If you need to make decisions based on data that is not known or uncertain (think of the fuel price, wind velocity, customer demand), a typical approach is to apply a margin of error to the decision, or to consider a few cases of the uncertainty to come to a balanced decision. Although this approach can be very helpful, it is very dependent on the skills of the decision maker or the completeness of the data at hand. Even personal emotion can play a large role when selecting cases or adding a margin of error to your decisions (do you believe the data, do you believe the decisions following the data?).
Increased complexity and data inaccuracy
If the complexity of your business increases, it can become very time consuming and impractical to consider the various cases. Moreover, even when the data is only slightly inaccurate or just different from what is considered, the decisions can result in business infeasibilities or disturbances.
Techniques are available and integrated in AIMMS
If you face such problems it is good to realize that various optimization modeling techniques are available that can help you incorporate data uncertainty.
- Scenario Analysis: considers, instead of a few cases, a large series of cases independently (e.g. various demand scenarios, production rates, etc.) after which the decisions for all cases are analyzed (average, distribution etc.) and used to make a final decision.
- Stochastic Programming: considers a representative set of scenarios at once and provides one feasible solution that will hold under all scenarios. Typically, each scenario has a probability and represents the uncertainty of the model.
- Robust Optimization: considers, instead of explicit scenarios, data uncertainties against whose realizations the solution is required to remain feasible. This uncertainty may occur in any part of the model data (e.g. demand is between 10 MW-12MW). Partial feasibility can be included by adding probabilities to constraints (e.g. the chance that demand is met is at least 95%).
The AIMMS modeling platform offers these techniques as standard components making it possible to integrate the reality of uncertainty in your business models and thus improve your decisions.
We can help you
Including uncertainty in your decision process is not necessarily easy, but we can help. Contact our team of experts and learn more about uncertainty modeling in your business using AIMMS: email@example.com.