MCDM | Tunny docs

A set of ranking methods for selecting "overall superior trials" from multi-objective optimization results. Tunny Dashboard provides the following 3 methods plus the Entropy Weight Method for objectively determining weights.

Method List

Method	Characteristic	Score range	Cost
TOPSIS	Ranking by distance ratio from ideal solution	[0, 1]	O(m × n)
VIKOR	Compromise solution balancing utility and regret	[0, 1]	O(m × n)
PROMETHEE	Ranking by pairwise preference comparison	Φnet ∈ [-1, 1]	O(m² × n)
Entropy Weight	Objectively compute weights from data variance	—	O(m × n)

Method Summaries

TOPSIS

Scores each trial by how close it is to the "ideal solution (positive ideal solution)". Computes: normalization → weighting → Euclidean distance from ideal/anti-ideal solutions → relative closeness. A score closer to 1 means closer to the ideal solution. Intuitive and fast to compute.

VIKOR

Measures the "gap from ideal" for each trial using two distances — L1 (Manhattan) and L∞ (Chebyshev) — and adjusts the balance with parameter v to find a compromise solution. v > 0.5 emphasizes overall utility; v < 0.5 minimizes worst-case regret.

PROMETHEE I / II

Evaluates "how much is a preferred over b?" for every pair of trials using a preference function, and aggregates the results into flows (Φ+/Φ−/Φnet). PROMETHEE I is conceptually a partial ranking (allowing incomparable pairs); Tunny Dashboard displays a total order (Φ+ descending, tiebreak Φ- ascending) and reports per-trial incomparability counts (⇹N in the chart). PROMETHEE II provides a complete ranking. Intuitive due to pairwise comparison, but note the O(m²) computational cost.

Entropy Weight Method

Automatically computes the weight of each objective function from the variance in the data itself. Objectives with greater spread are considered to have higher "discriminatory power" and are assigned larger weights. Can be used as weight input for the 3 methods above.

How to Choose

flowchart TD
    w{"How should weights be set?"}
    w -- "Objectively, from data" --> wA["Entropy Weight Method<br/>(auto-compute weights)"]
    w -- "Manually" --> wB["Set weights with manual sliders"]

    r{"Which ranking method?"}
    r -- "Fast, intuitive score" --> rA["TOPSIS<br/>(distance from ideal solution, [0,1] score)"]
    r -- "Consider both overall balance and worst case" --> rB["VIKOR<br/>(adjust utility/regret balance with v parameter)"]
    r -- "Want detailed pairwise preference relationships" --> rC{"Number of trials"}
    rC -- "Small (<10,000)" --> rC1["PROMETHEE I/II"]
    rC -- "Large (>10,000)" --> rC2["Note computation time"]

Recommended Combinations

Scenario	Recommendation
Try first	TOPSIS + equal weights
Large scale differences between objectives	TOPSIS/VIKOR + Entropy weights
Minimize worst-case for a specific objective	VIKOR (v < 0.5)
Detailed pairwise comparison between trials	PROMETHEE I + PROMETHEE II
Want objective basis for weights	Entropy Weight Method + any method