This is the trust and distrust network of Epinions, an online product rating site. The network consist of individual users connected by directed trust and distrust links. Edges have the weight +1 for trust and −1 for distrust. Note that a user can trust or distrust himself, and therefore the network contains loops.


Internal nameepinions
Data sourcehttp://www.trustlet.org/wiki/Extended_Epinions_dataset
AvailabilityDataset is available for download
Consistency checkDataset passed all tests
Online social network
Node meaningUser
Edge meaningTrust
Network formatUnipartite, directed
Edge typeSigned, possibly weighted, no multiple edges
Temporal data Edges are annotated with timestamps
ReciprocalContains reciprocal edges
Directed cyclesContains directed cycles
LoopsContains loops


Size n =131,828
Volume m =841,372
Loop count l =573
Wedge count s =182,193,467
Claw count z =78,294,267,861
Cross count x =31,252,929,210,362
Triangle count t =4,910,076
Square count q =715,390,258
4-Tour count T4 =6,453,318,352
Maximum degree dmax =3,622
Maximum outdegree d+max =2,070
Maximum indegree dmax =3,478
Average degree d =12.764 7
Fill p =4.841 42 × 10−5
Size of LCC N =119,130
Size of LSCC Ns =41,441
Relative size of LSCC Nrs =0.314 357
Diameter δ =16
50-Percentile effective diameter δ0.5 =3.574 21
90-Percentile effective diameter δ0.9 =4.771 62
Mean distance δm =4.109 14
Gini coefficient G =0.835 782
Balanced inequality ratio P =0.144 853
Outdegree balanced inequality ratio P+ =0.159 785
Indegree balanced inequality ratio P =0.169 681
Relative edge distribution entropy Her =0.830 630
Power law exponent γ =2.128 76
Tail power law exponent γt =1.741 00
Degree assortativity ρ =−0.069 571 4
Degree assortativity p-value pρ =0.000 00
In/outdegree correlation ρ± =+0.644 512
Clustering coefficient c =0.080 849 4
Spectral norm α =87.200 1
Operator 2-norm ν =167.764
Cyclic eigenvalue π =143.123
Algebraic connectivity a =0.002 810 82
Spectral separation 1[A] / λ2[A]| =1.270 22
Reciprocity y =0.308 723
Non-bipartivity bA =0.755 904
Normalized non-bipartivity bN =0.017 168 2
Spectral bipartite frustration bK =0.000 656 904
Negativity ζ =0.147 028
Algebraic conflict ξ =0.031 081 6
Triadic conflict τ =0.095 043 7
Spectral signed frustration φ =0.000 659 724
Controllability C =79,193


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 Laplacian

Spectral graph drawing based on the normalized adjacency matrix

Degree assortativity

Zipf plot

Hop distribution

In/outdegree scatter plot

Item rating evolution

Edge weight/multiplicity distribution

Clustering coefficient distribution

Average neighbor degree distribution

Temporal distribution

Diameter/density evolution

Signed temporal distribution


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] Paolo Massa and Paolo Avesani. Controversial users demand local trust metrics: an experimental study on epinions.com community. In Proc. American Association for Artif. Intell. Conf., pages 121–126, 2005.