Link Prediction Survey
As part of the Ph.D. work by Víctor Martínez, he performed an extensive bibliographic study of link prediction techniques and, as far as we know, the most extensive empirical evaluation of link prediction techniques that is currently available.
Our survey has been published by ACM Computing Surveys, one of the top journals in the field:
Víctor Martínez, Fernando Berzal & Juan-Carlos Cubero:
"A Survey of Link Prediction in Complex Networks"
ACM Computing Surveys, Volume 49, Issue 4, Article No. 69, February 2017
ACM Digital Library: http://dl.acm.org/citation.cfm?id=3012704
According to Thomson Reuters Journal Citation Reports, in 2016, ACM Computing Surveys had an impact factor of 6.748, being the second top-ranked journal in the "Computer Science - Theory & Methods" category (i.e. 2nd out of 104 research journals). The JCR impact factor is a measure reflecting the yearly average number of citations to recent articles published in that journal. It is frequently used as a proxy for the relative importance of a journal within its field. Journals with higher impact factors are often deemed to be more important than those with lower ones. The impact factor was devised by Eugene Garfield, the founder of the Institute for Scientific Information.
Given that ACM Computing Surveys published 143 papers in the last two years (104 papers in 2015 and 39 papers in 2014) and those papers received 965 citations in 2016 (386 citations to papers published in 2015 and 579 citations to papers published in 2014), its impact factor is just the result of dividing the number of citations to recent items (965) by the number of recently-published papers (143): 965/143 = 6.748.
The NOESIS Network Analyzer offers a score of link scoring and prediction methods. Our tool now supports the following methods, which can be used both for link scoring (i.e. evaluating existing link) and link prediction (i.e. predicting new links):
- CN (Common Neighbors).
- AA (Adamic-Adar score).
- RA (Resource Allocation score).
- J (Jaccard score).
- PA (Preferential Attachment score).
- Salton score.
- Sorensen score.
- HPI (Hub-promoted index)
- HDI (Hub-depressed index).
- LLHN (Local Leicht-Holme-Newman score)
- GLHN (Global Leicht-Holme-Newman score)
- K (Katz score)
- RW (random walk score).
- RWR (random walk with restart score).
- FP (flow propagation score).
- PL (pseudoinverse Laplacian score).
- ACT (average commute time score).
- RFK (random forest kernel score).
You can access these methods using the "Analysis > Links > Link prediction" menu for predicting new links or the "Analysis > Links > Link scoring"menu for evaluating existing ones.