Skip to main content
. 2014 Dec 15;4:7476. doi: 10.1038/srep07476

Figure 8. The flowchart of the hybrid method.

Figure 8

The hybrid method combines two-stage SVM with K-means to identify the structural domains. At first, it extracts the features for modelling. Then, it trains two-stage SVM model. At the two-stage SVM, the features Den, Dis and NN represent the density, dispersion and the number of nodes. The numbers (1, 2, 3) in Den2a, Dis2a, NN2a, Den3a, Dis3a and so on are the number of segments in the chain, which are tentatively partitioned. The alphabet (a, b, c) represents the 1st, 2nd or 3rd segment. In the 1st-Stage SVM model, 9 features are used, but 6 of them about 2 segments are kept and some new features about 3 segments are added into the 2nd-Stage SVM model. Finally, K-means model completes domain assignments.