Summary
This new feature addresses the infeasibility problem caused by data. The key function is to generate insightful messages for the end-users as well as additional information which may be used in the WebUI. The app developers can embed this feature in an application by setting the main function to run behind an action in the WebUI and by using parameter annotations to highlight suspicious data.
What problem is addressed
In mathematical optimization, even the models which are correctly formulated can face the challenge of infeasibility caused by incorrect input data. For app developers, AIMMS offers already for a long-time several effective tools for dealing with the model infeasibility. However, for end-users, this situation could lead until recently to frustration and uncertainty, with questions like, “Why is the model infeasible?” and “What is wrong with my input data?”
What does the new feature do
To address this problem, AIMMS introduces a new feature which may be used by the app developers to allow even the end-users to identify potentially incorrect input data and this way, help in resolving infeasibility issues. The focus is on empowering end-users to identify and correct problematic data in a systematic, user-friendly manner, ensuring their optimization models achieve feasible solutions efficiently.
How does it work
By leveraging advanced techniques such as:
- calculating Irreducible Inconsistent Subsystems (IIS) or solving feasibility problems,
- analysing constraints while isolating those with changeable parameters, and
- using a new mechanism to pinpoint parameters influencing constraints,
the new AIMMS function can deliver actionable insights directly within the Web User Interface (WebUI).
On the one hand, a structured message about the suspicious input data is generated and may be displayed for end-user’s information directly in the WebUI. On the other hand, the new function assigns suspicion levels (High, Medium, or Low) to relevant parameters. This information may be further used to highlight in the WebUI widgets potentially incorrect values by using intuitive color-coded visual cues (e.g., ranging from light pink to deep red).
An illustrative example can be found here.
What may the end-users expect
After adjusting the highlighted suspicious values to more realistic numbers, the end-users can re-solve the model and observe the impact. If infeasibility persists, the function can be re-invoked to uncover further insights. This iterative process helps addressing the potential issues with input data, eventually leading to a feasible and reliable model.