From Topic Networks to Distributed Cognitive Maps: Zipfian Topic Universes in the Area of Volunteered Geographic Information

Table 6

F-scores of classifying TTNs into seven target classes (CITIES, REGIONS, OTHERS, WP-REGIO-1, WP-REGIO-2, WP-OTHERS-1, and WP-OTHERS-2) by means of SVMs using RBF kernels.

	Measure	all	opt	ext	B1	B3 all	B3 opt	B4 all	B4 opt
1	GES	0.653	0.753	0.798	0.143	0.130	0.286	0.121	0.213
2	WAL	0.649	0.751	0.788	0.143	0.130	0.286	0.109	0.216
3	VEO	0.677	0.773	0.816	0.143	0.130	0.286	0.120	0.221
4	wges	0.559	0.620	0.650	0.143	0.130	0.286	0.120	0.199
5		0.638	0.722	0.764	0.143	0.130	0.286	0.119	0.211
6		0.729	0.768	0.853	0.143	0.130	0.286	0.125	0.223
7		0.694	0.766	0.832	0.143	0.130	0.286	0.127	0.229
8		0.642	0.681	0.717	0.143	0.130	0.286	0.122	0.212
9		0.742	0.773	0.790	0.143	0.130	0.286	0.102	0.156
10	NetSimile	0.479	0.629	0.722	0.143	0.130	0.286	0.127	0.229
11	ToSi	0.390	0.433	0.465	0.143	0.130	0.286	0.108	0.229

Column “all”: F-scores, if all features are used by the similarity measure (row). Column “opt”: F-scores, if a subset of features selected by the genetic search is used. Column “ext”: F-scores, if a subset of features selected by the extended genetic search is used. The last five columns display the F-scores of the random baselines B1, B3, and B4, in the case of B3 and B4 differentiated for the variants all and opt.