I am pondering if a machine learning problem might be better formulated as a reinforcement or control problem. I have no experience with the latter, so bear with me.

Let's say I am organising a tour and each time the tour has a different number of participants. My responsibility is to make sure that there is enough water for everybody during the tour. I know historically how much water is used per person per tour. So that is a good rule of the thumb.

But it is also known that the morning tours needs more water than the evening tours. And that there are seasonal differences.

What I would like to build is a mechanism that takes the average water usage per person per tour as a starting point. And then per morning and evening tour starts adjusting the average amount of water to be taken on the tour. The goal is to have enough water on 95% of the tours.

What kind of algorithm is this? I would be interested in the name of the class of the problem (stochastic discrete time continuous state control?) or the name of an applicable algorithm.

  • 2
    $\begingroup$ In the absence of some sort of constraint on procurement of water, it's just a statistical estimation problem. Are there constraints on how much water you can procure, how often you can procure it, how much you can store, or anything else related to supply? Do you incur discounts (or penalties) for larger orders, pay for storage, suffer pilferage, ...? $\endgroup$
    – prubin
    May 22, 2021 at 19:04
  • $\begingroup$ Thanks for your comment, it made me think about the question more thoroughly. Indeed without any constraints a time series forecasting projecting a 95% interval could solve the requirements. $\endgroup$
    – spdrnl
    May 24, 2021 at 13:04


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.