build classification and regression models using scikit learn and Keras, but this is a problem where I'm not quite sure where to go.

I have a mountain of data that can be used for retrospective analysis and/or used for training a model, but not quite sure on the direction to tackle this problem.

How would I go about thinking how to solve this optimal path problem considering varying products, buyers and buyer preferences to other buyer views?

Thank you for your input.

P.S. Due to the wide scope of this problem, I'm tagging several fields.

  • $\begingroup$ It's hard to recommend something specific without more details, but in general: This is an optimization problem, so you'd have to: 1) Parameterize your space of possible actions. This would be some representation of the products to offer for sale to each buyer, in some order. 2) Define an objective function that measures how good each choice is (i.e. the thing to optimize). This would be something like expected profit, so you need a way to calculate this for each action. (continued) $\endgroup$ – user20160 Sep 18 '18 at 1:15
  • $\begingroup$ 3) Decide how to search the space of possible actions to find the best. This is a tricky issue and depends heavily on the details of the problem. Off-the-shelf optimization algorithms may or may not work. You may have to rely on approximations, heuristics, etc. to find a reasonable balance between solution quality and computational feasibility. Different formulations of the problem lead to different strategies here. $\endgroup$ – user20160 Sep 18 '18 at 1:17
  • $\begingroup$ Thank you for the response. I wasn't sure if this problem would ring a bell with someone and point me to a research paper. You're outline is great and is how I'll approach the problem. $\endgroup$ – fcol Sep 19 '18 at 15:54
  • $\begingroup$ Did you manage to solve the problem? Could you maybe post your findings as an answer? $\endgroup$ – Jan Kukacka Sep 27 '18 at 12:55

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.