Fig. 5
From: A novel method for subgroup discovery in precision medicine based on topological data analysis
Visualisation of the proposed hotspot detection algorithm for a toy dataset. A The Mapper algorithm outputs two graph components corresponding to each circle of the graph. Nodes are coloured by the attribute and labelled from 0 to 149. Node size reflects the number of samples in a node. B The identified hotspot is highlighted in red in the Mapper graph. Nodes are unlabelled. C Dendrogram representing the output of the single linkage algorithm performed on the nodes of the Mapper component corresponding to the outer circle from Fig. 4 in the community detection phase. Two community clusters are present and node labels are coloured by average attribute value on the x-axis. The dendrogram is truncated to improve visualisation due to the large number of nodes with zero-value edge weights in the graph. D The confusion matrix comparing the samples contained in the hotspot to the samples originally labelled as Class 1