AI Glossary / AI and machine learning / Advanced
Dominance-based rough set approach
Definition
The dominance-based rough set approach, or DRSA, analyses decisions involving ordered criteria and ordered outcome classes. It uses dominance relationships and rough approximations to describe consistent patterns and uncertainty in preference-based decisions.
What it means
Some decisions have a natural direction. A supplier with better reliability and no worse cost would normally not receive a worse rating solely because of those criteria. DRSA is designed to represent this kind of preference order.
One option dominates another when it is at least as good on all the criteria being considered. The comparison respects each criterion's direction: lower cost may be preferable, while higher reliability may be preferable. Dominance does not require assigning a single weighted score first.
Real decision records can violate the intended order. Two otherwise comparable cases may receive inconsistent ratings because important information is missing, assessors disagree or the policy changed. Rough approximations distinguish conclusions supported consistently from cases that remain uncertain.
The method describes patterns in the information and preferences supplied. It does not prove that the original decisions were fair, sensible or complete. Those are questions for the people responsible for the decision process.
How it works
The analyst defines the objects, ordered decision classes and criteria, including which direction counts as better. Examples could be suppliers classified as poor, acceptable or preferred using delivery and service measurements.
Dominance comparisons identify which objects are no worse than others on the selected criteria. The analysis often considers unions such as at least acceptable or at most acceptable, rather than treating ordered classes as unrelated labels.
Lower and upper approximations capture different levels of support. Broadly, one describes cases securely included under the chosen consistency conditions, while the other includes cases that may belong. Boundary cases require care rather than forced certainty.
Rules can then express relationships between criteria and decision categories. Variants relax strict consistency or accommodate additional modeling requirements. The analyst should state the chosen variant and inspect the examples supporting each rule.
Three practical examples
- Supplier review: examine whether historical ratings respect agreed directions for reliability, delivery time and service quality.
- Product prioritization: explore rules connecting ordered quality assessments with ordered development priorities.
- Service assessment: identify records where improved performance measures nevertheless coincide with a worse assigned service category.
Think of it this way
Think of sorting climbing routes into easy, moderate and difficult using ordered measures. If a route is no harder on every agreed measure, assigning it a higher difficulty category deserves investigation.
The analogy has a limit. The apparent contradiction may disappear when an omitted factor, such as exposure to wind, is included. A dominance comparison can only use the criteria actually represented.
Worked example
Finding a rating that deserves explanation
Suppose two fictional suppliers are assessed on on-time delivery and complaint rate. Higher delivery reliability is better, and a lower complaint rate is better.
Supplier A records 96% on-time delivery and a 1% complaint rate. Supplier B records 92% and 3%. On these two measures, A dominates B. If A is rated acceptable while B is rated preferred, the recorded preference order conflicts with those criteria.
The analyst does not automatically upgrade A or downgrade B. Perhaps B offers a critical service that was omitted from the table. Perhaps the ratings were assigned at different dates. Alternatively, the assessment process may be inconsistent.
The percentages are invented. The exercise shows how an ordered comparison can expose questions for review without pretending that two measurements contain the entire business decision.
Illustrative case study
A fictional procurement team uses DRSA to examine past supplier classifications. The first analysis finds numerous boundary cases that appear inconsistent with its written assessment criteria.
Interviews reveal that some buyers used unofficial delivery-flexibility information. The team decides whether that factor belongs in the formal assessment and clarifies how it should be measured. It then reruns the analysis on comparable records.
The invented lesson is that uncertainty can reveal a documentation problem rather than a defective algorithm. Making the decision basis explicit is often more valuable than obtaining a confident-looking rule immediately.
Common misconceptions and mistakes
- Dominance means best overall in every possible sense. It applies only to the selected criteria and their stated preference directions.
- The method eliminates judgment. People still choose criteria, classes, data and how recommendations will be used.
- Inconsistent examples should simply be deleted. They may reveal missing information, policy changes or important exceptions that need investigation.
Business use
DRSA can help analyze preference-based classification when interpretability and ordered criteria matter. It is most relevant to structured decision-support work rather than ordinary chatbot use.
Start by checking whether the assumed ordering makes business sense. Inspect rule coverage and uncertain cases, and keep the responsible decision-maker involved when a historical pattern is converted into a future recommendation.
People also ask
Is it ordinary rough set theory?
It extends the framework to handle dominance and preference order rather than relying only on indiscernibility.
Must all criteria be maximized?
No. Each criterion needs an explicit direction, such as lower cost being preferable.
Does a decision rule establish causation?
No. A rule summarizes relationships in the represented decision data and assumptions.
Check your understanding
One supplier is better on every recorded criterion but receives a worse rating. What should an analyst investigate before changing the rating?
Answer: Omitted criteria, differences in timing, data errors and changes in policy or assessor judgment. The apparent inconsistency does not identify its cause by itself.
Related terms
- rough set
- dominance relation
- ordinal classification
- decision rule
- multi-criteria decision analysis
Learn next
Review the directions and definitions of three criteria in a familiar assessment process. Check whether all assessors apply the same ordering before selecting a modeling method.
More in AI and machine learning
- Divisive clustering - Divisive clustering is a top-down approach to grouping data: start with everything in one cluster and repeatedly split the least coherent group until each...
- Document classification - Document classification assigns one or more category labels to a document based on its content or other relevant information.
- Domain adaptation - Domain adaptation adjusts a machine-learning system so that knowledge learned from one data setting can be useful in another.
- Double delta feature - A double delta feature is the change of a change: it measures how quickly a measurement is speeding up or slowing down, rather than how fast it is simply...
- DPLL algorithm - The DPLL algorithm is a complete search procedure for deciding whether a propositional-logic formula in conjunctive normal form can be satisfied.
- Dynamic epistemic logic - Dynamic epistemic logic studies how knowledge and beliefs change when information is communicated or events are observed.
Sources
Błaszczyński, Greco, Matarazzo and Szeląg, Dominance-based Rough Set Approach, basic ideas and main trends: https://arxiv.org/abs/2210.03233. Checked 12 October 2026. Supplier scenarios are original teaching examples.
Previous: Domain (spatial) operation Next: Dominant resource fairness