European Proceedings Logo

Greedy Heuristics for the Choice of the Radius of Local Concentrations in FOREL-2 Algorithm

Table 1: Accuracy of clustering 1, 3, 8, and 9 batches of IСs 140UD25A data

k-means FOREL-2
time, sec. 1 10 60 1 10 60
Mixed batch (n=201) with normalization by standard deviation
Rand index
max 0.605 0.605 0.605 0.794 0.778 0.799
min 0.55 0.542 0.563 0.605 0.641 0.656
mean 0.596 0.593 0.591 0.712 0.711 0.741
st.dev 0.01 0.014 0.01 0.052 0.039 0.04
Objective function
max 85.1 80.6 77.2 85 81 78.4
min 66.9 66.8 66.8 76.1 72.5 72.2
mean 69.4 69 69.7 81.1 77.1 75.1
st.dev 3.5 2.9 3.2 2.5 2 1.3
var 0.051 0.041 0.046 0.03 0.026 0.017
spn 18.2 13.8 10.4 9 8.5 6.2
Mixed batch (n=807) with standard deviation normalization
Rand index
max 0.6 0.599 0.6 0.76 0.777 0.781
min 0.557 0.58 0.581 0.539 0.599 0.67
mean 0.584 0.594 0.595 0.669 0.718 0.738
st.dev 0.013 0.004 0.004 0.063 0.045 0.026
Objective function
max 326.5 296.9 298.4 411.6 348.2 336.7
min 273 269.7 269.8 325 323.5 312.7
mean 296.5 274 273.8 365 335.4 324.3
st.dev 14.6 6 5.9 21.1 7 5.7
var 0.049 0.022 0.022 0.058 0.021 0.018
spn 53.5 27.3 28.6 86.5 24.8 24
< Back to article