3 Fuzzy Numbers What we are going to learn today ● Fuzzy numbers ● Linguistic Variables ● Intervals ● Fuzzy Arithmetic Fuzzy Numbers ● e.g.: About 90 degree, 22 years old, ... ● Fuzzy sets define on the set of real number A : ℝ [0,1] ● @LimCK Must posses the following properties: – A must be a normal set – α – The support of A must be bounded A must be a closed interval for every α ∊ (0,1] (convex!) 3 / 25 Fuzzy Numbers ● @LimCK The concept can be extended to fuzzy intervals 4 / 25 MF of Fuzzy Numbers ● ● @LimCK Not necessarily triangular/trapezoidal and symmetric. Generally: 5 / 25 MF of Fuzzy Numbers @LimCK 6 / 25 MF of Fuzzy Numbers ● @LimCK Generally, l(x) and r(x) can be piecewise function as well. 7 / 25 Linguistic Variables ● ● ● @LimCK Fuzzy numbers can be connected to linguistic concepts, e.g. very small, small, medium, … Result – linguistic variables A linguistic variable is fully characterized by a quintuple (v, T, X, g, m): – v : name of the variable – T : set of linguistic terms/values of v – X : universal set – g : syntactic rule (a grammar) for generating linguistic terms – m : semantic rule that assigns to each linguistic term 8 / 25 Linguistic Variables Linguistic variable, v = “Age” Linguistic term, T = {very young, young, old, … } Universe, X = [0, 100] @LimCK 9 / 25 Linguistic Variables g – not shown here, is the rule/grammar to generate some other terms such as “not very young”, “very very old”, … m – rules/functions to define “old”, “young”, ... e.g.: @LimCK 10 / 25 Linguistic hedges / Modifiers ● Compare: “cold” ● ● @LimCK vs “very cold” vs “fairly cold” Strong modifier reduces the truth value of a proposition, e.g: “very” Weak modifier increase the truth value of a proposition, e.g.: “fairly” 11 / 25 Linguistic hedges / Modifiers ● Commonly used class of modifiers is: α hα(a) = a ● + where α ∈ R and a ∈ [0,1] If α < 1, hα is weak modifier e.g.: “fairly” : h(a) = √a ● If α > 1, hα is strong modifier e.g.: “very” : h(a) = a @LimCK 2 12 / 25 Arithmetic Operations on Intervals [a, b] + [c, d] = [a+c, b+d] [a, b] - [c, d] = [a-d, b-c] [a,b] • [c,d] = [min(ac,ad,bc,bd) , max(ac,ad,bc,bd)] [a,b]/[c,d] = [min(a/c,a/d,b/c,b/d) , max(a/c,a/d,b/c,b/d)] @LimCK note: / is not defined if 0 ∊ [c,d] 13 / 25 Example [2, 5] + [1, 3] = [3, 8] [2,5] - [1,3] = [ -1, 4] [-1,1] • [-2, -0.5] = [-2, 2] [-1,1] / [-2, -0.5] = [-2,2] @LimCK [0,1] + [-6,5] = [-6, 6] [0,1] - [-6, 5] = [-5, 7] [3,4] • [2, 5] = [6, 20] [4,10] / [1, 2] = [2,10] 14 / 25 Arithmetic Operations on Fuzzy Numbers ● ● One of the method is based on interval arithmetic. If A and B are fuzzy numbers and ◊ is denote any of the elementary arithmetic operations: α α α (A ◊ B) = A ◊ B for α in (0,1] α (again, / is not defined if 0 ∊ B) @LimCK 15 / 25 Arithmetic Operations on Fuzzy Numbers ● ● @LimCK With alpha-cut decomposition theorem: The result of the arithmetic operation is also a fuzzy number 16 / 25 Example ( before + - * / ) Consider lower bound of A(x) Consider upper bound of A(x) Same goes to B(x). So, their alpha cuts are: @LimCK 17 / 25 Example (addition) Addition of intervals, Lower bound, Upper bound, α is in (0, 1] When α=0, x=-10 When α=1, x=5 So x in (-10, 5] @LimCK α is in (0, 1] When α=0, x=25 When α=1, x=10 So x in [10, 25) 18 / 25 Example (subtraction) Subtraction of intervals, Lower bound, Upper bound, α is in (0, 1] When α=0, x=15 When α=1, x=30 So x in (15, 30] @LimCK α is in (0, 1] When α=0, x=50 When α=1, x=35 So x in [35, 50) 19 / 25 Example (multiplication) always negative for 0 ≤ α ≤ 1 always negative for 0 ≤ α ≤ 1 always positive for 0 ≤ α ≤ 1 always positive for 0 ≤ α ≤ 1 For this specific case, since A is always negative and B is always positive, BA is entirely in negative region. Lower bound – the “most negative” → negative with largest value Upper bound – the “least negative” → negative with smallest value (30 – 10α)(5α-20) = -50α2 + 350α - 600 (10α+10)(-5-5α) = -50α2 - 100α - 50 @LimCK 20 / 25 Example (multiplication) Choose – or + ? Depends on after the operation, is α still in [0,1] @LimCK 21 / 25 Example (division) always negative for 0 ≤ α ≤ 1 always negative for 0 ≤ α ≤ 1 always positive for 0 ≤ α ≤ 1 always positive for 0 ≤ α ≤ 1 For this specific case, since A is always negative and B is always positive, B/A is entirely in negative region. Lower bound – the “most negative” → negative with largest value (30 – 10α)/(-5-5α) Upper bound – the “least negative” → negative with smallest value (10α+10)/(5α-20) [ (30 – 10α)/(-5-5α) , (10α+10)(5α-20) ] @LimCK 22 / 25 Example (division) [ (30 – 10α)/(-5-5α) , (10α+10)(5α-20) ] (30 – 10α)/(-5-5α) = x @LimCK (10α+10)/(5α-20) = x 23 / 25 Example @LimCK 24 / 25 Questions? @LimCK 25 / 25