ETF5248 Chap.7 Model Selection and Exam Review
Model Selection and Exam Review
The final review is about choosing and defending a model, not attaching a memorised formula to a keyword.
Divisible activity with linear relationships suggests linear programming; movement over nodes and arcs suggests network flow; indivisibility or logic suggests integer programming; changing marginal rates suggest piecewise linearity; and valid curvature can justify convex optimisation.
This chapter combines the course into a repeatable response architecture: define the decision, state the objective, build named constraint families, specify domains, solve, audit and communicate.
It also separates published assessment facts from unknown examination details, uses the mid-semester practice set as skills practice rather than a format prediction, and closes with timed drills that require model selection before calculation.
A strong response is self-contained.
Its variable definitions include indices, units and domains; its objective states a direction; and its constraints are grouped and named so the reader knows what each family protects. Hybrid decisions may combine structures, such as binary depot openings with continuous network shipments.
After solving, the audit changes with the family: test every linear row, reconcile nodes, substitute both binary states, inspect segment breakpoints or establish convexity. The conclusion then gives the action, objective value, binding resource or logical explanation, and one limitation. Review should therefore alternate classification and calculation rather than repeat one topic in a block.
The published mid-semester practice set is useful for diagnostic work, but it explicitly does not predict test length. Final-examination duration, sections, question types and permitted materials are not stated in the acquired unit sources, so revision should cover the learning content while official instructions remain the authority for logistics.
What this chapter covers
- 01
Business cues and model families
- 02
Variable, objective and constraint workflow
- 03
Domain and logic checks
- 04
Independent solver-output audit
- 05
Sensitivity and limitation statements
- 06
Timed mixed-model review
Worked example · free
Select a model for a facility-routing decision
- 1The depot choices are discrete and the shipments are flows, so use a mixed-integer facility-location model.
- 1Define binary y for each depot opening and non-negative shipment x for every allowed depot-customer arc.
- 1Minimise fixed opening costs plus per-unit shipping costs across all arcs.
- 1Meet each customer demand and limit each depot's total shipment by its capacity multiplied by its opening binary.
- 1Audit demand totals, depot capacity, closed-depot flow, integrality, objective reconstruction and the reported gap.
Key terms
- Model family
- A class of formulations sharing a decision structure, such as network flow or integer programming.
- Validation
- Independent checking that a formulation and returned solution match the original business rules.
- Solver status
- The software report describing whether a solution is feasible, optimal, bounded or otherwise terminated.
- Limitation
- An assumption or omitted feature that constrains how the model result should be used.
- Managerial interpretation
- A plain-language statement of the decision, value, resource logic and conditions attached to a model result.
Model Selection and Exam Review FAQ
What model-selection cues are most reliable?
Use the underlying decision structure rather than one isolated word. Continuous divisible activity with constant rates is linear, arcs and conservation indicate networks, yes-or-no or count decisions require integers, changing marginal rates suggest piecewise segments, and smooth curvature requires a convexity check before claiming global optimality.
What is a complete formulation answer?
It defines every variable with index, unit and domain; states an objective with direction; gives named constraint families with index ranges; includes necessary boundary or linking conditions; and explains the returned decision. A list of equations without definitions or domains is not self-contained.
How should I prepare when final-exam format details are absent?
Prepare the published learning content and model workflow without inventing timing, sections, question styles or permitted aids. Use official examination instructions when they become available. Until then, practise mixed problems under several reasonable time limits and focus on producing concise, auditable formulations.
Can the mid-semester practice set predict the final examination?
No. It is useful evidence about skills and early topics, but its own instructions say it is not a mock test and is not indicative of test length. Treat it as a diagnostic practice resource, not as a blueprint for the final examination's format or coverage.
Exam move
Build a one-page selection matrix with columns for story cue, variable type, objective, signature constraints and validation check. Shuffle problems from different topics so you must classify each one before calculating. Use a fixed seven-step answer frame: define, objective, constraints, domains, solve, audit and recommend. After every timed attempt, mark modelling and arithmetic separately.
Rework any problem whose answer was numerically correct but poorly justified, because ETF5248 requires the model and interpretation as well as computation. Keep unknown examination logistics separate from content revision and rely only on official instructions for those details.
Working through Model Selection and Exam Review in ETF5248? Sia is AskSia’s AI Mathematics tutor — ask any ETF5248 Model Selection and Exam Review question and get a clear, step-by-step explanation grounded in how ETF5248 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
