Optimization in Supply Chain Management , the Current State and Future Directions-A Systematic Review and Bibliometric Analysis

Purpose: The purpose of this paper is finding the current state of research and identifies high-potential area for future investigation in optimization in supply chain management. Design/methodology/approach: In this paper we present Bibliometric and Network analysis to examine current state research on optimization in supply chain management to identify established and emergent research field for future investigation. The systematic research review which we used in our study have not grasp or assess by other researchers on this topic. Firstly, based on our methodology Bibliometric analysis began by identifying 1610 publications raised from scientific journals, included literatures from 1994 to March of 2016. Secondly, we applied PageRank algorithm in our data for citation analysis to indicate the significance of a publication. Thirdly, the topological decision variables analysis is done based on Louvain method for network data clustering, for this proposes we used the rigorous tools. Finding: Based on our Network analysis result, the optimization in supply chain management research can be divided into four clusters/modules that introduced fundamental skill, knowledge, theory, application and method.


Introduction
In the real world, business is highly competitive and dynamic.The rapid rate of innovation, technology, globalization and consumers expectations are modified the type of worldwide market competitions from traditional supply chains to new competitive supply chains (Scs) (Ponomarov & Holcomb, 2009).The past recent decades, the growing role of global supply chains was associated with increased interconnectedness among suppliers and manufacturers, which led to higher dependency among firms in the supply chains and a higher level of supply chain complexity.
Supply chain Management (SCM) is monitor processing and system implementation to manage the service and flow of the goods in order to capture maximize the efficiency and value added in SCs.Also SCM spans all operation, storage of raw materials, work-in-process inventory and finishing the goods from starting stage to customer point.Now supply chains network have faced challenges like as high demand variability, short life of products, and different expectations and requirements of customers; adapting to these challenges increased supply chain complexity and resulted in more instability and unpredictability (Stefanovic, Stefanovic & Radenkovic, 2009) SCM encompasses three decision levels: strategic, tactical and operative.In particular, in strategic stage, supply chain design comprises the making decisions following to the number and location of production and storage facilities, the amount of capacity at each facility, the conciliation of market demand analyzing and decision making on supplier selection to check out total cost feature (Chopra & Meindl, 2004).also they mentioned 37 papers to classified to supply chain decision making to addressing in different typical parts in supply chin such as rout, procurement and inventory.They showed that combination of inventory and production making decision is frequently use in facility location and allocation cases.In another part they taxonomy 18 papers in different network structure respected to closed loop or reverse logistic activity in different layers (Huscroft, Hazen, Hall, Skipper & Hanna, 2013).They reach this conclusion just few papers concentrated on both closed-loop and forward logistic models comprehensively.

Research Methodology
The aim of a systematic literature reviews is to direct toward the map of literature to distinguish the gap research to transparent the edge of sciences (Su, Chen & Yang, 2016).Key words search, literature searching and using different type typical analyzing made the structure of literature review which made by an iterative cycle processing (Saunders, Lewis & Thornhill, 2009), also (Pazhani, Ventura & Mendoza, 2016) suggested designing bibliography methodology to building mind road map for structure the literature review.As this approach also we use these steps, introducing the suitable search terms for data collecting and reach to initial result, after refinement and elimination, the result will send to data statistics analysis, at the end comprehensive data evaluation will be done to reach clustering classification to show direction for future scholars in this field.

Introducing the Appropriate Research Terms
Six main keywords are used to starting for our research and data collection.Our keywords are "Supply Chain", "Framework Design", "Sustainable development", "Optimization OR Optimisation", "Network Design" and "Modeling OR Modelling".Combined these key words including, (1) "Framework Design" AND "Sustainable development" AND "Supply Chain", (2) "Network Design" AND "Sustainable development" AND "Supply Chain", (3) Optimization OR Optimisation AND "Supply Chain", (4) Modeling OR Modelling AND "Supply Chain".There are different decisions -making in supply chain network design but suppose that the most important one is facility locating in different layers of supply chain (Eskandarpour et al., 2015), in continue perhaps logistics system configuration, re-organization, outsourcing modelling and re-configuration are critical issue in supply chain networks optimization.Using different method of optimization for taking these strategical decisions are essential for supply chain network design.So authors believe that with employ of these keywords will completely cover the main interest.

Initial Research
For defining our research terms, using "Abstract, Given that the first debate on facility location and supply chin network is traced to before 1980s (Melo et al., 2009) it is not surprise that in our observation be the same as Figure 1.
Figure 1.Articles trend in the field of our review

Initial Statistic Data
Figure 1 shows the numbers of article trend in optimization in supply chain network.In our first survey shows that all the 1610 papers published in 501 journals, the 11 top journals which published at least 10 articles during in 22-years ago is appeared in Table 3.We found that 308 identified papers published in these 11 journals, it is about 20% of all papers publication.Also there are some more statistic data for each journal which we did not present in

Data Processing
Bibliometric is systematic way to help us to measure the impact of scientific publications.Also with context of this toolkit the impact of scholar and productivity is measured by the number of citation with bibliometric.Our data processing divided in two sections; first it started with bibliometric analysis and then network analysis which presented in section 4 and 5 respectively.We used BibExcel software package to bibliometric analysis, it could provide easily statistics data such as title, author, abstract, research area, topic and affiliation.BibExcel selected among the other software package because of its flexibility and compatibility with the some application like as Gephi and Pajek (Costa, Celano, Fichera & Trovato, 2010).
After bibliometric processing all data conduct to network analysis, Pajek (Dabkowski, Breiger & Szidarovszky, 2015) VOSviewer (Jie, Xiaohong, Shifei & Jovanovic, 2014) and Gephi (Beiler, 2016) are existing software package for literature network analysis.Gephi is exploration and leading visualization free open-source package for all type networks and could runs on different operating system.We chose Gephi as network analysis software due to high capability to visualization in various graph (Wehbe, Hattab & Hamzeh, 2016) and efficiently to work with dataset.
Furthermore based on our requirement for future steps, we were looking for powerful software with high level degree flexibility to analysis comprehensive information and convert input data from different online database like as Web of Science and Science of Direct systematically.Also it should have enough ability to generated data output files can be imported to Microsoft Excel and network analysis software in Gephi.BibExcel designed to easy assist a user in bibliometric analyzing data, or any data of a textual nature formatted in a similar manner.This toolkit enables the generation of data files that can be imported to Excel or any program that further processes or visualizes tabbed data records.As above explanation we have chosen BibExcel software for our study.
The output searching in Web of Science has different file format, we selected RIS format with full record content which BibExcel could extract the bibliographic information.Our study is concentrated on following information: Journal name, location of organization, authors name, key words and reference.In continue 3 below section will show the output files of BibExcel analysis to provide statistic data.

Ranking of Authors in Our Study
The input data of bibliographic can be sort and analysis by BibExcel; it has some toolbars which help us to extract different data from RIS file.

Affiliation
One of the ability of BibExcel is extracting the city of each organization from RIS files.For this order, we sort our data based on "AD tag" in BibExcel and set the other toolbar as BibExcel comment, the output result will be a window which present the affiliation of authors in one column.
We could easily export this data to Microsoft Excel for future process.After this stage the analyzed data conduct to GPS (Global Positioning System) visualizer web site for built multiple geocodes, for this purpose we used the "gpsvisualizer.com"web site.It is free GPS visualizer online to make maps from geographic data, there is some limitation to draw a map for multiple geocodes but with asking keyword mapping from Google Map or Bing Map website we could create our geographical map with large number of cities. Figure 2 shows the map which created by "GPSvisualizer" online website, it shows the institutions location which working on optimization in supply chain network.The large numbers of literatures is located in western country in Europe and then west states in USA, also the numbers of publication in west of Asia is impressive.In Figure 2 the diameter of each circle express the relative degree of each institution to the contribution.In general view, the map geographical distribution of these institutions expresses that location facility in supply chain network attracted many researcher around the world.

Most Common Words Statistics
In this section we analyzed our data based on most popular words and phrases which used in the title/keyword of articles.The top 13 words used in the title and top 12 words or phrases used in keywords of papers are indicated in Table 7 and Table 8 respectively.With the compare of these tables we observe uniformity of the nature between them, such as: sustainable supply chain management, nonlinear programming optimization and life-cycle assessment.Clearly, it is not strange that all the chosen keywords in this research appear in both tables and also "Optimization" is most interesting word which appear is both lists, our implication is: in the supply chain, network optimization is absolutely necessary for effectiveness.

Network Analysis
Now after Bibliometric analysis we continue our study with network analysis in this section.Publication network analysis is the process of graphical investigation though the defined algorithms and graphical in Social Network Analysis, e.g.see (Doreian, 2006) and (Dabkowski et al., 2015) for more review see (Batagelj & Mrvar, 2011).Gephi is visualization, exploration and manipulating toolkit for many kinds of graphs and networks.It is open source software also we can count these main features for Gephi: analysis the Biological Network, ability to import many different file format such as GEXF, NET, GML and GDF (Gephi, 2013).For our study we selected Gephi as our network analysis software, because the design of this package is much broader performance to having a usual distinct toolkit, we doubted that maybe Gephi being user unfriendly but we found it designed very logical structure with ease functionality but user should understand the command sequences during to working with Gephi.
The bibliographic output file which getting from the Web of Science, is in RIS format, these files cannot use directly to Gephi software.For this order, we need one mediator software to create dataset, BibExcel used as operator for reformatted the data and present graphic dataset.For this purpose different information such as authors names and references extracted from BibExcel then based on the software instruction we continue to reach output files: ".NET" which it need for Gephi for network analysis.With the Web of Science bibliographic output we could conduct our study for citation analysis too.

Citation Analysis
Citation analysis is a systematic way of measuring the relative importance an author or a publication by counting the number of times that author and publication has been cited by other, for this purpose 1610 nodes is created to examine the grade of connection between each node.The primary output of Gephi revealed that near 45% of all publication cited to each.Table 9 shows 13 publications with most citation in Web of Science database."Average per Year" is the quantity number of local citation divided to the number of years from publication in Web of Science, "Reference Citation" show the number which each paper cited within 1610 paper-network, "Global Citation in all databases" means number of times which mentioned paper cited among all of Web of Science data base such as: "SciELO Citation Index" and "Chinese Science citation", and "Web of Science Core collection" is the overall Web of Science citation for publication.As it shown in Table 9, the rank of papers in "Global Citation in all databases" is more or less the same as "Reference Citation", it maybe means that that selection of keywords is done with proper way.
Another point is the huge difference numbers between these two columns indicated that optimization in supply chain management got more attention from authors in different disciplines.We see that in Table 9 ( Melo et al., 2009) dominated the list but following to Table 4 his name is not in the 15 top authors but also in this list we see Shah, Pishvaee & Torabi as 15 top authors.In Table 9 (Melo et al., 2009) is the top of all columns, it seems that, the invited review which made by Melo and S. Nickel, presented significant contribution of integration local decision with other strategic decisions and received huge attention between researcher.Also the average citation per year for this paper is near two times duo to second paper.
Pishvaee & Torabi who have co-authored represented three out of 13 high cited publication in Table 9, both of them also did these study in the University of Tehran which this university has highest ranking between in top institutes in our research study (see Table 6 and Figure 2).Both of them have engineering background and working on different type numerical experiments method optimization.Also it is not surprising to see that in Table 9, 4 out all top global cited papers published in "European Journal Of Operational Research" in Table 3 we saw that this journal has second top ranking in publication with 61 papers publication.Table 9 indicated all top cited papers at least published 5 years ago so maybe in near future some papers with high average citation per year will be replaced with in Table 9.Interestingly, we see that 8 out of ten top cited papers published in journal which their filed is Engineering/Mathematic, this evidence shows that perhaps in optimization for supply chain management still we need more intensive research as strategic supply chain planning due to solution method.

PageRank Analysis
Very usual method for understanding the level of paper is to counting the number of citation in different database which we indicated in section 4. 1, but also there is a evaluated model to measure average accuracy in in dataset.(Buckley & Voorhees, 2005) explained "MAP" -Mean Average Precision -based on probabilistic model (Huang, Huang, Wn, An, Liu & Poon, 2006) with citation graph using, e.g.see Figure 4. C, the papers represented as nodes and citation between the papers indicated as directed edge.
So, with apply linkage analysis algorithm in citation graph the ranking grade of each node/paper could be computed.There are three famous linkage analysis algorithm to measuring the publication significance  (Yin, Huang & Li, 2011): Degree Distribution, HITS and PageRank.Degree Distribution, this linkage analysis algorithm uses the definition of "Popularity of a document".It means the number of citation for each paper show the ranking of paper (Borodin, Roberts, Rosenthal & Tsaparas, 2005).For first time HITS (Hyperlink-Induced Topic Search) presented by (Kleinberg, 1999), the idea behind of this algorithm is computing the importance of deferent type of Web Pages.We have two type attributes for identifying the importance of Web Pages, hub and authority, hub attribute record the page quality as useful resources and authority attribute record the page quality as source itself.These two authorities could compatible to paper citation network analysis.In general world we can say that "PageRank C(T i ), presented the number of citation in T i .
N, the number of publication in citation network.
Attention if C(T i ) = 0 then in above formula, the number of publication will be instead of C(T i ).
Now if we consideration PageRank for P 1 , P 2 , …, P N , the equation is as fallow: (2) M(p i ) is the set of papers which cited to p i .
PR(p j ) PageRank value of p j .
In this case we need to introduce eigenvector.PR(p i ) numbers dominated in right of eigenvector, this made PageRank a practically proper metric.
The eigenvector is: (3) In Table 10 pretest the 10 top PageRank value in our study.With consider Table 9 and 10 we find that seven of publications are the same, these publications included, as new item in Table 10 ( Papageorgiou, 2009), (Costa et al., 2010) and (Prakash & Deshmukh, 2011).These paper published after 2005 and cited by high-cited papers.The interesting point is, not only these three papers (Costa et al., 2010), (Pishvaee & Torabi, 2010) and (Sarkis et al., 2011) located in both tables (Table 9 and Table10

Co-citation Analysis
Henry Small proposed the co-citation in 1973 (Small, 1973).Co-citation analysis is an assessment of similarity the publications which share to each citation.Definition of co-citation is to finding the relationship of two publication which cited to each by another paper (Wang et al., 2016) As indicated in respectively.Bibliographic coupling employ in citation analysis to make a similarity relationship between papers is a similar measure as co-citation (Wang et al., 2016).

Publication Classification with Clustering Method
Clustering is one the most critical approach in data mining analysis, clustering means the data fall into same module/cluster with more similarity than other (Chen & He, 2016).The case of our study, where represented papers in the citation network is indicated in a group with same research field, a cluster can be identified, into the cluster the density of edge and arrow is high.There are many fields that clustering algorithm extended in them, for example genetic engineering, customer segmentation and publication analysis, normally these type data are numerical and classified attributes (Hsu & Chen, 2007).The optimization of modularity tools in Gephi is under Louvain method, an iterative method to optimized modularity as the algorithm progresses.Modularity index is defined between -1 and 1 which present the density of edges inside due to outside of communities.According to (Uchimaru et al., 2015) modularity index is defined as: (4) Here is sum of link weight penetrating node i, is sum of link weight penetrating node j.
The "Kronecker" delta d(c i , c j ) is 1 and 0 when nodes i and j are assigned to the same community and .
With using of Force Atlas (as discussion in 4. 3 section), all connected nods move to the middle of the network, also the isolated nods move to the corners.With excluding outlier nodes in our network, it will remains with total of 233 node out of 1610 nodes.

Evolution of Clusters Over Years
To help understand the evolution of optimization in supply chain management evaluation over time, we also complete a dynamic co-citation analysis covering all papers in the four clusters.The evolution is graphically presented in Figure 5.The size of a circle represents the PageRank value of each paper, so the larger the size of a circle, the more highly-cited and prestigious the corresponding paper.It can be expected that the research papers in Clusters 3 and 4 will continue to grow in next decade and in other side we see stabilize in Clusters 1 and 2.

Conclusion and Some Suggestions for Future Study
This paper provided a systematic analyzed review of optimization within context the supply chain management.There are many literature reviews published within the scope of this topic, but there are a handful of researchers involved in bibliometric and network analysis to evaluate information and data clustering.Moreover, in this paper we identified classification, characterized and labeled our data (section 4.3.1) to show which part of this topic has been paid more attention, or rarely considered by researchers to guide the further investigation and strengthen SCM planning needs.
Based on our bibliometric analysis, our initial efforts presented the geographical global map.The dispersion of affiliation represented by Western Europe and North America has impressive publications, with Asia not very far behind.Perhaps in near future the rate of diffusion of the jobs will increase rapidly.
Looking to Table 6, this shows that 9 out of 17 top institutions belong to Asia (Turkey not included), while countries such as Iran, Taiwan, Singapore and china have more contributions.With this map and the information displayed in Table 6, the students and scholars who are interested in this topic could conduct their research at various institutes.
Furthermore, we found out that the majority of significant papers were publishing at the beginning of the last decade.Considering Figure 1 it is not surprising to see that in 2010 we faced a high rate of increase in publications regarding the supply chain management area.International Journal of Production Economics and European Journal of Operational Research are more popular journals in this field; we see that many scholars in United States and Asia attend to publish their papers in European journals.
Our network analysis conclusion shows that, optimization in supply chain management can be divided into four clusters/categories, labeling of the popular cluster (Cluster 1) is: Conceptual analysis and analytical modeling.We recognize that facility optimization problem may be getting saturated in this field.
The name of the second cluster, which is far behind is: Environmental sustainability and green supply chain network, articles which were published in this cluster are younger compared to the first cluster; it seems that green supply chain could be as an emerging area, as our research result experimental study and design strategy pays less attention in to this topic.
However, this type of research has some limitations and needs further investigations.
First, designing "Key words search model" to ensure capturing the most effective keywords and field area.
We can make a pool of papers but we should be sure to maximize coverage in our research.It is fundamental approach for this type review.Second, bibliometric and network analyzing is associated with a huge number of papers, so using professional software to enhance reliable results is necessary.Third, database plays the main role in systematic review, hence the importance of database should be considered, Scopus is relatively comprehensive due to others (Wang & Waltman, 2016), so to achieve suitable advantage, this database is suggested.We believe this paper can create an objective and professional understanding in order to navigate the further mapping and direction for broaden research.

Figure 2 .
Figure 2. (Red circles) graphical location of all institutions theories.For this purpose different software can be employ, but four toolkits are more familiar for researchers, these packages consist of: HistCite Graph maker, VOSviewer, Gephi and Pajek.Each of them has many toolboxes with different flexibility for example: VOSviewer is a comprehensive view on the scientific activities software for visualizing, mapping and clustering techniques in networks analyzing, it could import directly data from Web of Science and Scopus and supported RIS file.Gephi is an open source Windows software, proper for large network, it could analysis and visualization millions of vectors in any networks system.It implemented in Delphi (Pascal) and just accepted ".Net" format, beside the citation network and internet network, one of the most popular usage of Pajek is identifying social roles algorithm" is a new version of "Degree Distribution algorithm" which considers separated weight for each node/paper.In our study we used Page Rank algorithm as publication network analyzer.Page Rank algorithm presented by Larry Page and Sergey Brain in 1998.The primary application of PageRank was ranking web pages in Google based on the keywords searching to indicated page's relevance or importance and indicated the relationship between them, also, many authors study the Search Engine Optimization (SEO) topics based on Page Rank algorithm.These properties could the expanding to explore in citation network.The PageRank algorithms formula with N publication is: ), donated PageRank of paper A which has been cited by paper T i in the citation network.
) also they published after 2010.The outcome of this section given us the impact of citation from the high-cited publication is greatly important for PageRank algorithm and also can see that maybe PageRank give us better general view as prestige of papers.Although this topic is most popular between the researchers and discuss by expert to explore to find new solution for more accuracy in this field for example see(González-Pereira, Guerrero-Bote & Moya-Anegón, 2010)

Figure 3
Figure 3(b) Paper C and Paper D are co-cited by Paper A, and Paper B. So, Paper C and Paper D have cocitation strength of 2. Also Citation and Bibliographic coupling is shown in Figure 3(a) and Figure 3(c),

Figure 4 .
Figure 4. (a) The positioning of the four clusters without arc, (b) The positioning of the four clusters with arc

Table 1 .
Summarized some reviewed papers in related area

Research Key Words Result of Search (Papers) Result of Search after refinement (Papers)
Title and Keywords" in advance research of electronic bibliographical sources Web of Science, in this survey we refine or search based on "journal" and all the articles saved, but "Conference paper, Book and Chapter of book" is not included.Based on this refining, our initial research result achieve 2196 articles totally.The breakdown of research result is shown in Table 2. Web of Science Database web site could save essential information in different formats like as RIS and Plaintext.With the export the related file from Web site, we accesses the all information such as, references and author's names, Citations and abstracts.

Table 3 .
The 11 top journals with 10 or more publications in location facility in supply chain management area

Table 4 .
Table4shows the top 15 contribution authors with 8 or more publication in our records; also beside of each we mentioned the number of publication.It is clear to see that Shah and Tiwari are in high raking in this table and dominated over the list.The authors with 8 or more publication chain network, planning and scheduling, future challenging and infrastructure design especially in chemical industry.Tiwari with engineering background, study in computational Intelligence in manufacturing and logistics.Farahani and Rezapour contributed on facility location and distribution planning, transportation planning, graph theory, production operations management.In the middle of list Papageorgiou with 11 number publication works on these fields: chemical engineering, biotechnology, mathematics modeling and distributed energy resources.Less than 1% of all publication is provided by these 15 top authors.Majority leading scholars in this list tend to study in optimization in different section of supply chain network issue, such as network design, mathematical approach and production plan.Grossmann and Lam have published some papers in the field of process systems engineering and green sustainable and broad mathematical programming tools in sustainable processes.The other researcher usually trend to work on general Based on our records, this point should be consider Shah and Tiwari in large number of papers are co-authors, they cooperated in the publication as the second or third author, in the other hand Farahani and Rezapour co-authors with each other in many publication.Shah focuses on industry processing in supply modeling framework, empirical approach and environment issue.For example, Pishvaee used fuzzy theory in his case study and using multi-objective optimization in environmental supply chain network design.

Table 5
shows greater breakdown in geographical area in globe, like as Europe, USA and Asia.Also we add Oceania and Canada to this list, the remain of the country and the articles without affiliation records indicated in 4 and 5 row, respectively.

Table 5 .
Contribution of Institutions based on geographical areaThe 18 top institutions with most contributed for publication is mentioned in Table 6.Consider with this table and Table 4, we see that Imperial College London, Indian Institute of Technology and Amirkabir Journal of Industrial Engineering and Management -http://dx.doi.org/10.3926/jiem.2035University of Technology are presented productively scholars with Shah, Tiwari and Farahani.It should be noted during several years, different publication for one researcher may have submitted from different organization.

Table 6 .
The 17 top organizations with 6 or more publication

Table 7 .
The most common words in title of articles Table8.The most common words in keywords of articles

Table 9 .
Top authors with more global citation in Web of Science Core database