Directed Closeness Centrality
Directed Closeness Centrality (In-closeness and Out-closeness on Directed Graphs) · Also known as: directed closeness, in-closeness centrality, out-closeness centrality, directional closeness
Directed closeness centrality extends the classical closeness measure to directed networks by separately quantifying how quickly a node can be reached by others (in-closeness) and how quickly it can reach all others (out-closeness). It is a foundational node-level metric in social network analysis and graph theory, used wherever link direction conveys meaningful asymmetry such as citation flows, information cascades, or authority hierarchies.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use directed closeness centrality when your network has meaningful arc directionality — citation networks, Twitter follower graphs, hyperlink structures, food webs, supply chains, or any influence flow where direction matters — and you want to identify nodes that are efficient broadcasters (out-closeness) or easy targets of diffusion (in-closeness). It is especially useful when undirected closeness would mask asymmetric roles. Do not use it on undirected networks where direction is absent or artificial. Avoid applying it to sparse or highly disconnected directed graphs without the harmonic-mean correction, as unreachable pairs inflate raw scores misleadingly. For very large graphs, computational cost of all-pairs shortest paths can be prohibitive; approximation algorithms should be considered.
Strengths & limitations
- Captures asymmetric reach that undirected closeness completely obscures, revealing sink versus source roles.
- In-closeness and out-closeness together provide a two-dimensional centrality profile per node.
- Interpretable as average directed geodesic distance — a concrete, intuitive quantity.
- Applicable to any domain with directional flow: information, citations, resources, authority.
- Harmonic-mean variant handles disconnected components without discarding data.
- All-pairs shortest-path computation is O(n(n+m)) — expensive for networks with millions of nodes.
- Undefined or degenerate on graphs that are not strongly connected unless the harmonic-mean correction is applied.
- Does not account for the volume of paths (only shortest ones), unlike betweenness or flow-based measures.
- Sensitive to the choice of normalization and disconnected-pair treatment, making cross-study comparison difficult.
Frequently asked
What is the difference between in-closeness and out-closeness?
In-closeness measures how easily a node can be reached from all others along directed paths — a high in-closeness node is an efficient information sink or target. Out-closeness measures how quickly a node can reach all others — a high out-closeness node is an efficient broadcaster or spreader. The two scores can differ greatly for the same node.
What should I do if my directed graph is not strongly connected?
Use the harmonic-mean closeness formulation: instead of 1 divided by the sum of distances, compute the sum of 1/d(u,v) over reachable pairs only. This treats unreachable pairs as contributing zero, giving a valid score even in disconnected graphs. Most modern packages (NetworkX, igraph) offer this as an option.
How does directed closeness differ from directed betweenness centrality?
Closeness reflects how short a node's directed paths are to (or from) others — it is about speed of access. Betweenness centrality counts how often a node lies on directed shortest paths between other pairs — it is about brokerage and control of flows. A node can have high closeness but low betweenness if it reaches others quickly but is rarely on others' optimal routes.
Should I use directed closeness or directed PageRank to find influential nodes?
Both capture reachability in directed networks but from different angles. Directed out-closeness treats all nodes as equally important sources. PageRank weights incoming links by the importance of their source, capturing recursive authority. Use closeness when geodesic efficiency matters; use PageRank when link prestige and recursive endorsement are theoretically meaningful.
Is directed closeness centrality suitable for weighted directed graphs?
The standard formulation uses unweighted shortest-path hop counts. For weighted directed graphs — where arc weights represent distance, cost, or dissimilarity — substitute Dijkstra-based weighted shortest paths. If weights represent strength or similarity rather than distance, convert them to cost (e.g., take the inverse) before applying the algorithm.
Sources
- Wasserman, S. & Faust, K. (1994). Social Network Analysis: Methods and Applications. Cambridge University Press. ISBN: 978-0-521-38269-4
- Freeman, L. C. (1979). Centrality in social networks conceptual clarification. Social Networks, 1(3), 215–239. DOI: 10.1016/0378-8733(78)90021-7 ↗
How to cite this page
ScholarGate. (2026, June 3). Directed Closeness Centrality (In-closeness and Out-closeness on Directed Graphs). ScholarGate. https://scholargate.app/en/network-analysis/directed-closeness-centrality
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Closeness CentralityNetwork analysis↔ compare
- Directed Betweenness CentralityNetwork analysis↔ compare
- Directed Eigenvector CentralityNetwork analysis↔ compare
- Directed PageRankNetwork analysis↔ compare
- Directed Social Network AnalysisNetwork analysis↔ compare