Content uploaded by Piotr Potiopa
Author content
All content in this area was uploaded by Piotr Potiopa on May 08, 2015
Content may be subject to copyright.
The Assessment of the EPQ Parameter for
Detecting H-Index Manipulation and the
Analysis of Scientific Publications
Rafał Rumin1, Piotr Potiopa2
Abstract The work presents the analysis of mechanisms for determining the sus-
ceptibility of parametric indices (such as the h-index) of evaluation of scientific
articles published on the modification of parameters not resulting from essential
value of the research work. Currently, most methods for verifying the article is fo-
cused on the selection of works potentially strongly influence the international po-
sition of a journal. To this end, editorial offices wide use of parametric methods of
assessment. In addition, the work attempts to identify the used criterion functions,
namely the assessment parameters and guidance, the risks associated with using
this type of method to change the popular parametric indexes for authors and jour-
nals. These parameters are divided into categories and offered their initial verifica-
tion based on statistical analysis of already published articles in various journals.
Each parameter has attributed weight function, which allows to define its impact
on the total evaluation of an article, and also adaptation of formula to any academ-
ic journal. Weight functions will be determined with the usage of neural networks
or genetic algorithms, aiming to their individual adaptation to particular journal.
1 Introduction
Relentless pursuit of scientific journals to obtain the greatest possible number of
points in the created rankings enhances continuous improvement of parametric al-
gorithms to verify the quality of the article and the assessment of its author (Phila-
delphia List, Impact Factor, quoting indicators etc.) [2-4].
All the time created a new methods of evaluation of journals and modifications
of existing criteria result in a situation that merits evaluation of the article can be
replaced by a parametric assessment forced by the publisher [8-16].
1AGH University of Science and Technology, Al. Mickiewicza 30, 30-059 Kraków, Poland,
rumin@agh.edu.pl
2AGH University of Science and Technology, Al. Mickiewicza 30, 30-059 Kraków, Poland,
ppotiopa@zarz.agh.edu.pl
2
The result is a situation that good, exploratory research publications may be as-
sessed or unfairly withdrawn from the publications on the ground that have been
poorly prepared for parametric criteria. Thus, the following aspects parameteriza-
tion are to determine the influence of subjective factors in the evaluation of scien-
tific articles in specific journals. Furthermore, there were presented series of fac-
tors which, if they are taken into consideration during writing of scientific articles,
they have a chance to increase probability of obtaining positive review and in ef-
fect the acceptance of publication in renowned journals. In the further process of
research works, realization of automatic information system is planned, which role
will be connected with the verification of the working version of an article, before
sending it to the journal and the definition of the probability of obtaining high pa-
rametric evaluation. Described parametric evaluation will determine the coeffi-
cient EPQ – Estimated Paper Quality. This co-efficiency will be helpful for scien-
tists who concentrate mainly on essential, and less over the editorial part of their
scientific article. The low value of EPQ should induce the author to analyze and
supplement his publication before sending article to editorial office of the pre-
viously chosen journals.
2 Mathematical models for authors evaluation
Authors of scientific articles are subject to verification by placing in the ranking
reflects their contribution to the development of the field of scientific work. One
of main parameters applied in relation to authors of publication is proposed in year
2005 the Hirsch index (h-index) [1]. As easily can be envisaged, such evaluation
can be sensitive on manipulation on the side of several cooperating with each oth-
er authors, who mutually will quote their works (apart from their essential contri-
bution into researches). The parametric evaluation of publication issues from the
category of scientific research. Scientific researches require financing, and one of
the popular sources of learning financing are exploratory grants. To obtain financ-
ing it is expected that the scientist will carry out planned investigations and their
effect will have visible influence on a given exploratory field. How such influence
is measured?
Most legible measures are publications and their quotations. For this rea-
son, scientists who have a suitably high Hirsch index, are treated as trustworthy to
commit to them public money on carried researches. Legible dependencies appear
between the financing of research, with the quantity of publication, and with their
quotations which put each other greater chance for future financial resources.
3
2.1 Mathematical models for journals evaluation.
The high quality journal tries to be visible for society of scientists. To found a
difference between quality of journals, the special parametric factors was pro-
posed. Below are some of them [20-26]: Source-Normalized Impact per Paper
(SNIP), Relative Citation Rates (RCR), SCImago Journal Rank (SJR), Journal to
Field Impact Score (JFIS), Article Influence (AI), and the most popular - Impact
Factor (IF) (1).
“IF” counts all citations from particular calendar year, and it divides them by the
amount of “cited” publications from last two years (C).
IF=B
C
(1)
Other indicators also reflect the parametric quality rating of journal, but they are
not so popular. Each of them characterizes different factors which influence final
evaluation of journals. Different journal evaluation criteria cause the inhomo-
geneity in resultant rankings. Furthermore, algorithms of evaluation are subjected
to continuous changes aiming to the most reliable definition of publications quali-
ty. For this reason, the aim of publishing companies, instead of valuing scientific
publications, having less 'popular' character (though substantially equally good if
not much better), could be the wish of achievement of the highest parametric coef-
ficients evaluating the other of their publications.
It can be accepted that, as the evaluation of the given journal is higher in the rank-
ing, an article published in it has the chance to obtain greater range, and conse-
quently receive greater quantity of quotations. It seems that there exists the con-
formity of business among a journal and an author of the article, however this
concerns only the wishes of obtaining the maximum quotations quantity at other
publishers through the large number of scientists.
Willing to check our chances for the publication in the given journal, we often set
incorrect question - Will this journal publish my article?
To show existing dependences and conflicts of interests between an author and an
editor, one ought to set himself the question:
How will my article help the journal to obtain better position in the ranking (more
points in the parametric evaluation of journals)?
The answer depends on many factors, which can subjectively influence the
evaluation of an article, apart from its essential value. Figure 1 shows a general
scheme of the relations between a publishing house and an author.
4
PUBLISHING HOUSES
•EVALUATE THE QUALITY OF
JOURNALS
JOURNALS
•EVALUATE THE QUALITY OF
ARTICLES
PUBLICATIONS
•CITE OTHER ARTI CLES
•AUTOCITING
CITING
•INFLUENCES THE EVALUATION
OF PUBLICATIONS, AUTHORS,
JOURNALS A ND PUBLISHING
HOUSES
THE EVALUATION OF
AUTHORS
•INFLUENCES THE EVALUATION
OF RESEARCH UNITS
•INCREASES THE CHANCE OF
RECEIVI NG GRANTS
GRANTS
•FINANCE RES EARCH WORKS ,
RESEARCHERS AND THEI R
PUBLICATIONS
RESEARCHES
•SOLVE FACTUAL SCIENTIFIC
PROBLEMS
•CONSTITUTE THE TOPIC OF THE
PUBLICATION
Fig. 1.The scheme of the relations between a publishing house and an author
2.2 Mathematical models for article evaluation.
Each article has its own essential value which cannot be measured automatically.
The article evaluation is limited to a group of parameters defining its quality from
the interest of a journal point of view. Unfortunately this can cause a conflict of
interests between publishing houses and authors [17].
During the evaluation of an article, the essential value can be estimated by addi-
tional parameters: the range of carried out researches, description of theoretical
models, simulation models, experiment. If it has only theory it can be classified
lower than articles containing simulation or experiment.
Articles containing the experimental verification of carried researches will be eva-
luated as the best ones. Separately, articles containing rich and complex reviews of
the literature from the given field can display significantly high classification, be-
cause this type of articles are quoted often many times. This results from the spe-
cific approach of scientists to carried out researches and wishes of using elabo-
rated earlier literature review, which often requires a lot of time and belongs to
„less attractive” researches.
Thereby, at the evaluation of articles’ value, nobody can foresee how often he will
be quoted in the future. To put it simply, it can be assumed, that at the initial phase
5
of article analysis, each has the evaluation for the essential value on the same lev-
el. Since the quantity of elaborated article future quotations cannot be influenced,
it can be influenced who the author quotes in his own publication. This way the
quantity of "gained' quotations from the journal's point of view, can be controlled .
The issue here is the period of time in which journals are subjected to evaluation
in rankings. For the calculation of Impact Factor, last 2 years are taken into con-
sideration which means that the auto quotation of other articles which appeared in
the same publishing house within a period of last 2 years have a positive influence
on IF indicator increase. Therefore, the publishing house will be willingly promot-
ing articles which already show quotations from their own journal, is a method to
obtain higher place in the ranking. However, if there exists a group of journals
given by the common institution, then cross quotations of other journals belonging
to the same publisher constitute also an added value. Here arises a threat regarding
the reliability of 1 published articles, because one can apply the mechanism which
would permit ranking speculations between journals. Following the paragraphs of
this article, they contain the case study describing such situations.
3 The new indicator for the Parametric Evaluation of an Article
– EPQ.
All articles can be parameterized by the EPQ coefficient (Estimated Paper Quali-
ty). This model can indicate many factors which participate in the evaluation of
given article. It can be presented as weighted mean of individual parameters, with
suitably assorted weights functions. The value of parameters is standardized so
that it contains itself in the range from 0 to 1. This type of method descends from
Churchman and Ackoff (1954) researches, under the name SAW (Simple Addi-
tive Weighting) [18-19]. SAW is one of most popular solutions in MADM type
(Multi-Attribute Decision Making) problems of which undoubtedly is the problem
described in the work. Elaborated process of EPQ calculation is similar to the
above methods, however differences in designating of individual parameters ap-
pear. Differences are caused by different ways of Pi parameters of values determi-
nation.
=1
=1 (2)
where Pi is appropriate parameter of evaluation with following index n appointed,
and wi is the weight for a given parameter. Below, in the table (cf. Table 1-3) the
list of parameters together with their asserted values and ranges is presented. All
parameters Pi are situated in the same range:[0,1].
6
3.1 The Methodology of calculation the EPQ
Calculation of the EPQ coefficient is based on a lot of other indicators described
bellow. Particularly essential from the usage of EPQ indicator point of view, is the
possibility of weights definition wi in way compatible to parametric evaluations
applied by the given journal. The large number of academic journals cause differ-
ent approach to the parametric evaluation of accepted articles to editorial office
and the review of article. Basing on the data from previous years, considering all
publications printed within the framework of one publishing-title , we are able to
determine weights of individual parameters individually for the given journal.
For that purpose we will use neural networks with the feedback which will learn to
recognize the influence of the given parameter on the positive acceptance of ar-
ticle to the publication. In case of the analysis, already printed publications, we
will subordinate the quantity of published articles from the value of individual pa-
rameters. The more articles will have e.g. the high parameter P6, the greater influ-
ence on the printing of publication has the quantity of archival articles quotations
laded from the same journal.
Initial values of the weight parameter w1amount to 1. Due to the implication of
matching algorithm, the weights shall be modified in the 0 to 1 bracket through ar-
tificial neural networks. The aim of the modification is the selection of appropriate
levels of weights to a given journal.
Table 1.Defining parameters for the calculation of the EPQ indicator – basic parameters:
Pi Meaning of value substituted to Pi Range Formula on Pi
P1 H – authors’ Hirsch index
H=[0:inf]
1=11
11
P
2
I – the quantity of authors’ indexed publications
I=[0:inf]
2=11
12
P3 C – quantity of authors’ indexed quotations
C=[0:inf]
3=11
1 + 3
P4 S - degree/ the scientific title of the author
(none/engineer/MSc/the doctor/assistant
professor/professor)
S=[0:5]
4=11
1 + 20 4
Table 2.Defining parameters for calculation of the EPQ Indicator –content rating of an article:
Pi Meaning of value substituted to Pi Range Formula on Pi
P5
The calculated Gaussian distribution basing on
the quantity of all quotations contained by an
author in the article, where:
d- the height of the Gaussian curve top,
x- quantity of all quotations contained by an
author in the article,
σ- standard deviation of Gaussian distribution,
d=[1]
x=[0:inf]
σ=[0:inf]
5=()2
225
7
μ- expected value, equal average quantity of
quotations devolving on one article in the given
journal,
a - quotations devolving on one article (k) in the
given journal.
μ=[0:inf]
a=[0:inf]
k=[0:inf]
=1
1
(
)
2
=1
=1
=1
P6 A - the quantity of quotations coming from
archival numbers of the same journal to which
the publication is submitted
A=[0:inf]
6=11
16
P7 B - the quantity of quotations coming from
archival numbers of remaining journals
belonging to the same publishing house to which
publication is submitted
B=[0:inf]
7=11
17
P8
The indicator of the publication originality.
O - the quantity of similar articles earlier
published by the author.
D - the sum of "duplicates”, measured by the
coefficient of similarity of genuine text and
small pictures between previous articles of the
author, and with his current publication
O=[0:inf]
D=[0:inf]
8=11
1 + 8
=
=1
P9 Rd - the quantity of cited publications of the
current editor of journal to which publication is
submitted
Rd=[0:inf]
9=11
1 + 9
P10 Rc - the quantity of cited publications of current
reviewer of journal to which publication is
submitted
Rc=[0:inf]
10 =11
1 + 10
Table 3.Defining parameters for calculation of the EPQ indicator – other parameters:
Pi Meaning of value substituted to Pi Range Formula on Pi
P11 J - the quantity of authors’ publications quoted
by a current editor or reviewer of the journal to
which publication is submitted
[0:inf]
11 =11
1 + 11
P12 K - the quantity of authors’ common publication
articles and a current editor or reviewer of a
journal to which publication is submitted
[0:inf]
12 =11
1 + 12
P
13
Z –the quantity of elements from the range of
carried out researches (the form of survey):
review, theory, model, simulation, experiment,
lack/other.
[0:5]
13 =11
1 + 20 +13
3.2 The Example of the EPQ calculation
The definition of the exact value EPQ does not decide about “the success” and the
publication of the given magazine article. This will permit however finishing up
and improving of the editorial part which could not take into the above-mentioned
factors influencing decision of editors and reviewers. System elaborated in such a
8
way, using the IT network will permit quick definition of the article modification.
Outwardly, basing on obtained result EPQ it will enable to propose the alternative
academic journal which parameters answer to the result. The value EPQ was cal-
culated basing on the example of the publication based on the Matlab software.
Below, the calculated value of EPQ was presented for a model publication.
a) Data concerning the author:
The current Hirsch index of the author amounted to 5, for 100 indexed publica-
tions and 500 of all his connective cited publications. The author obtains the aca-
demic title of professor.
b) Data concerning a publication:
The article contains 90 of citations, from which 2 citations come from archival
journal, to which the publication is being composed. In total, there are 2 citations
from other archival issues of journals belonging to the same publishing house, to
which the publication is being composed. The publication demonstrates original
quality, since there have not been any of its duplicate samples and publications of
similar content of the same author. In the publication there are no citations from
the works of the members of editorial board, however there have been 2. citations
of works developed by reviewers. The publication contains at least 3. elements of
scientific publication (e.g. review, theory, model).
c) Data concerning the journal:
The statistical average number of citations in a single article published in this
journal amounts to 100. Editors and reviewers have not cited any other works of
the author, they have not had any mutual articles with the author.
The range of the selected parameters together with separate results of the calcula-
tions are depicted below.
Table4. The results from this case are as follows:
No Parameter Pi Initial data Pi results
1
P
1
H=5
0,833
2
P
2
I=100
0,990
3 P3 C=500 0,998
4 P4 S=5 0,990
5 P5 d=1; x2=90; a1=80; a2=100; a3=120 0,882
6 P6 A=2 0,667
7 P7 B=2 0,667
8 P8 D=0; O=0 1
9 P9 Rd=0 0
10 P10 Rc=2 0,667
11
P
11
J=0
0
9
12 P12 K=0 0
13
P
13
Z=3
0,984
Average 0,667
EPQ 0,67
4 SEO, Hirsch Index and Impact Factor
4.1 The similarity of the Hirsch Index and Impact Factor to Page
Rank, and Threats Resulting from Black Hat SEO Methods.
The growth of the Hirsch index and IF strictly depends on the quantity of the giv-
en author’s publication quotations. This model can be compared to the published
ranking of websites (PR - Page Rank) used in Google search engine [5-7]. The si-
milarity refers to the quantity of quotations which correspond to quantities of re-
turnable links indicating given page of data sources.
There are known general methods of influencing the algorithm of search engine in
this way, so that the indicated page will be higher in the SERP ranking (Search
Engine Results Page). These methods are divided on the so-called white and
black. White Hat SEO - means the positioning of the website in compliance with
official guidelines of search engines, which should result in better page adaptation
to Web-crawler's and engines of search engines requirements. Good preparation of
the website facilitates, quick indexing of it in the search engine base of data, how-
ever increasing number of valuable references to page (gained naturally and re-
sulting from its popularity and uniqueness) permits its positioning and obtaining
of high place in the SERP ranking. As valuable references are acknowledged,
links from pages about high PR are often visited by users (e.g. thematic, commu-
nity websites). There also exists Black Hat SEO which is characterized by the use
of all possible gaps in the search engine, for the purpose of raising the ranking of
given website. Such effects are achieved through the manipulation with the quanti-
ty of returnable links and their “artificial” addition through generating large quan-
tity of pages with links. So many of manipulation methods constitute the necessity
of continuous algorithms change of search and qualitative selection of websites.
From obvious reasons, exact parameters of the algorithm are not revealed for the
purpose of their protection before the manipulation. There can be only estimated
general dependencies and on their base there can be created algorithms improving
the position of website in ranking of searches. Methods of rankings creating e.g.
PR and IF, and also H- index cause the risk of appearing methods taken from
SEO, which in the artificial way will manipulate results of the above-mentioned
10
rankings. Probably there is no possibility of obtaining 100% reliable and objec-
tive ranking not burdened with the above risk.
From this reason, the essential evaluation of publication can be shaken, in the in-
terest of the parametric evaluation. This can cause the reverse to intended effect
i.e. these rankings will promote less ambitious scientific discoveries, but artificial-
ly will overvalue indexes across the elaboration of their manipulation method. The
case study is presented below, which in the mental experiment, could result with
“artificial” increasing of IF for the journal or with “artificial” increasing of the H-
index for given scientist.
4.2.Threats resulting from the usage of artificial methods of
increasing the indexes
Among academic researchers, there is no unambiguous method of scientific
achievements’ evaluation, which would credibly and objectively determine the
value of a research work. There are numerous publications that describe threats
connected with manipulation of indexes [27, 28]. A short analysis of the case that
highlights the susceptibility of the used indexes for artificial manipulation is
shown below. It is a similar situation that the search engine Google encountered. It
is subjected to continuous attempts of manipulation of websites’ rankings that are
displayed in the first 10 results of a search. The search engine algorithm was
evolving, taking into account increasingly different parameters so as to extract ar-
tificial positioning. If the algorithms of calculating the indexes are not modified,
the intentional manipulation to increase indexes is very likely to occur, which is
shown below.
4.3 Hirsh Index Manipulation
The Hirsch index has lots of advantages, however it is also subjected to the risk of
being manipulated. Suppose there are at least 2 researchers working in similar
field of study, they may cite each other’s work (Fig. 2), only to increase each
other’s Hirsch index.
Of course the publication that contains several or a dozen of citations of the same
author may focus the attention of the reviewer as far as the legitimacy of citing is
concerned. In such a case there exists the possibility of developing the model for a
specified group of authors, where division of citations between each other will be
evenly determined so as not to undermine the legitimacy. Furthermore, it is worth
to add that the publication drawn up by an author A that contains more or less 5
citations of an author B is fully sufficient to increase the Hirsch index from 0 to 3
level. In the cooperation of at least 6 authors, the increase of H index to the level 5
11
for each of the author will demand only the cooperation in the scope of citing of 5
publications (6 counting from the first one, which has not been cited before by no-
body).
H-index of
Person A
Number of
publications
of Person A
Number of
citations of
Person B*
Number of
citations of
Person A*
Number of
publications
of Person B
H-index of
Person B
0
1
0
0
1
0
1
2
1
1
2
1
1
3
2
2
3
1
2
4
3
3
4
2
2
5
4
4
5
2
3
6
5
5
6
3
3
7
6
6
7
3
Fig. 2. Mutual citations of 2 authors.
*- for each next publication one person has to cite other person’s all publications
(the number of citations per new publication).
5 Application Realizing EPQ Designation
Determination of the individual parameters can be achieved through the use of
available databases of scientific publications and names of authors like: SCOPUS,
WEB OF KNOWLEDGE, Google Scholar and other smaller databases. The sys-
tem collects this data, it will turn collecting the following information: author, ci-
tations, journals, publications, and then assessed the parametric analyzed publica-
tion. Based on this evaluation, it can propose suggestions, ref. introductions of
changes in the article or present proposal of the alternative journal to which the
parametric evaluation was better. The system architecture may be built based on
the client-server methodology which is presented on Fig. 3.
12
5.1 The architectural Schema
On the after-mentioned figure3 the general architectural schema of the system is
presented.
Fig. 3.The architecture of a proposed IT system to determine EPQ indicator.
In presented architecture system we distinguish:
1. Presentation layer - a layer of the application responsible for the presen-
tation of results and communication with user, receiving data from user
(proposed article, survey for the author)
2. Application layer - layer responsible for the resumption of data and
processing of results which consists of:
• citations’ analyzer (module processing the quotation categorizing and
counting quotations of authors works.
• authors’ data analyzer (module processing data of authors (also review-
ers and editors), checking relations of author with journals across quota-
tions as well as categorizing his achievement)
• authors’ analyzer (module being supposed for the task to process availa-
ble data sources information of used in the algorithm coefficients for
journals and authors)
3. DB–a layer of database recording source data and results of calculations
with application layer, permitting caching of data sources in the situation
when data don't need to be refreshed at every operation of weight-
coefficients calculation weight- coefficients.
4. Sources–a layer of gaining data from chosen sources dividing into
sources of quotations gaining ("citations sources"), given authors ("au-
13
thors the date sources") and coefficients used in the algorithm of EPQ
count ("factors sources").
The system architecture in case of further development can be calibrated because
the module of processing may receive partial results of calculations (weights of
component parameters) from individual modules which can be found on separate
instances of servers. Each module of gaining data can have the separate database
in which it will store the received results of the data sources indexing. In case of
presentation layer, the system can communicate with software of the thin client
type in case of approach users (authors of articles) and with the software of the fat
client type in case of the administrator who can control work of the processing
module (settings control).
6 Conclusion
Determining the actual accuracy of parametric evaluation for scientific journal ar-
ticle by calculating the proposed EPQ parameter is possible only after verification
on figures. Methodology is based on foundations that a substantially good article
can be evaluated wrongly due to remaining factors on which reviewers and editors
of journals pay attention to. The elaborated system, proper verifying of targets and
correction of an article before its delivery to publishing houses will permit to carry
out essential research on equally high level and to regard subjective 'expectations'
from the side of publishing house in relation to an author. So an improved article
has greater chance for printing in a renowned journal, which can positively re-
bound on future publications of many authors.
On the other hand, it will be possible to verify retrogradely articles that have too
high parametric evaluation to indicate potential authors or journals, which were
subjected to artificial mechanisms that are used to increase index parameters.
Through this process it will be possible definition of what elements have been
streamlined parametric evaluation of the use of illegal solutions, which include the
use of observed SEO methods.
REFERENCES
1. [1] J. E. Hirsch, “An index to quantify an individual’s scientific research
output.” Proc. Nat. Acad, Sci. (PNAS), 2005, vol. 102, nr 46.
2. [2] L. Egghe, R. Rousseau, “Introduction to informetrics”. Amsterdam:
Elsevier, 1990.
3. [3] H. F. Moed, M. Vriens, “Possible inaccuracies occurring in citation
analysis”. Journal of Information Science, 1989, 15, 95-107.
4. [4] R. Todorov, W. Glanzel, “Journal citation measures: A concise re-
view”. Journal of Information Science, 1988, I4. 47-65.
5. [5] S. Brin, L. Page. “The anatomy of a large-scale hypertextual web
search engine”. In WWW7: Proceedings of the seventh international con-
14
ference on World Wide Web 7, pages 107–117. Elsevier Science Pub-
lishers B. V., 1998.
6. [6] L. Page, S. Brin, R. Motwani, T. Winograd, “The PageRank citation
ranking: Bringing order to the web”. Technical Report, Stanford Digital
Library Technologies Project, 1998.
7. [7] S. Maslov, S. Redner, “Promise and Pitfalls of Extending Google’s
PageRank Algorithm to Citation Networks”, Journal of Neuroscience
2008, 28, 11103.
8. [8] D. Zhou, S. A. Orshanskiy, H. Zha, C. L. Giles, “Co-Ranking Authors
and Documents in a Heterogeneous Network”, INTERNATIONAL
CONFERENCE ON DATA MINING, 2007.
9. [9] P. Chen, H. Xie, S. Maslov, S. Redner. “Finding Scientific Gems with
Google”. J.INFORMET., 1:8, 2007.
10. [10] E. Garfield. “Citation analysis as a tool in journal evaluation”.
Science, 178(60):471–479, November 1972.
11. [11] X. Liu, J. Bollen, M. L. Nelson, and H. Van de Sompel. “Coauthor-
ship networks in the digital library research community”. ar-
Xiv.org:cs/0502056, 2005.
12. [12] M. Bianchini, M. Gori, and F. Scarselli. “Inside pagerank”. ACM
Trans. Inter. Tech., 5(1):92–128, 2005.
13. [13] De Moya, F. “The SJR indicator: A new indicator of journals’ scien-
tific prestige”, 2009, Arvix. (arxiv.org/abs/0912.4141)
14. [14] E. Garfield, “The history and meaning of the journal impact factor”.
JAMA 2006;295:90-3.
15. [15] M. Amin, M. Mabe, “Impact factors: Use and Abuse”. Perspectives
in Publishing, 2000 - October, 1, 1-6
16. [16] Mu-Hsuan Huang, Wen-Yau Cathy Lin, "The influence of journal
self-citations on journal impact factor and immediacy index", Online In-
formation Review, 2012, Vol. 36 Iss: 5, pp.639 – 654
17. [17] C. Wenneras, A. Wold ”Nepotism and sexism in peer-review”. Na-
ture. 1997 May 22;387(6631):341-3.
18. [18] C.W. Churchman, R.L. Ackoff, “An approximate measure of value”.
Journal of the Operational, Research Society of America, 1954, 2(2).
19. [19] D. Widayanti , S. Oka, S. Arya, “Analysis and Implementation
Fuzzy Multi-Attribute Decision Making SAW Method for Selection of
High Achieving Students in Faculty Level” IJCSI International Journal of
Computer Science Issues, Vol. 10, Issue 1, No 2, 2013, ISSN 1694-0784
20. [20] SNIP and SJR at Journal Metrics (www.journalmetrics.com)
21. [21] SCImago Journal Rank (SJR) (http://www.scimagojr.com/)
22. [22] Source – Normalized Impact per Paper (SNIP)
(www.journalindicators.com)
23. [23] Impact Factor (IF) (http://thomsonreuters.com/products_services/sci
ence/free/essays/impact_factor)
15
24. [24] h-index: (http://help.scopus.com/robo/projects/schelp/h_hirschgraph.
htm)
25. [25] Article Influence (AI) (www.eigenfactor.org)
26. [26] Relative Citation Rates (RCR)/Journal to Field Impact Score (JFIS)
27. [27] Bartneck, Christoph, and Servaas Kokkelmans. "Detecting h-index
manipulation through self-citation analysis." Scientometrics 87.1 (2011).
28. [28] Jin, Bihui, et al. "The R-and AR-indices: Complementing the h-
index." Chinese science bulletin 52.6 (2007): 855-863.
