Can the core contain several values?

Yes. A trapezoidal fuzzy number can have an interval of membership-one values.

Are fuzzy numbers always triangular?

No. Many shapes meet the chosen definition; triangular forms are just convenient examples.

What is an alpha-cut?

For positive alpha, it is the ordinary set of values whose membership is at least alpha. For a fuzzy number, that set is typically a closed interval.

AI Glossary / Graphs and symbolic AI / Intermediate

Fuzzy number

Definition

A fuzzy number is a fuzzy set on the real line that represents a numerical quantity through graded membership. Under a common definition, it is normalized, fuzzy-convex and upper semicontinuous, with compact support; its positive alpha-cuts are nested closed intervals.

What it means

A fuzzy number can express a description such as approximately ten without selecting one exact value. Values near the center may fit the description strongly, while more distant values fit weakly or not at all.

Normalized means that at least one value has membership one. Fuzzy-convex means that the values between two members fit at least as strongly as the smaller of their membership grades. It does not mean the graph of the membership function is a convex mathematical function.

Triangular and trapezoidal shapes are convenient examples. A fuzzy number is not simply a probability distribution drawn with an unusual label: membership grades need not integrate to one, and a grade does not automatically represent a frequency or probability.

How it works

Specify a membership function over numerical values. A triangular number is often written (a, b, c), with a below b below c: membership rises linearly from zero at a to one at b and falls linearly to zero at c.

For positive alpha, its alpha-cut is [a + alpha(b - a), c - alpha(c - b)]. Increasing alpha selects a narrower interval of values fitting the description more strongly. The closed support is [a, c], including its boundary points even though their memberships are zero in the nondegenerate triangle.

Arithmetic requires a specified extension of ordinary operations. Standard noninteractive, minimum-based fuzzy arithmetic can operate on these intervals at matching alpha levels. Relationships between quantities matter: treating repeated occurrences of the same uncertain quantity as independent can create spurious spread.

Three practical examples

A project-estimation exercise records approximately ten hours as a declared fuzzy description rather than presenting ten as an exact duration.
A material-modeling study represents an imprecisely specified positive stiffness using a documented fuzzy number.
A design comparison expresses approximate numerical preferences and examines how their graded ranges influence the result.

Think of it this way

Think of a target on a ruler with a central region that best matches a description and surrounding values that match progressively less well. The ruler remains precise; the description has gradual boundaries.

The analogy does not imply that a dart will land according to those grades. A fuzzy number describes compatibility with a numerical concept or specification, not automatically a random-generating process.

Worked example

Adding two triangular descriptions

Let an invented preparation-time description be A = (8, 10, 14) hours. At alpha = 0.5, its cut is [8 + 0.5 × 2, 14 - 0.5 × 4] = [9, 12]. A second duration is B = (2, 3, 5), whose corresponding cut is [2.5, 4].

Assume the standard minimum-based extension with no additional interaction restricting the two durations. Interval addition gives the sum's 0.5-cut as [9 + 2.5, 12 + 4] = [11.5, 16]. The resulting triangular sum is (10, 13, 19), which gives the same cut.

This calculation does not say there is a 50% chance that the total lies between 11.5 and 16 hours. It follows the selected membership arithmetic. If the two durations share a constraint that prevents their largest values from occurring together, that relationship must be represented before calculating the total.

Adding triangles this way preserves triangular shape, but nonlinear operations such as multiplication generally need not.

Illustrative case study

The fictional Birch Planning team records several approximate task durations as triangular fuzzy numbers. Its report initially converts every total to a single center value, making the output look much more precise than the inputs.

Reviewers ask to see selected cuts and the assumptions behind combining tasks. They find that setup and review share the same staffing condition, so treating their uncertainties as unrelated is questionable. The team documents this relationship and investigates an appropriate joint representation.

Birch keeps the graded descriptions visible when comparing schedules and avoids labeling them confidence intervals. It also compares the approach with ordinary scenario planning. The case demonstrates transparent estimation assumptions, without claiming that fuzzy durations predict real completion times better.

Common misconceptions and mistakes

A triangular fuzzy number is a triangular probability distribution. The shapes can resemble each other, but their heights and interpretations differ.
Membership one means the value is known with certainty. It means full compatibility under the chosen membership function.
Every fuzzy set on the real line is a fuzzy number. The selected definition imposes additional shape and regularity conditions.
Subtracting a fuzzy quantity from itself should be handled as two unrelated inputs. That loses the identity relationship; the same actual quantity minus itself is zero.

Business use

Fuzzy numbers can make approximate numerical judgments explicit when a graded representation is meaningful. Their usefulness depends on a clear explanation of how the shape, endpoints and core were chosen.

Ask whether the inputs reflect expert judgments, engineering specifications or another information source. Distinguish that evidence from the arithmetic used to combine them. When a final decision needs one number, document how it was selected and what information was discarded. A convenient triangular shape should not be mistaken for a validated description of real variation.

Check your understanding

For A = (8, 10, 14), is the membership-one cut [8, 14]?

Answer: No. The membership-one cut is [10, 10], containing the triangle's peak. The closed support is [8, 14].

Related terms

Defuzzification
Fuzzy set
triangular membership function
alpha-cut
fuzzy arithmetic
extension principle
interval arithmetic
uncertainty representation

Learn next

Calculate the 0.25-cut and 0.75-cut of A in the worked example. Check that the higher-level interval is contained in the lower-level interval, then explain what information would be needed to assign a probability to either interval.

Sources

A primary arithmetic paper states fuzzy-number properties and examines how extension choices affect operations: https://doi.org/10.1016/j.ijar.2019.01.005

MathWorks documents the triangular membership calculation: https://www.mathworks.com/help/fuzzy/trimf.html

A research article discusses fuzzy-number definitions and level-set conditions: https://link.springer.com/article/10.1155/2009/395714

The duration arithmetic and Birch Planning case are original illustrations using explicitly stated combination assumptions.

Previous: Fuzzy measure theory Next: Fuzzy relation