Table 1 Explanation of the graph-theoretical features used in ≥10 prediction modelling studies. Local features can be averaged over all nodes to obtain the global-scale counterpart
From: Individual-specific networks for prediction modelling – A scoping review of methods
Graph-theoretical feature | Abbrev. | N | Scale | Explanation |
|---|---|---|---|---|
Clustering coefficient | CC | 82 | Both | Ratio of the connected triangles to the maximum possible number of triangles |
Characteristic path length | CPL | 60 | Both | Average of all shortest paths over all pairs of nodes |
Global efficiency | GE | 55 | Global | Average of the reciprocals of the shortest path lengths |
Local efficiency | LE | 45 | Local | Global efficiency applied to the neighbourhood of a node |
Small-world index | SWI | 42 | Global | Ratio of the CC normalized by that expected in a random graph and the CPL normalized by that expected in a random graph |
Degree | Dg | 38 | Local | Number of links of a node |
Betweenness centrality | BC | 36 | Local | Ratio of all shortest paths with and without the node |
Edge weight | EW | 34 | Local | Strength of the connection between two nodes |
Modularity | M | 21 | Global | Degree to which nodes tend to form relatively independent modules |
Density | Ds | 18 | Global | Percentage of observed connections from the maximum number of possible connections |
Assortativity | A | 10 | Global | Pearson correlation coefficient of degree between pairs of connected nodes |