Twitter mentions

This is the directed network of "@username" mentions on Twitter. Each node is a Twitter user, and each directed edge from user A to user B means that user A has mentioned user B in a tweet using the "@username" syntax. Multiple edges are allowed, and each edge is annotated with the timestamp of the tweet. Since it is possible to mention one's own username, this network contains loops.


Internal namemunmun_twitterex_at
NameTwitter mentions
Data source
AvailabilityDataset is not available for download
Consistency checkDataset passed all tests
Online contact network
Dataset timestamp 2010
Node meaningUser
Edge meaningMention
Network formatUnipartite, directed
Edge typeUnweighted, multiple edges
Temporal data Edges are annotated with timestamps
ReciprocalContains reciprocal edges
Directed cyclesContains directed cycles
LoopsContains loops


Size n =2,919,613
Volume m =12,887,063
Unique edge count m̿ =7,301,101
Loop count l =235,300
Wedge count s =736,717,625
Claw count z =690,658,257,498
Cross count x =1,640,036,248,049,757
Triangle count t =1,449,797
Square count q =81,161,696
4-Tour count T4 =3,610,479,352
Maximum degree dmax =39,753
Maximum outdegree d+max =1,697
Maximum indegree dmax =39,753
Average degree d =8.827 93
Fill p =8.565 21 × 10−7
Average edge multiplicity m̃ =1.765 08
Size of LCC N =2,893,623
Size of LSCC Ns =98,784
Relative size of LSCC Nrs =0.033 834 6
Diameter δ =23
50-Percentile effective diameter δ0.5 =4.987 02
90-Percentile effective diameter δ0.9 =5.908 45
Mean distance δm =5.454 36
Gini coefficient G =0.798 378
Balanced inequality ratio P =0.164 585
Outdegree balanced inequality ratio P+ =0.281 465
Indegree balanced inequality ratio P =0.229 310
Relative edge distribution entropy Her =0.894 306
Power law exponent γ =2.845 35
Tail power law exponent γt =1.861 00
Degree assortativity ρ =−0.010 567 6
Degree assortativity p-value pρ =0.000 00
Clustering coefficient c =0.005 903 74
Spectral norm α =1,114.49
Operator 2-norm ν =1,112.63
Cyclic eigenvalue π =345.070
Reciprocity y =0.031 959 8
Non-bipartivity bA =0.002 874 42
Normalized non-bipartivity bN =0.000 598 655
Spectral bipartite frustration bK =6.019 94 × 10−5


Degree distribution

Cumulative degree distribution

Lorenz curve

Spectral distribution of the adjacency matrix

Spectral distribution of the normalized adjacency matrix

Spectral distribution of the Laplacian

Spectral graph drawing based on the adjacency matrix

Spectral graph drawing based on the normalized adjacency matrix

Hop distribution

In/outdegree scatter plot

Edge weight/multiplicity distribution

Clustering coefficient distribution

Temporal distribution

Diameter/density evolution


Inter-event distribution

Node-level inter-event distribution

Matrix decompositions plots



[1] Jérôme Kunegis. KONECT – The Koblenz Network Collection. In Proc. Int. Conf. on World Wide Web Companion, pages 1343–1350, 2013. [ http ]
[2] Munmun De Choudhury, Yu-Ru Lin, Hari Sundaram, K. Selçuk Candan, Lexing Xie, and Aisling Kelliher. How does the data sampling strategy impact the discovery of information diffusion in social media? In ICWSM, pages 34–41, 2010.