Choquet Integral: Non-additive aggregation

Overview

Sugeno λ-measure. Output typically utility (higher value = preferred).

Strengths

• •Method-specific: Sugeno λ-measure

Limitations

• •Assumes: Singleton fuzzy densities g_i are elicited from domain experts or estimated from data
• •Assumes: Criteria exhibit interaction (superadditive if Σg_i<1, subadditive if Σg_i>1, additive if Σg_i=1)
• •Assumes: All decision matrix entries are real-valued, finite, non-missing

Assumptions to verify

• •Singleton fuzzy densities g_i are elicited from domain experts or estimated from data
• •Criteria exhibit interaction (superadditive if Σg_i<1, subadditive if Σg_i>1, additive if Σg_i=1)
• •All decision matrix entries are real-valued, finite, non-missing

When not to use

• •Criteria are independent → use SAW/WPM (Choquet collapses to additive weighted sum anyway)
• •Fuzzy densities g_i cannot be estimated: method becomes unidentifiable

Edge cases

• •ties g_i = g_λ({x_i}) for each singleton criterion.

Common pitfalls

• •Hatalı: 'CHOQUET-INTEGRAL bu varsayımı kontrol etmeden uygulamak'. Doğrusu: Singleton fuzzy densities g_i are elicited from domain experts or estimated from data
• •Hatalı: 'CHOQUET-INTEGRAL bu varsayımı kontrol etmeden uygulamak'. Doğrusu: Criteria exhibit interaction (superadditive if Σg_i<1, subadditive if Σg_i>1, additive if Σg_i=1)
• •Hatalı: 'CHOQUET-INTEGRAL bu varsayımı kontrol etmeden uygulamak'. Doğrusu: All decision matrix entries are real-valued, finite, non-missing
• •Hatalı: CHOQUET-INTEGRAL'yi 'Criteria are independent → use SAW/WPM (Choquet collapses to additive weighted sum anyway)' durumunda kullanmak: recommendation_metadata.not_recommended_when alternatif öneriyor.
• •Hatalı: CHOQUET-INTEGRAL'yi 'Fuzzy densities g_i cannot be estimated' durumunda kullanmak: recommendation_metadata.not_recommended_when alternatif öneriyor.

Worked example

1. 1.Adım 1 (F1): Step 1: Specify fuzzy densities g_i = g_λ({x_i}) for each singleton criterion. Formül: g_{i} = g_{\lambda}(\{x_{i}\}),\ i=1,\ldots,n Anchor: Sugeno 1974, p.15
2. 2.Adım 2 (F2): Step 2: Solve 1 + λ = Π (1 + λ g_i) for the Sugeno λ-parameter. Formül: 1 + \lambda = \prod_{i=1}^{n}(1 + \lambda g_{i}),\quad -1\le \lambda < \infty Anchor: Sugeno 1974, p.16 Eq.(2)
3. 3.Adım 3 (F3): Step 3: Build λ-fuzzy measure g_λ on every subset. Formül: g_{\lambda}(H_{i}) = \dfrac{1}{\lambda}\Big[\prod_{x\in H_{i}}(1+\lambda g_{x}) - 1\Big] Anchor: Sugeno 1974, p.16 Eq.(3)
4. 4.Adım 4 (F4): Step 4: Order h(x_1)≥…≥h(x_n) and compute Choquet integral. Formül: \int h\,dg = \sum_{i=1}^{n}\big[h(x_{i}) - h(x_{i-1})\big]\cdot g_{\lambda}(H_{i}),\quad h(x_{0})=0 Anchor: Murofushi-Sugeno 1989, p.5 Eq.(1)

How to cite