Research Article

Wind Pressure Coefficients Zoning Method Based on an Unsupervised Learning Algorithm

Table 1

Common clustering validity indices.

No.NameFormulaOptimal clustering numberReferences

1CH indexMaximum value of the indexCaliński and Harabasz [22]
2C IndexMinimum value of the indexHubert and Levin [23]
3Gamma indexMaximum value of the indexBaker and Hubert [24]
4DB indexMinimum value of the indexDavies and Bouldin [25]
5Hartigan indexMaximum difference between hierarchy levels of the indexHartigan [26]
6Tau indexMaximum value of the index
7Scott indexMaximum difference between hierarchy levels of the indexScott and Symons [27]
8Marriot indexMaximum value of second differences between levels of the indexMarriot [28]
9Friedman indexMaximum difference between hierarchy levels of the indexFriedman and Rubin [29]
10Rubin indexMinimum value of second differences between levels of the indexFriedman and Rubin [29]
11KL index
Maximum value of the indexKrzanowski and Lai [30]
12Silhouette index

,
Maximum value of the indexRousseeuw [31]
13Dunn index

Maximum value of the indexDunn [32]
14SD index

σ = (VAR (V1), VAR (V2), …, VAR(VP))


Dmax = max(||ck − cz||) ∀k, z ∈ {1, 2, 3, …, q}
Dmin = min(||ck − cz||) ∀k, z ∈ {1, 2, 3, …, q}
Minimum value of the indexHalkidi et al. [33]