Indonesian J our nal of Electrical Engineering and Computer Science V ol. 39, No. 2, August 2025, pp. 973 986 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v39.i2.pp973-986 973 CriteriaCheck er: a kno wledge graph appr oach to enhance integrity and ethics in academic publication Garima Sharma 1 , V ikas T ripathi 1 , V ijay Singh 2 1 Department of Computer Science and Engineering, Graphic Era Deemed to be Uni v ersity , Dehradun, India 2 Cisco-NUS Corporate Lab, National Uni v ersity of Sing apore, K ent Ridge, Sing apore Article Inf o Article history: Recei v ed Apr 4, 2024 Re vised Mar 17, 2025 Accepted Mar 26, 2025 K eyw ords: Centrality graph analytics Information e xtraction Kno wledge graph Le gitimate publishers Predatory criteria ABSTRA CT Academic writing is an inte gral part of scientic communities. This is a for - mal s tyle of writing used by researchers and scholars to communicate critical analysis and e vidence based ar guments. This w ork sho wcased a graph-based approach for scraping, e xtracting, representing and e v aluating the a v ailable aca- demic writing for gery detection criteria and further enhanci ng the model by proposing a set of ne w age criteria. The proposed w ork is based on kno wledge graphs and graph analytics capable of selecting subset of 16 criteria from the a v ailable superset of a cent of criterias pro vided by Bealls, Cabells, Shreshtha, and Think.Check.Submit, Scopus, and other rele v ant authors. The process for detecting the inuencial parameters cons ists of 04 phases: dataset preparation, kno wledge graph representation and making inferences through graph analyt- ics and e v aluation of results. The e xperimental results are then compared to the retraction database that consisting of information about retracted articles. The w ork enables the construction of an e xperiential kno wledge graph that ef- fecti v ely identies inue ntial criteria, enhancing this list by incorporating ne w age criteria into current inuential set and concluding in result by successfully detecting the academic predatory beha vior . This is an open access article under the CC BY -SA license . Corresponding A uthor: Garima Sharma Department of Computer Science and Engineering, Graphic Era deemed to be Uni v ersity Dehradun, Uttarakhand, India Email: g arima vrm91@gmail.com 1. INTR ODUCTION An unethical prac ticing journal [1] commonly kno wn as predatory journal gets associated with some suspicious publisher called as predatory publisher [2] represents an e xploited publishing model in academics. The characteristics of a predatory publisher comprise e xpedited re vie ws, lacking professional re vie w mecha- nisms, decepti v e impact f actors, f alsely listed respected scientists on editorial boards, an e xtensi v e repository of articles, journal titles that mimic those of reputable journals, and persistent spam in vitations ur ging arti- cle submissions [3]. Predatory publishing has become more widespread issue that is ne g ati v ely impacting academics as well as research inte grity , and therefore dissemination of inappropriate kno wledge in dif ferent sectors [4]. One major concern f acing the academic research community is the proliferation of misinforma- tion and disinformation resulting from unethical publication practices. In the current en vironment, publishing houses frequently o v erlook le gitimate content concerns in f a v our of commercial considerations. The y claim to adhere to genuine academic protocols for closely e xamining research, b ut the y routinely generate articles that are poorly produced, f all outside of their purvie w , and contain glaringly frequent errors or re v ersals of impact J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
974 ISSN: 2502-4752 f actors. It erodes condence in scientic publications as a result. A pioneering compan y , named retraction w atch [5], is at the forefront of a re v olutionary initiati v e set to identify and retract predatory publications. This compan y has coined a ne w term, “paper mill, and has designated Hinda wi as a leader in the paper mill industry [6]. The disco v eries made by independent researchers indicated that the inltration of Hinda wi special issues occurred on a lar ger scale than initially e xpected. Thus making it one of the parameters for the identication of ne w age predatory . The analysis concluded with retract ion of more than 8,000 papers ha ving included se v eral parameters for predatory identication including issues in scope, research description, data a v ailability , cita- tions, coherence, and peer -re vie w inte grity , indicating potential problems with the quality and reliability of the reported research [6]. W ith e v ery research in this area, the author does encounter a fe w common and impor - tant aspects to easily nd the suspicious ones. The common practices emplo yed by such predators consititues unsolicited ‘spam’ emails, char ging author’ s high publication fees without conducting thorough assessments of articles for their quality and le gitimac y [7]. Predatory publishers e v en emplo y tactics such as distorting peer re vie w processes, misrepresenting editorial services, and f alsely claiming database-inde xing statuses [8]. From f alsifying the inde x es to for ging the impact f actor v alues with high arti cle processing char ges (APC), the a w areness between correct inde xing, ranking, editorial boards, and membership. Can help the researchers to refrain from ille gitimate journals as well as publishers. Ce rtainly , these j ournal types are inf amous for emplo y- ing decepti v e tactics to entice researchers into submitting manuscripts, later imposing e xcessi v e APC before publication leading to decei ving no vices [9]. Re g ardless of their mode of operation, open or not, the predatories lags in fullli ng the lack of le g al and essential editorial as well as publishing services. Earl y career researchers (ECR) are especially prone to f all victim to thes e tactics, gi v en the challenges the y f ace in securing emplo yment and promotions [10]. In 2017, Bealls shared a report to standardized a set of criterias to cate gorize predators [11]. He continues the updation of this set for another 5 years b ut discontinued this due to backslash from v arious publishers and a fe w other unkno wn reasons [4]. The establishment of Cabell’ s whitelist and blacklist in 2018 [12], Jiban Shrestha’ s set of predatory criteria in 2021 [13], re gularly updated criteria from Scopus [14], other authors [15] and [16], and public research communities [17] are just a fe w of the ongoing ef forts to identify predatory publishing that ha v e surf aced since Bealls shutdo wn. The practice of publishing has signi cantly increased as a reason of educational reforms in v arious de- v eloping countri es [18], with increasing rates ranging from 10% to 16% [19] and e v en higher today . Countries hea vily implicated in these unethical practices include the US, China, German y , and the UK. A core-periphery netw ork dynamics may be e vident among de v eloping nations [20]. A substantial reason for this rise is particu- larly notable in countries that ha v e implemented signicant structural funding reforms in the past tw o decades, such as China (2002), Norw ay (2003), Russia (2005), and German y (2006). The e xamination of the timeline of distractions and missed opportunities since Jef fre y Beall alerted to the risks associated with pseudo-publishers and identied the majority of those operating at that time is been highlighted by Do wnes [21]. Using a combina- tion of Beall’ s list and predatory publisher data supplied by researchers, an online plug-in from ispredatory .com emplo ys cro wdsourcing [22]. Users can retrie v e a manually updated list of v eried predatory publishers and search for publishers by name, URL, title, or journal ISSN. According to Cabells’ Predatory Reports database in 2021, around 15,000 predatory journals were acti v e, leading authors to collecti v ely pay hu ndre ds of thou- sands of dollars to publish their papers. The In v estig ation of the incursion of journals with suspected predatory practices into the citation database Scopus and e xplores v ariations across countries in scholars’ lik elihood to publish in such journals is sho wcased by Mach ´ a ˇ cek [23]. Prakash et al. [24] e xplores potential predatory jour - nals and those with poor scientic standards by analyzing citations to 124 such journals in Scopus. This study e xplores the geographic location, publi cations, and citations of citing authors. The ndings indicate that the characteristics of citing authors ha v e a close resemblance with those of the publishing authors in these journals. In one of the w ork [25], the author e xplores mentoring approaches for guiding graduate students in a v oiding predatory publications and dubious conferences. These conferences often of fer swift manuscript re vie w pro- cesses, commonly omitting the f act that the y de viate from standard peer -re vie w protocols [8]. There are ne w approaches that authors are nding no w adays to detect predatory publications. An open automation system for identifying predatory journals is been proposed by [26]. This AI-enabled system uses Feature Extraction and a Bag of w ords algorithm to distinguish between le git imate and predatory publishers. In one e xample, the connections between indi vidual articles and predatory/le gitimat e publishers and journals are analyzed while emplo ying a data-dri v en training model named PredCheck [26]. F or an y researcher , therefore, it is a matter of utmost importance to de v elop the right understanding of dif ferentiating betwee n ethical and unethical publishers. Identication of predatory publishers can be done Indonesian J Elec Eng & Comp Sci, V ol. 39, No. 2, August 2025: 973–986 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 975 using v arious parameters such as editors’ suspicious role during publication [27], manuscript writing le v el features [28], claimed to be peer -re vie wed [29], sending emails to resea rchers in an attempt to publish articles [30], f alsely claiming to maintain adequate quality control, pro viding subpar editorial services and inadequate cop y editing, all while imposing undisclosed and e xcessi v e publication fees on the researchers [31], etc. This study p os sess a comprehensi v e comprehension and recognition of k e y parameters for swiftly and ef fortlessly identifying predatory practices. It is vital to identify the source s and tak e appropriate action on a global and re gional scale ag ainst publishers eng aged in ille g al or unethical practices. While other studies conned their reliability solely on Beall’ s list or completely disre g arding it is insuf cient to address the issue ef fecti v ely . Numerous researchers are acti v ely in v estig ating the k e y parameters of research publication fraud. T eix eira Da Silv a [27], Machacek primarily relies on the Scopus database to identify and label journals as predatory . Ho we v er , this analysis o v erlooks the inte gration of other reputable databases such as W OS, UGC, MAKG, Publons, and predatory databases lik e Beall’ s. Do wnes [20] e xplores the lik elihood of a journal being both open-access and predatory simultaneously . The researc her depends on Beall’ s library and four prominent databases—W eb of Science, Scopus, Dimensions, and Microsoft Academics—as the primary and e xclusi v e means of comprehensi v ely analyzing and cate gorizing open-access journals as predatory or not. The author hea vily leans on Beall’ s library , b ut this reliance is considered inaccurate due to the library’ s f ailure to pro vide scientic reasoning for cate gorizing an y journal into the distrustful cate gory . Instead, Beall’ s library is criticized for presenting a list of baseless alle g ations when assessing journals, publishers, and v arious de v eloping re gions such as Asia and Africa. There are se v eral parameters in circulation claimed to be ef fecti v e in the identication of predatory in academic writing. Each author has presented a distinguished methodology and set of criteria. Pro viders include Beall, Cabell, Public Research Communities, Shrestha, Scopus, and Think.Check.Submit and other authors. A comparati v e study between Bealls and Cabells has been demonstrated in T able 1. T able 1. Comparati v e analysis of tw o prominent predatory criteria pro viders P arameter Bealls Cabells Output List of journals/publishers practicing predatory/suspicious practices Extensi v e information about v arious journal types, their suitability , range of quality metrics Last update 2017, 2021 Up to date Subscription No Y es Usability Predatory practi ce only Suitability of journal/publisher for publication. F ocus Predatory or possible predatory journals/publishers Ev aluation of journals on metrics Methodology V ague W ell-dened Metrics Non-systema tic Systematic A v ailability Free to acces s P aid access Maintainability P assi v e Acti v e Commercial service No Y es Since there are hundreds of w ays specied by dif ferent criteria pro viders hence there is a need to summarize them and nd the most inuential ones that can guide the researcher at early stage of publication as this has not be e xplicitly addressed to date. The present w ork aims to de v elop such intuiti v e set of criteria from the collection of more than hundred criteria pro vided by abo v e me ntioned authors. Our proposed solution in- v olv es a model dening and utilizing predat ory and le gitimate criteria constructed from dif ferent le gitimate and predatory journal websites using web scraping of websites of dif ferent criteria pro viders. The collected data is then pro vided the weights based upon the frequenc y of their occurrence at dif ferent instances so collected using the higher the occurrence, the higher the weighting methodology . Thi s weighted m atrix is used to const ruct a kno wledge graph [32], [33] to obtain a consolidated graph using t riplet as Le v el, P arameter , P aram Pro vider . W e analyzed this graph using centrality analytics [34] to nd the most inuential nodes among these. W e concluded our w ork by specifying 16 criteria inuencing more than 100 of the criteria pro vided by dif ferent pro viders at dif ferent le v els. W ith this recommendation, we ha v e also identied 12 ne w parameters to enhance the o v erall model for the identication of ne w-era suspicious journals. The remainder of this paper follo ws a sequential structure k eeping the rst section in order is pre- sentation of the main theoretical concepts related to le gitimate and ille gitimate publishers characteristics along with the w ork done by dif ferent authors so f ar in detecting the predatory . In the ne xt section, we then intro- duce our proposed research methodology of e xtracting the parameter , de v eloping a kno wledge graph, nding the inuential parameters and e v aluating the results with and without ne w parameters. In the third section, CriteriaChec k er: a knowledg e gr aph appr oac h to enhance inte grity ... (Garima Sharma) Evaluation Warning : The document was created with Spire.PDF for Python.
976 ISSN: 2502-4752 we then sho wcased the e xperimental results and the discussions displaying the inuential parameters and a winner of ille gitimate spreader . In the last section, we ha v e presente d the conclusion and future w ork on the e xtension of the presented w ork. 2. METHOD In this w ork, the criteria checking model is prepared and implemented to identify the inuential param- eters from a set of predatory detection criteria established by v arious authors. The proposed model, illustrated in Figure 1, comprises of four primary phases, each encompassing a distinct set of tasks. Figure 1. Proposed o v erall architecture W e started the data collection process through W eb Scraping approach. The authors’ data has es- sentially been web-scraped programmaticaly using re ge x and other functions which then been correlated with their a v ailable parameters. The e xtracted data w as subsequently correlated with the a v ailable parameters to en- Indonesian J Elec Eng & Comp Sci, V ol. 39, No. 2, August 2025: 973–986 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 977 sure coherence and v alidity . This informati on w as then structured into a kno wledge graph, wherein each node represents the e xtracted parameters, while the edges denote the weighted relationships between the le v el type and the criteria pro vider . The le v el types analyzed in this study include journals, publishers, and conferences. Through graph analytics, our ndings indicate that the highest de gree of for gery occurs at the journal le v el. F or this purpose, the comprehensi v e analysis has been systematically di vided into four major sections: 2.1. Data pr eparation The de v elopment of a rob ust kno wledge graph requires met iculous data preparation to ensure accu- rac y , consistenc y , and semantic richness. Initially , the scope of the graph w as dened by identifying rele v ant entities, relationships, and attrib utes aligned with the intended application domain. The typical three steps to achie v e this are as follo ws: 2.1.1. Identication of criteria pr o viders This pi v otal stage in the entire algorithm’ s operation and design serv es as a crucial data collection uni t. During this step, v arious automated techniques are emplo yed to retrie v e the signicant parameter pro viders for subsequent parametric analysis. W e ha v e de v eloped our web scrapping engine and e xtracted the details of v arious authors w orking on parametric designs for predatory publications. Since the data g athere d w as huge in size we prepared and sa v ed the same in the graph database, required for graph analytics. Further , we e xtracted and added the characteristics of v arious predatory publishers using web mining techniques and identied k e y parameter pro viders useful for the model design and further analysis. 2.1.2. Actual parameter extraction Using mark et bask et analysis, we ha v e identied the frequent characteristi cs and common parameters as discussed by dif ferent authors in the pre vious step. The actual parameters e xtracted is more than a 100 while a fe w of them has been listed in T able 2 belo w wherein J stands for journal le v el and P stands for Publisher le v el f alsication. T able 2. Glimpse of e xtracted parameters (P) catalogue as pro vided by dif ferent pro viders PNo Le v el PDetails Bealls PRC Cabells Shreshtha Scopus Think.Check.Submit Others 1 J/P Soliciting authors for publication vi a emails F T T T F F T 2 J/P Luring authors for f ast publication vi a emails T T F T F T T 3 J/P Requesting high/v ery lo w publicat ion char ges after re vie w T T F T F T T 4 J Claimed to be peer re vie wed F T T T T T T 5 J Short re vie w timing T T T T T T T 6 J/P Bogus impact f actors are GIF , Inde x Copernicus v alue, Citef actor , UIF F T F F F F F 7 J/P F alsify le gitimate impact f actors F T F F F F T 8 J/P V eried impact f actors are Google, Di- mensions, and W eb of Science F T F F F F F 9 J Editors lack/for ging qualications in the eld F T T F F T T 10 P Dif ferent journal, single publi sher , same editorial board F T F F F T F 2.1.3. Fr equency weight assignment This section forms the core of the operational principle for the proposed and implemented model. The minimum weight assigned to an y parameter is 0 and the maximum goes up to 3. The weight depends upon the pre vious section as the frequenc y of characteristics present in a parameter pro vider’ s catalog is directly proportional to the weight assigned to it. x = n X i =0 x i (1) Using abo v e, the updated table has been sho wcased in T able 3. CriteriaChec k er: a knowledg e gr aph appr oac h to enhance inte grity ... (Garima Sharma) Evaluation Warning : The document was created with Spire.PDF for Python.
978 ISSN: 2502-4752 T able 3. Glimpse of updated parameters (P) catalogue after weight assignment PNo Le v el PDetails Bealls PRC Cabells Shreshtha Scopus Think.Check.Submit Others 1 J/P Soliciting authors for publication via emails 0 1 3 1 0 0 1 2 J/P Luring authors for f ast publication via emails 1 1 0 1 0 1 1 3 J/P Requesting high/v ery lo w publi cation char ges after re vie w 1 1 0 1 0 1 1 4 J Claimed to be peer re vie wed 0 1 2 1 1 1 1 5 J Short re vie w timing 2 1 3 1 1 1 1 6 J/P Bogus impact f actors are GIF , Inde x Copernicus v alue, Citef actor , UIF 0 1 0 0 0 0 0 7 J/P F alsify le gitimate impact f actors 0 1 0 0 0 0 1 8 J/P V eried Impact F actors are Google, D i- mensions, and W eb of Science 0 1 0 0 0 0 0 9 J Editors lack/for ging qualications in the eld 0 1 2 0 0 1 1 10 P Dif ferent journal, single publ isher , same editorial board 0 1 0 0 0 1 0 2.2. Kno wledge graph and infer ences In the ne xt step, a global kno wledge graph w as constructed by e xtracting entities, relationships, and attrib utes from curated datasets, follo wed by data cleaning, normalization, and alignment of the ontology to ensure semantic consistenc y . The processed data were transformed into a graph-based representation, enabling a structured inte gration of the collected heterogeneous sources. In addition, centrality measures (e.g., de gree, betweenness, and closeness) were applied to assess the relati v e importance of nodes and parameters within the graph. These analytics f acilitated the identication of k e y entities inuencing netw ork connecti vity and supported subsequent inference generation. The detailed steps are mentioned belo w: 2.2.1. Kno wledge graph pr eparation A structured representation of captured information from the abo v e sections is prepared using a kno wl- edge graph approach wherein each le v el is a node and all the weights act as edges to the node. The typical algorithm follo wed in the de v elopment of the weighted graph is gi v en in Algorithm 1. Algorithm 1 Create a weighted kno wledge graph 1: Input: Data D containing entities and relationships. 2: Output: Graph G with weighted edges. 3: Initialize an empty graph G = {} 4: Step 1: Extract Entities 5: Extract the set of entities E from the input data D 6: f or each entity e E do 7: Add node e to the graph G 8: end f or 9: Step 2: Identify Relationships Between Entities 10: f or each pair of entities e 1 , e 2 E do 11: if e 1 ̸ = e 2 then 12: Calculate relationship strength w between e 1 and e 2 13: if w > threshold then 14: Add edge between e 1 and e 2 with weight w 15: end if 16: end if 17: end f or 18: Step 3: Retur n the Graph 19: Return the graph G W e visualize the graph obtained in the spring layout as sho wn in belo w Figure 2. Indonesian J Elec Eng & Comp Sci, V ol. 39, No. 2, August 2025: 973–986 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 979 Figure 2. Kno wledge graph representation about the relationship between criteria and le v els 2.2.2. Centrality analytics among parameters Centrality analytics helps in measuring the inuential nodes within a graph and helps to identify es- sential edges within a netw ork of information [34]. The higher de gree of centrality means the higher connected node. The highest de gree nodes ha v e been k ept in the se v ere cate gory , the a v erage lik ed nodes ha v e been put into the moderate cate gory , and the least link ed into the lo west cate gory . Figure 02 sho wcases ho w the journal type node becomes the center node of all the parameters where publisher -le v el and indi vidual-le v el nodes ha v e a fe w connections. d = N X i =0 d i / N (2) Wherein, d i = number of edges connected to node i, N = total number of nodes. The proposed met hod tended to observ e the cate gory of le v el at which for gery is happening. While analyzing the updated parameters it has been assessed that these cat e gori es of for ging belong to a specic le v el such as journal or publisher or indi vidual wherein fe w of them are common between these le v els. K eeping this in vie w , a kno wledge graph so that the relationship connection between these parameters can be performed. Our graph triplets < subjects, predicate > consisted of < le v el, parameter , each parameter pro vider > ha ving weights assigned to each parameter as per the T able 3. The combination of these triplets formed a netw ork of interconnected information is sho wn in Figure 2. 3. RESUL TS AND DISCUSSION Our proposed approach through centrality analytics of kno wledge graphs, w as capable of nding the most inuential parameters that directly promotes the publication for gery . As there are so man y criterias pro- posed by indi vidual authors, this approach benets the no vice to look upon only 16 such inuencial highly weighted parameters includes Soliciting Authors for publication via emails, luring authors for f ast publication via emails, requesting high/v ery lo w publication char ges after re vie w , short re vie w timing, citef actor , UIF , f al- sify le gitimate impact f actors, v eried impact f actors are Google, Dimensions and W eb of Science, irre gular publication frequenc y , rapid increase in the publication in recent year , claimed open access, no dened Cop y- right Polic y/License, parent compan y information hidden, dead links on journal/publisher website, presence only in pre-print serv ers and disciplinary repos, authors and publishers are cross countries, lack of transparenc y in editorial board de v elopment, restricted F ocus on some countries. Additionally this has been observ ed that the majority of these belongs to the journal le v el, the model concluded journal le v el is the highest contrib utor to academic writing for geries. Figure 3 sho wcases the e xclusi v e dif ferent de gree analytics present between these parameters. CriteriaChec k er: a knowledg e gr aph appr oac h to enhance inte grity ... (Garima Sharma) Evaluation Warning : The document was created with Spire.PDF for Python.
980 ISSN: 2502-4752 Figure 3. De gree analytics between all parameters 3.1. Cor e entities of k ey parameters of intelligent help system in the kno wledge graph Further , e v aluating the centrality analysis, it w as identied that the ‘journal-le v el vulnerability’ is the most central f actor in the kno wledge graph with a centrality score of 0. 8. This points out to centrality which is high and it means that this particular product plays a v ery important role within the netw ork of predatory journals. Others include the follo wing parameters; ‘soliciting authors being 0. 7, ‘high publication char ges’ at 0. 65, ‘bogus impact f actors’ at 0. 7, and ‘f alsied editorial board’ at 0. 55 as presented in Figure 4. The f act that these parameters are so highly rated points to their importance in the functioning of these unscrupulous journals. Figure 4. Centrality of k e y parameters in obtained kno wledge graph 3.2. Le v eled parameters kno wledge graph fr om multiple pr o viders The le v eled paramet ers kno wledge graph as sho wn in abo v e Figure 5 sho wcase the identied f actors and their interconnections, pro ving that predatory journal acti vities are a multif aceted issue. The comple xity of these netw orks is signicant, since the central practices include man y f actors that depend on each other . This mak es it v ery hard to point out the predatory journals and therefore there is the need for a multi-pronged strate gy to deal with this issue. Indonesian J Elec Eng & Comp Sci, V ol. 39, No. 2, August 2025: 973–986 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 981 Figure 5. De gree of centrality for inuential parameters 3.3. Effect of ‘jour nal-le v el vulnerability’ on the structur e of a kno wledge graph Based on the analysis of Figure 4, one can conclude that ‘journal-le v el vul n e rability’ has a high le v el of impact f actor on the general formation of the kno wledge graph. The absence of this parameter af fects the graph’ s connecti vity and its topological structure, thereby underlining its role in the predatory journal en vironment. 3.4. Extended description of the most signicant v ariables and their interdependence Thus, Figure 6 is an analogous representation that sho ws the detailed interactions between the para me- ters, thereby elucidating the relations within the kno wledge graph. Such specics contrib ute to the vie w of ho w precisely relational elements lik e ‘soliciting authors’ and ‘high publication char ges’ connect and contrib ute to the o v erall netw ork. Figure 6. Comparati v e analysis of parameters from dif ferent criteria pro viders 3.5. Comparing centrality scor es of the nodes concer ning differ ent parameters and v alidation T o v alidate the proposed model selected 16 inuential criteria, a repository of retracted papers has been prepared using retract ion database [6], [7] wherein a random e xtraction of 50 retracted paper details such as w ork title, author name, journal title, publisher title, year of its publication and year of retraction is done. Further , the criteria pro vided by dif ferent authors is check ed to measure the accurac y of predatory identication using a set of 16 inue n t ial criteria and a hundred-plus a v ailable criteria pro vided by dif ferent authors. The similarity score found between randomly e xtracted lists from Bealls and randomly e xtracted list retraction database is only 5% stating the commonness present between the random sampling carried out between the tw o sets. This score subsequently increased to 30% between Bealls and Shrestha’ s proposed w ork. CriteriaChec k er: a knowledg e gr aph appr oac h to enhance inte grity ... (Garima Sharma) Evaluation Warning : The document was created with Spire.PDF for Python.
982 ISSN: 2502-4752 Similarity Score = Set i Set j (3) Where i and j are tw o randomly generated sample sets from a dif ferent repository . Upon analysis thoroughly with acti v e suspicious journals, a set of ne w emer ging parameters has been proposed that can be utilized for catching the fraudsters at the three dened le v els along with the specied inuential parameters set. T able 4 describes t he ne wly identied parameters raising a bo w to w ards suspicious predatory acti vity at dif ferent le v els e xclusi v ely seen no w adays. T able 4. Ne wly identied parameters for intercepting ne w era predatory S. No. Ne w age predatory criteria Description 1 Sho wcasing b usiness address of de v eloped countries, major editors are from de v eloping countries The of cial address on the website of the publisher or journal is claimed to be from a de v eloped country wherein all the editors under the publisher belong to de v eloping countries. 2 Self-citations The self-citation of the indi vidual author , journal, publisher , society , and institutions. 3 Evidence of multiple publishers in a single journal Single journal title is claimed to be part of man y publishers. 4 Single publisher with dif ferent numbers of journals on dif ferent websites As man y online addresses are a v ailable for a single publisher , there could be dif ferent numbers of journal listings at dif ferent addresses. 5 Single journal title with tw o dif ferent ISSN numbers Dif ferent ISSNs are listed at dif ferent web links. 6 Common editor among all the journal-title of a single publisher Single editor for all the broad areas starting from medical to engineering as well as educational and social. 7 Journal is publishing articles without ISSN No ISSN w as present e v en after the publication of the article 8 All the associated journal titles are accepting ne w manuscripts throughout the year Besides special sessions, the journal is ready to accept a ne w manuscript and release it in special editions with higher APCs. 9 Publisher re ady to pro vide membership with- out author af liation The membership is generally free of cost and can be tak en without mentioning an y af liation the author is associated with. 10 F alsely claiming under Scopus after e xpiration Journal is present in the discontinued list of Scopus and on the website, it is claiming to be Scopus. 11 No editorial board is listed on the website The journal or publisher’ s website lacks in pro viding information about their editors. 12 Dif ferent journal name on the publisher’ s website and journal website The journal name on the publisher’ s website is dif ferent and on opening its web link the name and details are dif ferent. T o measure the accurac y of predatory identication by studying and analyzing dif ferent criteria present so f ar , a ne w term called strength score has been coined here. A higher strengt h score means that an y one of these is capable enough to identify the maximum of the predatory present in either list. S = np X i =0 a (4) a = p i pr i (5) Where, S = strength score, n = a positi v e inte ger v alue, a = an inte ger score assigned to an indi vidual parameter , np = number of identied parameter , pr = parameter title, p = publisher title, i = a positi v e inte ger v alue. The system attains a nominal strength score of approx. 40% of these randomly prepared lists are matched with 16 inuential parameters so e xtracted. The o v erall ef cienc y of the system increases by 20% if we incorporate the ne wly identied parameters. Consequently , incorporating the ne wly identied parameter with inuential paramete rs forms an ef fecti v e system to detect predatory journals. Comparing the centrality scores of all the parameters is presented in Figure 7, b ut that bar chart sho ws the importance of each parameter clearly . Comparison of data collected on v arious institutions and parameters indicates which f actors ha v e the most impact and hence should be gi v en undue emphasis while trying to put mechanisms in place to address the issue of predatory journals. The present study e xplored a comprehensi v e approach to detect the 16 inuencial Indonesian J Elec Eng & Comp Sci, V ol. 39, No. 2, August 2025: 973–986 Evaluation Warning : The document was created with Spire.PDF for Python.