Skip to main content


Springer Nature is making Coronavirus research free. View research | View latest news | Sign up for updates

Fig. 3 | Applied Network Science

Fig. 3

From: From free text to clusters of content in health records: an unsupervised graph partitioning approach

Fig. 3

The top plot presents the results of the Markov Stability algorithm across Markov times, showing the number of clusters of the optimised partition (red), the variation of information VI(t) for the ensemble of optimised solutions at each time (blue) and the variation of Information VI(t,t) between the optimised partitions across Markov time (background colourmap). Relevant partitions are indicated by dips of VI(t) and extended plateaux of VI(t,t). We choose five levels with different resolutions (from 44 communities to 3) in our analysis. The Sankey diagram below illustrates how the communities of documents (indicated by numbers and colours) map across Markov time scales. The community structure across scales present a strong quasi-hierarchical character—a result of the analysis and the properties of the data, since it is not imposed a priori. The different partitions for the five chosen levels are shown on a graph layout for the document similarity graph created with the MST-kNN algorithm with k=13. The colours correspond to the communities found by MS indicating content clusters

Back to article page