RANKING ASSOCIATION RULES BY TOPSIS METHOD

RANKING ASSOCIATION RULES BY TOPSIS METHOD WITH COMBINATION WEIGHTS
ZHEN ZHANG, CHONGHUI GUO
Institute of Systems Engineering, Dalian University of Technology
Dalian, 116023, People’s Republic of China
E-mail: zzupchd@gmail.com; guochonghui@tsinghua.org.cn
Various rules can be generated from databases by using association rule algorithms, but only a small number of these rules may be selected for implementation due to the limitations of resources. Thus, evaluating the quality of these rules becomes a hot topic in data mining field. Based on multiple criteria decision making theory, a framework for ranking the mined association rules using TOPSIS method with combination weights, considering both objective interestingness measures and the users’ domain information, is proposed. An example of market basket analysis is applied to illustrate the applicability of this method.
1. Introduction
Rapid advances in data collection and storage technology have enabled organizations to accumulate vast amounts of data. To extract useful information from large databases, data mining has become an essential tool for data analysis. In recent years, data mining has been widely applied in many fields, such as business, science and engineering (Tan, Steinbach, & Kumar, 2006).
Association rule mining, known as one of the most important techniques in data mining field, is useful for discovering interesting relationships hidden in large data sets. Various rules can be mined from databases by using association rule algorithms (Srikant, Vu, & Agrawal, 1997). But only a small number of rules may be selected for implementation due to limited business resources (Choi, Ahn, & Kim, 2005). Thus, evaluating the quality of these rules becomes a hot topic in data mining field.
Generally speaking, rule quality can be evaluated according to some interestingness measures, which consist of objective measures and subjective measures (Geng & Hamilton, 2006). Objective measures, which are mostly based on theories in probability, statistics, or information theory, just only consider the raw data, such as support and confidence used in association rule mining. Subjective measures which take into account both the data and users of these data, require users to input domain knowledge manually as constraints, or distinguish rules as interesting or uninteresting by interacting with the data mining system.
Most association rule mining algorithms consider these interestingness measures as constraints to obtain interesting patterns. But it may be difficult to select appropriate interestingness measures before data mining is performed. Thus, it is more appropriate to carry out post-processing for the mined association rules. An efficient method is to rank these mined rules based on some interestingness measures and subjective criteria in order to obtain rules that are more interesting to users. Multiple criteria decision making methods can be used to solve such problems. Choi et al. (2005) used ELECTRE-II approach combined with AHP for association rules prioritization which considered both objective criteria and subjective preferences of users. They believed the proposed method made synergy with decision analysis techniques for solving problems in data mining field. But this method may be less efficient when large number of association rules are generated. Chen (2007) developed their work and proposed a DEA based method to rank association rules. This method firstly uses a DEA model to obtain efficient association rules, and then applies another discriminated model to rank the efficient association rules. Obviously, this method needs to solve considerable numbers of linear programming models, and includes redundant computations and considerations. Based on Chen’s work, Toloo, Sohrabi, & Nalchigar (2009) used an integrated DEA model which was able to identify the most efficient association rule by solving just one mixed integer linear programming and proposed a new method for ranking association rules with multiple criteria. Both the two DEA-based methods can’t rank all these association rules and can’t reflect the importance of these criteria quantitatively, which is important for multiple criteria decision making. Besides that, the computation may be complex when the number of these association rules is very large. Thus, based on the previous work, this paper uses TOPSIS method with combination weights to rank the mined association rules synthetically and proposes a new framework for ranking association rules.
The rest of this paper is organized as follows. In section 2, association rule mining is described briefly. TOPSIS method with combination weights is presented in section 3, and section 4 introduces a framework for ranking association rules. Next, an example of market basket analysis is applied to illustrate the applicability of our method in section 5, and section 6 concludes this paper in the end.
2. Association Rule Mining
Association rule mining proposed by Ageawal, Imielinski, & Swami (1993), is a useful technique for discovering relationships between items in databases. It has been widely used in business applications such as cross-marketing, attached mailing and catalog design (Choi et al., 2005).
Let I= {i1, i2, .. , im} denote a set of items. Moreover, let D indicate a set of transactions, where each transaction t is a set of items so that t  I. A unique identifier, named tID, is associated with each transaction. An association rule can be formulated as the following implication relationship:
X  Y, where X  I, Y  I, and X∩Y=Φ.
The rule X  Y holds in the transaction set D with confidence, c, where c % of transactions in D that contain X also contain Y，and the rule has support, s, in the transaction set D if s % of transactions in D contain X∪Y.
The Apriori algorithm proposed by Agrawal et al. (1993) is one of the most widely used techniques for finding associations rules, which restricts search space by setting minimum support and confidence threshold and obtains relatively strong rules. The algorithm consists of two phases.
Firstly, all itemsets with minimum support, called frequent itemsets are generated, which can be described as follows:
(1) L1=find_large_1-itemsets;
(2) for(k=2; Lk-1≠Φ; k++) do begin
(3) Ck=apriori_gen(Lk-1); //generating k-itemset
(4) For all transactions t∈D do begin
(5) Ct=subset(Ck,t); //Ct is all the candidate itemsets that are contained in t (6) for all candidates c∈Ct do c.count++;
(7) end
(8) Lk={ c∈Ck|c.count≥minsup_count }
(9) end
(10) L=∪Lk;
After that, association rules can be generated from the set of all frequent itemsets by setting minimum confidence threshold.
3. TOPSIS Method with Combination Weights
TOPSIS (Technique for Order Preference by Similarity to Ideal Solution), initially proposed by Hwang and Yoon (1981), is one of the commonly used classical multiple criteria decision making methods for ranking and selection of a number of possible alternatives through measuring Euclidean distances. The concept of TOPSIS is that the most preferred alternative should not only have the shortest distance from the positive ideal solution, but also have the longest distance from the negative ideal solution. The basic TOPSIS methodology can be described as follows.
First of all, the decision matrix is built by aggregating the selected criteria values, and then, based on the weighted normalized decision matrix, the positive-ideal solution (PIS) and the negative-ideal solution (NIS) can be found from these limited alternatives. After that, the distances from each alternative to PIS and NIS are computed. In the end, the closeness coefficients for all the alternatives can be obtained, and these alternatives can be ranked according to the closeness coefficients. In the rest of this section, we will introduce this method specifically.



3.1. Data Preprocessing
Assume that there are n alternatives to be evaluated with regard to m selected criteria. The criteria are assumed to be benefit criteria, as TOPSIS requires that the utility of each criterion be monotonic. This assumption does not cause any loss of generality as other types of criterion can be easily transformed into a beneficial one. The criterion value of the i th alternative for the j th criterion is denoted as vij (for i=1, 2, …, n; j=1, 2, …,m). Therefore, the initial decision matrix can be obtained as
(1)
To ensure that the selected criteria are comparable, normalization for the initial decision matrix is needed so that the incomparability among criteria can be avoided. Here we use the following approach to carry out normalization,
(for benefit criteria) ; (for cost criteria) (2)
where xij is the normalized criterion value for vij, and a normalized decision matrix X can be obtained as
(3)
3.2. Criteria Weights Determining
Criteria weights reflect the decision makers’ subjective preference as well as the objective characteristics of the criteria themselves (Zeleny M, 1982). Whether the weights are reasonable can directly influent the reliability and effectiveness of the decision result. Up to now, many weights determining methods have been developed, which can be divided into subjective ones and objective ones. Subjective methods which can reflect the decision makers’ subjective preference are usually based on the decision makers’ own experiences, knowledge and perception of the problem, but objectivity is lacked. For objective methods, although they take into account the information implicated in the data and can reflect the characteristics of the criteria themselves, the weights determined by them may deviate from the actual situation sometimes and can not be interpreted easily. Thus, a method that combines the two methods together may be more reasonable.
There are many subjective weights determining methods such as AHP, Delphi method and DARE method. In this paper, group AHP is used to determine subjective weights. AHP, developed by Saaty (1980), addresses how to determine the relative importance of a set of criteria in a multiple criteria decision making problem. This method requires the experts to make pair-wise comparison for the selected criteria and a judgment matrix for each expert can be obtained. Then the relative weights are given by the right eigenvector corresponding to the largest eigenvalue for each judgment matrix. After that, the consistency for pair-wise comparisons is examined. If the consistency examination can’t pass, this judgment may be eliminated or remade. Afterwards, hierarchical clustering method is used to generate the weights for each expert in the group. And the ultimate subjective weights p= (p1, p2, …, pm) can be obtained by weighted calculation. More details can be found in Wu, Hua, & Zha (2003).
In common with subjective methods, many objective methods are also used to determine criteria weights, such as principal component analysis and entropy method. This paper uses entropy method to determine objective weights (Deng, Yeh, & Willis, 2000). This method computes the standard deviation of the data for each criterion to measure the information content such that most of the information in initial data can be kept as much as possible. The weights can be obtained by this method as follows:
(i) Compute the entropy value for each criterion: the entropy value for the j th criterion ej can be obtained through the following calculation:
. (4)
(ii) Compute the weight for each criterion: the weight value for the j th criterion uj can be computed as
. (5)
After that, the criteria weights can be obtained by using the convex combination approach, that is to say, the final weight for the j th criterion can be denoted as wj=αpj+(1-α)uj (0≤α≤1). The value of  can be set according to users’ preferences.
Then, the weights matrix can be written as
(6)


3.3. Procedure for TOPSIS Method
After data preprocessing and criteria weights determining, we can use TOPSIS method to rank the limited alternatives. TOPSIS method consists of the following steps (Hwang et al., 1981; Deng et al., 2000):
(i) According to the normalized decision matrix and weights matrix, the weighted normalized decision matrix can be obtained as Z= XW = (zij) n×m.
(ii) Compute PIS and NIS for each criterion according to matrix Z, which can be obtained as follows:
(7)
where and denote PIS and NIS for the j th criterion respectively, j* and j’ denote the sets of benefit criteria and cost criteria respectively.
(iii) Compute the Euclidean distances from each alternative to PIS and NIS.
(8)
where and denote the distances from the i th alternative to PIS and NIS respectively.
(iv) Compute the closeness coefficients for all the alternatives. Let Ci denote the closeness coefficient for the i th alternative, and Ci can be computed as
(9)
(v) All the alternatives can be ranked with regard to the closeness coefficients computed above, and the decision makers can make decisions based on the ranking result.
4. A Framework for Ranking Association Rules
Traditionally rules are selected only due to the thresholds of support and confidence, which only considers the database perspective. Actually the interestingness of an association rule is application-dependent (Srikant et al., 1997). The domain information in application areas can potentially provide useful criteria for selecting useful rules and can be adopted to select rules which are more interesting to the users. In this section, a new framework for ranking association rules which considers both the objective interestingness measures and the users’ domain information is proposed.
The proposed framework can be illustrated with the flow chart in fig. 1. Firstly, the transaction database is constructed, and association rules can be generated by using Apriori algorithm with minimum support and minimum confidence. After that, the users are required to select criteria taking into account both the objective interestingness measures and the users’ domain information, and determine the weights for the selected criteria using the method proposed in this paper. Finally, TOPSIS method can be used to rank the mined association rules, and the rules at the top of the ranking list can be selected for implementation.

Fig. 1. Framework for ranking association rules
5. Illustrative Example
To show the applicability of their proposed approaches, an example of market basket analysis is presented in Chen (2007) and Toloo et al. (2009). In this example, association rules are discovered using Apriori algorithm, in which minimum support and confidence are set to 1.0% and 10.0%, respectively, so that some infrequent but interesting rules can be kept. Forty-six association rules are mined from the market basket database.
The criteria selected for ranking consist of support, confidence, itemset value, and cross-selling profit. For association rules like XY, support is the percentage of transactions that contain X∪Y, confidence is the ratio of the percentage of transactions that contain X∪Y to the percentage of transactions that contain X. Itemset value refers to the sum of values of all items in the itemset X∪Y. Cross-selling profit can be described as the recommending that customers purchase Y, if they have bought X.
Next, we can use the proposed TOPSIS method to rank these 46 mined association rules. Firstly, the subjective weights determined by using group AHP can be denoted as p= (0.118, 0.167, 0.262, 0.453).
Using the above mentioned entropy method, the objective weights can be obtained as u= (0.196, 0.259, 0.122, 0.423).
For convenience, we set =0.5, and the weights vector can be written as w= (0.157, 0.213, 0.192, 0.438) by using the convex combination approach.
Finally, TOPSIS method is applied to rank the mined 46 association rules, and the closeness coefficients and ranking result for these rules are shown in table 1.
Table1. Ranking result and closeness coefficient for each association rule

Table2. Ranking results of Chen’s method and Toloo’s method
Ranking Rule no.
Chen's method Toloo's method
1 26 18
2 22 23
3 18 26
4 17 12
5 7 31
6 23 43
7 6 22
8 43 6
9 31 17
10 12 1
From table 1, we can see that the top 10 association rules are rules 17, 18, 12, 31, 28, 29, 27, 34, 33, and 46. The ranking result is different from the results with Chen’s method and Toloo’s method, which are shown in table 2. But it should be noted that rules 17, 18, 12 and 31 rank at the top of the ranking list no matter with our method or with the other two methods. Thus we can conclude that these rules are strong association rules which can be selected for implementation.
6. Conclusion
Association rule mining algorithms can generate amounts of rules, but only a small number of rules may be selected for implementation. There is a need for developing techniques to obtain rules that are more interesting to the users. Based on multiple criteria decision making theory, a framework for ranking the mined association rules using the proposed TOPSIS method with combination weights, which considers both the objective interestingness measures and the users’ domain information is presented, and a market basket analysis example is applied to illustrate the applicability of this method. The result shows that our method is computationally efficient and applicable. What’s more, our method takes into account both the objective weights and the subjective weights for the selected criteria, so the ranking result should be more credible.
References
Agrawal, R., Imielinski, T., & Swami, A. (1993). Mining association rules between sets of items in large databases. Proceedings of the ACM SIGMOD Conference on Management of Data, 254–259.
Agrawal, R., & Srikant, R. (1994). Fast algorithms for mining association rules. Proceedings of the international conference on very large data bases, 407-419.
Chen, M. C. (2007). Ranking discovered rules from data mining with multiple criteria by data envelopment analysis. Expert Systems with Applications, 33, 1110–1116.
Choi, D. H., Ahn, B. S., & Kim, S. H. (2005). Prioritization of association rules in data mining: multiple criteria decision approach. Expert Systems with Applications, 29, 876–878.
Deng, H., Yeh, C. H., & Willis, R. J. (2000). Inter-company comparison using modified TOPSIS with objective weights. Computers & Operations Research, 27, 963-973
Geng, L., & Hamilton, H. J. (2006). Interestingness measures for data mining: A survey. ACM Computing Survey, 38(3): 9, 1-32.
Hwang, C. L., & Yoon, K. (1981). Multiple attribute decision making methods and applications. Berlin, Heidelberg: Springer.
Saaty, T. L. (1980). The analytic hierarchy process. New York: McGraw-Hill.
Srikant, R., Vu, Q., & Agrawal, R. (1997). Mining association rules with item constraints. In Proceedings of the third international conference on knowledge discovery and data mining, KDD-97 (pp. 67–73).
Tan, P. N., Steinbach, M., & Kumar, V. (2006). Introduction to data mining. Beijing: Post & Telecom Press.
Toloo, M., Sohrabi, B., & Nalchigar, S. (2009). A new method for ranking discovered rules from data mining by DEA. Expert Systems with Applications, 36, 8503-8508.
Wu, Y. Y., Hua, Z. S., & Zha, Y. (2003). Calculation of the weights and the amalgamation of the matrixes in AHP group decision. Operations Research and Management Science, 12, 16-21.
Zeleny M. (1982). Multiple criteria decision making. New York: McGraw-Hill..

TOTAL VALUE MANAGEMENT IN ENTERPRISES

TOTAL VALUE MANAGEMENT IN ENTERPRISES OF VALUE CREATING MODEL
NAIJING WANG
Shandong Economic University, 7366 Erhuan Road
Jinan, Shandong, China
E-mail:cappussino@163.com
LU JING
Department of Civil Engineering and Architecture, University of Jinan, 106 Jiwei Road
Jinan, Shan Dong, China
E-mail:babycappussinor@163.com
In the background of knowledge-based economy and economy globalization, comparing with the traditional manufacturing industry, the modern manufacturing industry takes basic changes in operation character and manufacturing technique. To realize the target of “Enterprises of Value Creating Model”, the value management of modern manufacturing industry should rise to higher level than the traditional VA——“Total Value Management”(TVM).
1. How to State Total Value Management
Experts in Rand Corporation spent more than twenty years tracking 500 companies all over the world, and they found that there is a common feature in successful companies. In addition to the profit, they also pursued social objectives. American Management Master Peter Drucker pointed out: "Business exists in society; business is not only a livelihood, but a life." A new generation of management Master Peter Senge in "The Fifth Discipline" is also emphasized the learning organization is the true meaning of "living out the meaning of life." Since the enterprise is a kind of life, and to live out the meaning of life, we must pursue the value of life. The value mentioned here is the total value (or the overall value), that is the enterprises in their production process, not only to consider its own interests, but also to consider many interests of stakeholders. Enterprises will be able to survive and develop as long as they continue to create total value. Therefore, since the 80's of the 20th century, whether to create the total value had became new standards for the valuation of businesses.
Japan 4th Management Consulting World Conference pointed out that enterprise must do "6S" in the 21st century, that is CS (Customer Satisfy); ES (Employee Satisfy); MS (Manager Satisfy); SS (Society Satisfy); IS (International Satisfy); NS (Nature Satisfy). In fact, total value theory embodied in the "6S" idea.
In October 1997, Social Accountability International (SAI), European-American transnational corporations and other international organizations developed the Social Accountability 8000 International Standard (SA8000), which was the first international standard in code of ethics in the world. It requires companies to take up the environment and the responsibility of stakeholders as they pursue their own interests.
In our opinions, it should be defined as a kind of managing pattern, which seeks maximum enterprise value creation, basing on enterprise product value management, by means of customer value creation, with ideas of upgrading enterprise society value and sustainable development and aiming at enterprises of value creating model. It’s obvious that “total value” contains enterprise product value (EPV), customer value (CV) and enterprise society value (ESV).
2. Basis of the Total Value of Management Theory
Value Management has been studied by many disciplines, and modern manufacturing management studied it with a view of the product value, customer value and social groups. Therefore, the research for them was more unilaterally, and for integrating these three aspects of the “Total Value Management” study was rare.
The research for product value original by monographs "Value Engineering - Value Analysis Technology", which written by Lawrenoe • D • Miles published in 1961. Since then, after the United States set up the value engineering organization in 1959 , Japan, the United Kingdom, Australia, South Korea, India, Kuwait have also set up organizations. Especially in China, major breakthroughs have been made in the theory and business since value engineering adopted in the mechanical industry in 1979. In 1985, Professor Porter stated "enterprise value chain" which is a analysis tool of competitive advantage, a variety of value management models based on value chain have emerged in the modern manufacturing industry, such as Porter Hines re-definite the value chain, and Jeffery F. Rayport and John J. Sviokla proposed virtual value chain. Among these, the most common research focused on the value analysis in production or manufacturing operations for a single product, a small number of scholars have put forth to conduct a comprehensive analysis of the product value for the entire "value chain".
Since the 1980s, theoretical and practical areas focused on the customer value. The current study focused on some areas such as the definition, nature description and driver factors, Gronroos, Parasuraman, Woodruff, Zeithaml, and Professor Bai Changhong of Nankai University and other scholars have done a lot for this work. They reached Consensus in the following three areas: Customer value is perceived value of customers, which is the trade-off between perceived benefits and perceived sacrifices; Customer value has three characters as follow: Sequential attribute, Dynamic attribute and Relationship attribute; relationship marketing could create a more value than pure transaction marketing.
Social responsibility is a hot topic at home and abroad over the past decade, especially in the social responsibility of modern manufacturing research. British scholar Oury Shelton proposed "Corporate Social Responsibility" concept first in 1923, from the 1950s so far, Keith Davis, Joseph W. McGuire, Edwin M. Epstein, Stephen P. Robbins, Archie B. Carroll have been studied social responsibility. In 1997, British-American non-governmental organizations - the Council of the priority areas of the economy (CEP) issued SA8000. Subsequently, Denmark, the Netherlands, France, the European Union have legislation to regulate the enterprise social responsibility. In recent years, legislation such as environment, population and resources on the domestic, as well as the "Scientific Outlook on Development" put forward, including on the norms of enterprise social responsibility from the micro-level understanding, at the same time, is also forcing companies to achieve social value with varieties ways.
3. Enterprises of Value Creating Model Framework in Modern Manufacturing Industry
The concept of “Enterprises of Value Creating Model” was put forward in the 33rd Value Engineering Annual Convention held by Japanese Value Engineering Society in Tokyo from October 24 to 25 in 2000. The idea of TVM in modern manufacturing industry shows the inner meaning of enterprises of value creating model clearly. On the basis of above analysis of TVM, combining the basic frame of enterprises of value creating model set forth in the convention, we construct the framework of “enterprises of value creating model in modern manufacturing industry” (As Figure 1).

We can see from above frame that modern manufacturing industry is in the global and social environment. To survive and develop better, it must protect and respect the whole environment and keep the ideas of increasing ESV and sustainable development. It includes: to protect environment, to save energy, to protect basic human rights and interests and to standardize enterprise behavior with SA8000 criterion. The customer of modern manufacturing industry is a part of society public. They deal with modern manufacturing industry directly or indirectly. The successful transaction and the good relationship is the final approval of EPV. Therefore, the modern manufacturing industry must be committed to enhancing customer value and core customer value innovation with customer value creation and customer value chain analysis tool. The increase of EPV is the basis of achievement of maximum value for modern manufacturing industry. At the same time of seeking ESV and creating CV, modern manufacturing industry should analyze, evaluate and create systematically all the activities in the value chain to achieve rectification of high-cost economy, with enterprise value chain as managing tool and VA as managing method.
4. TVM in Enterprises of Value Creating
Dialectical unifies of EPV, CV and ESV in modern manufacturing industry. With reference to its essential, TVM in modern manufacturing industry is not confined to EPV, but the integration of EPV, CV and ESV. This is determined by the multi-target pursuit of value in modern manufacturing industry. EPV is the ratio of profit the enterprise gains to the expenses. It refers to the value of all the activities relative to product value creation in the enterprise value chain, like production operation, product price, delivery time and after-sale service. CV is the perceived value of customers about the whole operation of enterprise, that is, customers’ trade-off between perceived benefits and perceived sacrifice. It refers to the value of all the activities relative to CV creation in the customer value chain. ESV means in the process of creating value, modern manufacturing industry must give consideration to social accountability, lay emphasis on social effectiveness, such as vindication of man’s dignity, attention to environment protection and energy conservation. As far as their relationship, fundamentally speaking, they are consistent: only if CV is increased, customers can accept the enterprise, EPV can be realized, so the enterprise can develop thoroughly, increases ESV and consequently seeks higher CV; only if ESV is increased, society can accept the enterprise, so the enterprise can develop in more favorable circumstances, realizes and consolidates EPV.
4.1. EPV management
4.1.1. Value chain in modern manufacturing industry
Since Porter put forward “value chain”, the tool of analyzing enterprise competitive advantage, in his book Competitive Advantage, various management styles based on value chain have been coming up continually in modern manufacturing industry. The following is the value chain of modern manufacturing industry.



In fact, each part of chief and auxiliary business in the value chain of modern manufacturing industry can be subdivided into several concrete activities, that is unfolded form of enterprise value chain (as Table 1). The purpose of these classifications is to find the potential aspects in which enterprise can gain or create product value. According to this, the resources and capability that can increase EPV will be identified.

4.1.2. Method of EPV―Method of VA
4.1.2.1. Definition of EPV
Production value is generated by the features, characteristics, quality, variety and style of product. The formula is: Product + Service = product value, which is the center of the customer needs, and also the primary factor for customer purchasing products.

4.1.2.2. Analysis of EPV
As a carrier of brands and enterprises, product is the intangible value. It makes value competition among commodities, and form into the corporate culture competitive in the final. We can evaluate EPV in particular from the following aspects.
Functional (F): customer needs. Customer needs products with core function and product quality, that is the core layer, which reflects a key part of customer needs, the specific performance is product structure and composition of elements which determinate the function and quality.
Product Additional Interests (U): availability. This is additional part of the products or additional products, a variety of services associated with product supply. Product additional interests meet the customers needs better, and usually is different part. Its can attract customers, establish a good corporate image, increase the rate of repeat purchase.
Shape (A): seductive packaging. The value of production is mainly composed of two parts: one is using value, which can be percept only after using it; the other is the brand value which conveyed through the packaging design, which is products physical characteristics layer. There are different physical characteristics layers in same core layer, and able to meet the preferences of different customer needs.
We must also consider the product cost C in addition to the above factors: transfer value in physical work, live and create the value of labor cost to the dominant part of the individual workers, labor protection costs, downtime losses, etc. The formula is as follows:
(1)
In formula 1, EPV is value; F is core layer which is function, can be seen as the demand for products; U is product additional interests, which is availability, can be decomposed into after-sales service SE, product safety SA, product' environmental G; A is shape, a brand factor,; C is the cost.
If design-manufacturing cost is Cm; sales-service cost is Cs; transaction cost is Cd; the use-maintenance cost is Cu, the total cost is as follows:
TC= Cm + Cs + Cd + Cu (2)
In formula 2, design-manufacturing cost and sales-service cost component of the total cost of vendor Cj:
Cj = Cm + Cs (3)
Formula 1can be written as:
(4)
4.2. CV Management
CV is in fact customer perceived value (CPV), whose core is the trade-off between perceived benefits and perceived sacrifices. Perceived benefits is a series of benefits the customers can perceive from certain product or service; Perceived sacrifices is a series of cost the customers pay out in evaluating, obtaining and using certain product or service. CV can be expressed with this formula:
CV＝CPV＝ (5)
Customer value chain is an integration of CPV formed in transaction process and relationship process. Its structure should give full expression to the core idea and basic character of CV (Just as Figure3).

In fact, as each part of chief and auxiliary business in enterprise value chain can be subdivided into several concrete activities, the eleven steps of trade value and four parts of relationship value in customer value chain (we call them “a-class factor”) can be also subdivided into several concrete CPV, including perceived benefits and perceived sacrifices (we call them “second-class variation”). As Table 2:
Management of CV also shows the idea of VA. The structure and unfolded form of customer value chain provides a quantum basis for our research of CV management. For each second-class variation, we can set customers’ perceived weight and perceived score so as to measure CV effectively and discover core customer value. According to this, we start CV creation.


4.3. ESV Management
People has put the goal from the pursuit profit maximization to enhance the social value with the development of society, that is concerning on business-to-staff, the environment, the community.
Norsk Hydro story "four circles" described the connotations of ESV. In "Four Circle" story, the business stakeholders are the customers, employees, government, local communities, public institutions and powers. Founded in the early stage, enterprise focused only on the products, then enterprises begin to pay attention to the working environment of employees, as well as the protection of the environment to the community with the passage of time, and ultimately, enterprise developed through their own performance, such as respect for local culture, respect for human rights, as well as the product sustainable products. Thus, we can understand the social value as "modern enterprise strike a balance between social responsibility in production and management process, such as people-oriented, pay attention to environmental protection, energy-saving society, participate in charity, etc." and "four circles" theory summarized enterprise social values as follows:
ESV= Attention Employees+ Attention Environment + Attention Society (6)
Designing Indicators of ESV management system should attach importance to the value of resources and environmental protection, economic growing value, society stability and harmonious development value, and we can try to establish a ESV evaluation system, which is combined of personality indicators and common indicators, qualitative indicators and quantitative indicators, the basic indicators and amendment indicators, process indicators and outcome indicators, as shown in table 3.


(7)
（8）
In formula 7 and formula 8, WBi and Wcij are the weight for Bi and Cij.
5. " Soft Value Chain of the Reasonable Quality" based on TVM
"Reasonable quality" means a scientific, systematic, comprehensive and long-term planning for operation of the quality and future development built on grasp customer needs and exceeds customer demand dynamically, based on the economic of quality cost and technical state of practice. Product quality determines the degree of customer satisfaction. But at the same time, products quality impacts of life-cycle, and affects the products sales. In the pressure of demand and supply flexibility, it restricts enterprise to achieve the goal of profit maximization, which affects product value in turn.
In addition, the level of product quality affects the level of environmental pollution further so as to determine the social value. So, how to restrict the products quality within reasonable limits, and reached the balance of EPV, ECV and ESV, then obtains the enterprises integrated value maximization? Measure of reasonable quality provide new depth thinking for the total value management in modern manufacturing industry. "Soft Value Chain of the Reasonable Quality" can be seen as Figure 4.

"Soft Value Chain of the Reasonable Quality" reflects the relationship between enterprise profits and EPV, ECV, ESV. In short terms, EPV is a dominant factor, the upgrade of EPV, ECV, ESV makes decline in business profits, but in long terms, ECV and ESV will enhance profit greatly.
How to develop enterprises of Value Creating in modern manufacturing industry? Total value management (TVM) has provided a new management paradigm and the development space, it also develops brand-new thoughts worthy of reference for enterprise circles to discuss and research this subject and build up management concept that can make enterprises alive forever. Sustainable development for both enterprises and society, the win-win for enterprise development and social interests will be the principal means for enterprise enhancing competitiveness.

References
Porter．Competitive Advantage.[M]．Translated by Xiaoyue Chen．Poking: Hua Xia Press, 1997.
Kotler．Marketing Management[M]．Translated by Ruhe Mei．Poking: Ren Min University of China Press, 2001.
Naijing Wang．Basic Principle and Method of Value Engineering[M]．Ji’nan: Enterprise Society of Shandong Province, 1991.
Naijing Wang．Development Strategy and Mode Choice of Our Country’s Small Enterprise in Knowledge-based Economy Ages[M]．Ji’nan: Education Press of Shandong Province, 2002.
Xinjian Gu．Knowledge-type Manufacturing Industry[M]．Poking: National Defense Industry Press, 2000.
Naijing Wang, Ma Gang．Haier’s Management Innovation and Constructing of Enterprises of Value Creating Model[J]．Value Engineering, 2001, 4.
Naijing Wang, Du Tao．Customer Value and Structuring of Customer Value Chain[J]．Value Engineering, 2004, 4.
Fang Zhuo．Value Analysis and Value Management Based on Value Chain[J]．Value Engineering, 2001, 4. Therefore, the modern manufacturing industry must be based on customer value creation as a means to customer value chain analysis tool, is committed to enhancing customer value and core customer value innovation.
Guangming Zhang．New Development, New Characteristics and New Enlightenment of Management in Modern Manufacturing Industry[J]．Hua Dong Economy Management, 1995, 5.
Yuanzhen Xiao．Knowledge-based Economy in 21 Century and Sustainable Development of Modern Enterprise[J]．Hu Nan Economy, 2003, 3.
Feelders•A•J，Daniels•A•H•M．A general model for automated business diagnosis[J]．European Journal of Operational Research, 2001, 130(3).
Valarie•A•Zeithaml．Consumer perceptions of price, quality and value: a means-end model and synthesis of evidence[J]．Journal of Marketing, 1988, Vol.52.
Gronroos．Value-driven relation marketing: from products to resources and competence[J]．Journal of Marketing Management, 1997, 13.
Parasuraman．Reflections on gaining competitive advantage through customer value[J]．Journal of the Academy of Marketing Science, 1997, 25(2).
Woodruff．Customer value: the next source for competitive advantage[J]．Journal of the Academy of Marketing Science, 1997, 25(2).
Pankaj Ghemawat．Sustainable Advantage[J]．Harvard Business Review, 1986, Vol.64.

FUNCTIONAL DYNAMICS IN SYSTEM OF INNOVATION

FUNCTIONAL DYNAMICS IN SYSTEM OF INNOVATION:
A GENERAL MODEL OF SI METAPHORIZED FROM TRADITIONAL CHINESE MEDICINE
Xi Sun and Xin Tian
School of Management, Graduate University of Chinese Academy of Sciences
Beijing, 100190, China
E-mail: sunxi-b08@mails.gucas.ac.cn; tianxin05@ mails.gucas.ac.cn
Xingmai Deng
School of Management and Economic, Beijing Institute of Technology
Beijing, 100081, China
E-mail: dxingmai@gmail.com
Based on a comparison between two strands in SI and an analysis on the notion of “ideal model”, this paper builds a general model of SI through the metaphor form function-defined ideal model in Traditional Chinese Medicine (TCM). This general dynamic model metaphorized from TCM is built on the base of Yin-Yang-Five-Phases theory. It is made up of five sub-systems and the supporting networks of innovation. Both static descriptions and evolutional dynamics of SI are discussed. Policy implications are given based on the basic features of SI as a complex adaptive system.
1. Introduction
The concept of “systems of innovation” (SI) is still in the state of art because of conceptual diffuseness and strands of methodologies. This restricts its effect in policy making. This paper tries to develop a model to study SI in a systematic manner. Although path-dependence is important in SI, we argue there is rationality to develop an ideal model of SI because of the convergent global competition and common features of innovation. The ideal model introduces the systematic thought of Traditional Chinese Medicine (TCM) by metaphor based on the function-defined approach. The archetype in TCM is an ideal model of human body, so SI model here is also ideal to manifest what SIs should be as complex adaptive systems (CAS). That needs a description of dynamics among sub-systems of SI. The central topic of both models is harmonies among all function-defined sub-systems. Disharmony would be adjusted by self-organization or outside interventions. Human body is intervened by therapy, while SI policy.
The paper is organized as follows. Section 2 will present the two main strands in SI research, i.e. the actor-defined approach and function-defined approach, and analyze the rationality to develop an ideal model of SI. Section 3 will give an introduction to the systematic thought in TCM, which is mainly the Yin-Yang thought and the Five Phases theory. The metaphorical model will be developed in Section 4. Modeling is conducted by the basic logic of TCM: to identify Yin and Yang in SI firstly, then to metaphorize different functions of SI into different functional sub-systems in human body. Analyses on systematic dynamics are possible when the fundamental metaphors are established. Section 5 discusses policy implications of the metaphorical model with some examples in China. The last part concludes this paper from the perspective of methodology.
2. Divergent Systems of Innovation (SI) Approach in the Past Two Decades
SI approach emerged in the 1980s when the systemic nature of innovation is increasingly realized. Nowadays SI is still the most attractive realm in innovation research. But it is still not a formal theory, but an approach or conceptual framework.
2.1. Origins of SI and the Actor-defined Systems of Innovation
All the earliest researches on national innovation systems (NIS) could be classified into two strands. Freeman (1987) and Nelson (1993) could be seen as the springs of SI in macro-level those emphasize the arrangement of institutions and the role of R&D. While Lundvall (1992) highlighted the theoretical analysis on interactions between producers and users, what he called the micro-foundation in NIS. These two strands get no agreement on the key issues in NIS: Nelson (1993) stresses the balance between public and private technology, while Lundvall (1992) put learning in the core of NIS. This reflects the lack of a general definition of NIS (Edquist, 2007).
But there are some similarities between them, one of which is that they defined NIS by actors and relations among actors, what are partly influenced by institutions (Freeman, 1992; Lundvall, 1992). This has deep influence on the SI approach. From then on, Patel and Pavitt (1994), Metcalfe (1995) and OECD (1997) define NIS as actors-based innovation networks regulated by institutions. Because of relative success on the research of the systemic nature of innovation, SI approach developed several other specifications. The notion of sectoral systems of innovation (SSI) focuses on the innovative and production processes in a border defined by sectors (Breschi and Malerba, 1997); regional innovation systems focus on interactive learning embedded in region (Cooke et al., 1997); technological systems focus on the generation, diffusion and utilization of specific technologies in technology networks (Carlsson and Stankiewitz, 1995). All these SIs are defined and analyzed by institutions and actors, or agents.
We call all researches above as actor-defined SI approach. It considers actors and institutions as the main components of SI. It has obvious shortcomings: (1) actors in SI tend to play multiple roles, this will make the inter-relations complex (Galli and Teubal, 1997); (2) the difficulty in description of system’s dynamics makes this approach lack a system-level explanatory factor (Liu & White, 2001); (3) the border of system is set a priori to nation, region or technology, while actors are hybrid of innovation functions and unrelated ones, thus the presupposed border is illogical (Johnson, 2001). Thus, a new approach emerged, which define and analyze SI by functions.


2.2. Function-defined System of Innovation
In the retrospect of SI research, Edquist (2005) compares function-defined approach and actor-defined one. He admitted the importance to study functions in SI in a systematic manner. Actually, the function-defined approach in SI emerged in late 1990s.
Galli and Teubal(1997) is the earliest paper gave attention to functions in SI. They emphasized the role functions played in SI research, and developed a function-based componential approach. They categorized functions into hard functions and soft ones, and all components in SI are linked by different linkages. Liu & White (2001) inherited this approach, constructed a framework including 5 fundamental activities— R&D, end-use, education, implementation, linkage—and focused on the performance implications of a system’s structure and dynamics. This model focused more on system-level characteristics, but little on inter-functional relations and dynamics in SI.
Johnson (2001) sorted functions in SI to basic and supporting ones. Basic functions affect the innovation process directly, supporting ones indirectly. In the following empirical work, the list is modified to 5 functions (Jacobson et al., 2004): to create new knowledge, to guide the direction of search processes, to supply resources, to facilitate the creation of positive external economies and to facilitate the formation of markets.
Hekkert et al. (2007) focuses on the processes important for well performing SIs. The authors depict three motors of change in SI in terms of processes and functions, which are demonstrated as A, B, C in Fig.1. Bergek et al. (2008) added insights from political science, sociology of technology and organization theory to describe a systematic approach. 6 steps constitute the scheme (Fig.2). They clarified each functions in step3a. In their latest research on functional dynamic of TIS, Suurs and Hekkert (2009) identified 7 functions in technological innovation systems. Interactions between functions result in cumulative causation in the formation of TIS. The authors acknowledge that complex interactions may lead to complex development process.

Fig.1. Functional dynamics in Hekkert et al.(2007) Fig.2. Policy schemes in Bergek et al. (2008)
We can conclude that functional dynamics approach (FDA) calls for not only the classification of functions, but also the inter-functional dynamics in SI. A clear landscape of functional dynamics derives from a clear definition of innovation process and corresponding functions. Dynamic relations among functions are complex, this is one of the difficulties to build FDA. But we are sure that all functions should not be inter-reinforcing ones, there are not only positive feedbacks but also negative ones.
The two main approaches are both in dilemmas: the actor-defined approach is weak to explain systematic performance in innovation, while the function-defined one is still deficient to manifest fundamental rules and mechanisms of innovation system. Simply, the systems of innovation approach have paid most attention on the nature of innovation, but few on the systematic methodology that is important to model system.
Then, another question comes forth: is there any “ideal model”? In terms of Todtling and Trippl (2005), does one size fit all? They think there should exist no general system model in SI which are highly specific because of different development paths. But some theoretical argument at micro-level may give us some new insights in this problem. Just as Eisenhardt et al. (2001) states, dynamic capabilities of firms are “idiosyncratic in their details and path dependent in their emergence, they have significant commonalities across firms (popularly termed ‘best practice’).” Commonalities are determined by the interaction between fundamental nature of activities and competition. Different SIs have different actors and relations among them, international competition environment and fundamental nature of innovation are alike more or less. It is possible to develop a model to manifest the commonalities among systems based on a proper methodology of system. Summarily, heterogeneities of SI are largely related to the specific forms of actors and institutions, while functions of SI are similar. Thus, we may find an ideal model of SI by FDA under the help of a suitable systematic methodology.
3. Traditional Chinese Medicine: the Potential Proper Methodology of System
TCM may be the methodology we need. It is a philosophical thinking of ancient China. It is holistic to build a functional model embedded in the environment rather than to open the black box of human body reductively . Human organism was seemed “not a machina with a single deux”, and “for any recognizable continuance of identity its parts were not separable” . Chinese doctors give every patient a unique treatment based on specific diagnosis and constitutions. Such “philosophy of organism” is a world-view “derived from the biological, evolutionary and holistic facets of natural science which … have been contributing to ‘a rectification of the mechanical Newtonian universe’” .
3.1. Yin-Yang Theory in Traditional Chinese Medicine
TCM begins and ends with Yin and Yang and never goes outside Yin and Yang. Yin and Yang are a general term for two opposites to describe how things function in relation to each other and to the universe. All things have both Yin and Yang aspect. The Yin aspect is associated with such qualities as cold, rest, constringency and downwardness. The Yang aspect is associated with heat, stimulation, radiation and upwardness. They contain within themselves the possibility of opposition and change, and depend on each other for definition, mutually create and control each other. Furthermore, they transform into each other. It is only through this kind of mutual creation and restriction that dynamic equilibrium can be established. Harmony means that Yin and Yang are relatively balanced; disharmony means that the proportions are unequal.
TCM believes that normal physiological functions of human body result from unified and opposite relation between Yin and Yang, and the internal viscera, both functional and corporeal, are of the two counter-reacted aspects which are interdependent, inter-supporting and inter-consuming in harmony. Disharmony would result in disease. Both of them are always in dynamic balance in which one waxes while the other wanes. Even under normal conditions, Yin and Yang can not be in absolute balance.
The basic Yin and Yang substances of the body include Qi, Blood, Essence, etc. Qi could be seemed as matter on the verge of becoming energy, or energy at the point of materializing. It promotes all organic activities, thus is the source of transformation in the body. Blood in TCM is not same to “blood” in West. The major activity of Blood is to circulate continuously through the body, nourishing, maintaining, and moistening its various parts. Essence is the Substances that underlie all organic life. It is the source of organic change. It is supportive, is the basis of reproduction and development.
3.2. The Theory of Five Phases (Wu Xing) and its Applications in Medicine
The Five Phases theory is to classify phenomena in terms of 5 processes, represented by wood, fire, earth, metal, and water. It is a system of correspondences and patterns that subsume events and things, especially in relation to their dynamics.
Each Phase is an emblem that denotes a category of functions and qualities. Wood is associated with active functions that are growing. Fire designates functions that have reached the most active state and are about to decline or rest. Metal represents functions in declining. Water represents functions that have reached a maximal state of rest and are about to start a new cycle. Earth designates balance or neutrality, is a buffer among other Phases. In terms of Yin-Yang theory, Wood is the Yang in Yin, Fire is the Yang in Yang, Metal is the Yin in Yang, Water is the Yin in Yin, and Earth is the buffer between Yin and Yang. The Five Phases can also be used to describe annual cycle in terms of biological growth. Wood corresponds to spring and is associated with birth, Fire corresponds to summer and growth, Metal corresponds to autumn and is associated with harvest, Water corresponds to winter and is associated with storage, Earth corresponds to the change from one season to the next and the activity of transformation.
The Five Phases generate sequences and movement, as well as qualities which are important in TCM. These correlations (solid lines in Fig.3a), are known as the Mutual Production order. They represent the way in which the Five Phases interact and arise out of one another. Production implies that one activity can promote or bring forth another. Another sequence is known as the Mutual Checking or Mutual Control order. In this sequence, each phase is to control or restrict the corresponding Phase (broken lines in Fig.3a). Production and Control have inseparable correlations in the Five Phases. They oppose each other and yet also complement each other. Production leads to growth and development; while Control balance and coordination during development and change. However, once any one of the Five Phases becomes excessive or insufficient, there would appear abnormal counter-control known as insult and humiliation. By insult is meant that one of the Five Phases over-controls upon another one when the latter is weak. Humiliation means that the strong bullies the weak. It is also a morbid condition in which one phase fails to control the other in the regular order, but in reverse order. It is clear that the order of humiliation is just the opposite to that of insult.

Fig. 3a. inter-dynamics of Five Phases Fig. 3b.inter-dynamics of Five Yin Organs
In TCM, the Five Phases Theory is used to explain different kinds of medical problems by analogizing and deducing their properties and interrelations. In the sense that the Phases correlate observable phenomena of human life into images derived from macrocosm, they serve a similar function as that of elements in other medical systems. TCM recognizes five Yin Organs (wu-zang) and six Yang Organs (liu-fu), all are defined by function but remote to the anatomic reality. The Yin Organs are Liver, Heart, Spleen, Lungs and Kidneys, by sequence which corresponds to Wood, Fire, Earth, Metal and Water in the Five Phases (Fig 3b). Because the Yin Organs are generally more important in medical theory and practice, thus we only give further illuminations to them.
-Liver rules flowing and spreading. Liver or Liver Qi moves the Qi and Blood in all directions, sending them to every part of the body.
-Heart dominate Blood and Vessels. It propels and regulates the circulation of Blood. Under the impulse of Heart-Qi, Blood is transported to all parts of the body.
-Spleen rules the transformation and transportation. It is the primary organ of digestion and is the crucial link in the process by which food is transformed into Qi and Blood, thus it is viewed as the source for the production and transformation of Qi and Blood. Besides this, Spleen governs Blood. It keeps Blood flowing in its proper paths.
-Lungs dominate Qi and respiration. They are the place of exchange between the gases inside and outside the body. They regulate Water channels, govern dispersing and descending, and communicate with numerous vessels to coordinate functional activities of the whole body, assisting Heart to adjust normal circulation of Qi and blood.
-Kidneys store the Essence and are seemed as the "foundation of prenatal life". They are the foundation of the entire process of Water movement and transformation, they rule Water through their Yang aspect.
The Mutual Production order of the Five Phases in five Yin Organs describe normal generative functions, that is the different phases of Qi’s movement, i.e. birth, growth, transformation, harvest and storage. The producer here is called Mother and the produced, Child. Some patterns of disharmony can be explained by reference to the Mutual Production order, especially patterns of Deficiency. And a disharmony within the Control order might mean that an Organ controls excessively over the Organ it regulates (insult), then it would lead to a Deficiency in the regulated Organ. Or the Organ that should be regulated may become the regulator, such conditions usually happen when the former regulator is deficient (humiliation), while the regulated one excites excessively.
3.3. Meridian in Traditional Chinese Medicine
Meridian comprises an invisible network that links together all the Substances and Organs. These networks are unseen but important: the Substances Qi and Blood move along them, carrying nourishment and strength. Because Meridian makes all tissues and organs in body an organic whole, it is essential for maintenance of harmonious balance. Furthermore, the Meridian connects the interior of the body with the exterior. Thus we can say the Meridian is the interface of human body which is a dissipative structure. Meridian theory assumes that disorder within a Meridian generates derangement in the pathway and creates disharmony along the Meridian, or that such derangement is a result of a disharmony of the Meridian’s connecting Organs.
In TCM, human body is in dynamic balance of Yin and Yang in both the whole system and its function-defined sub-systems. All sub-systems, especially the Five Yin Organs interact with each others through Meridian to seek a relative balance among different functions at the system level. Health is the result of harmony between Yin and Yang and among different functional sub-systems. Thus, there exists an ideal model of human body which is in the dynamic harmony in TCM. All diseases are resulted in the deviation from the ideal model. Slight deviation could be self-adjusted by internal systemic dynamics; while serious ones need therapy as an intervention.
4. A Metaphorical Model of SI Based on Traditional Chinese Medicine
This section try to build a ideal model of SI derived from TCM. Because the archetype is used to describe the ideal state of human body as a CAS, the ideal model of SI also describes a self-organized process of innovation in nation, region or sector. The metaphorical model following will take TCM as an archetype and a reference in methodology, and the model here has not to correspond to the archetype strictly. It is necessary to declare that the main purpose of this paper is to introduce the fundamental methodology in TCM which is in a systematic manner into the SI research. So this paper only gives a conceptual framework and some tentative explanation of system dynamics in SI which are analyzed qualitatively without any numeric experiment or simulation.
4.1. Description of basic static features of sub-systems
It is important to make proper metaphors of Yin and Yang in SI. Because Yin is usually associated with functions which tend to form concrete bodies, we can link it with finance-related functions in SI which usually resulted in the formation of capital goods; another support to this metaphor is that the Essence stored in Kidneys is considered as money that can finance any Organs to function normally. Yang is typically related with stimulation, radiation and increase. In knowledge economy, the most critical engine of economic growth and commercial success is knowledge. Thus it is reasonable to link Yang with the movement of knowledge. There needs a network to link all functional modules together like Meridian in human body so that different functions could interact with each other. Here we call it the Supporting Networks of Innovation.

Fig. 4.Metaphorical model of System of Innovation
Based on the definitions of Yin and Yang in SI, the metaphorical model corresponding to the Five Phase theory has five sub-systems defined by functions linked by the supporting networks. Functions are assorted into knowledge creation, knowledge diffusion, transformation from knowledge to value, value realization and ex-post selection of innovation, and financial support. They correspond to Liver (Wood/birth), Heart (Fire/growth), Spleen (Earth/transformation), Lungs (Metal/harvest) and Kidneys (Water/storage) respectively. It has to note that the sequence is just a logical path to describe innovation, but not innovation process in reality. Supporting networks of innovation comprise social culture, social structure, infrastructure and legislation (Fig. 4). Although the model is not suitable to enterprises, some micro-dynamics will be discussed in view of their importance as micro-foundation in the holistic mechanism.
- Knowledge creation (KC). R&D has an increasing importance today among the various ways to create knowledge. It is the main part in KC dynamics. R&D also relates to knowledge assimilation, which is significant for followers to create knowledge. Knowledge base embodied on capital goods and talents is necessary for and restricts knowledge creation. Comparatively, the input of knowledge base is the Yin aspect in KC, while the emergence of new knowledge is the Yang one. The Yin aspect links KC with FS, that is to say capital support will come true when the capitalists believe the knowledge base may be enough to innovate. At the micro-level, new knowledge concludes tacit knowledge which is personal techniques derived from recursive practices (Ziman, 2000). At the macro-level, knowledge is mostly created by public research institutions and enterprises. The most important issue here is to keep the balance between public knowledge and private ones (Nelson, 1988).
- Knowledge diffusion (KD). Education and training, personal transfer among universities and industries, information exchange through social networks, licenses of patents under IPR protect and other non-market approaches, e.g. imitation, are the main approaches to diffuse knowledge at macro-level (Freeman, 1987; Morone and Taylor, 2004). Those approaches constitute a knowledge diffusion network. The nucleus of KD is to govern and utilize this network suitably, just as the Heart rules Blood and the Vessels. “To govern” may be the Yin aspect of KD, while “to utilize” the Yang one. The paths of diffusion are context-specific. The structure of interaction is determined by the relation between knowledge suppliers and users, which is affected by the supporting network, including IPR legislation and technological infrastructure. Efficiency of KD is influenced by various factors, from spontaneous reform on routines to legislation, the former of which may be initiated by competition. Knowledge itself would make no sense for innovation until it transforms into product what is valuable for the customers.
- Transformation from knowledge to value (KT). Enterprises are the main actors to transform knowledge into value. They “digest” resources got outside of the SI under the help of knowledge and capital in SI, transform them into a new kind of knowledge carrier. Knowledge embodied in processes and routines are the most important determinant in this period. There are two aspects of this kind of knowledge: to govern or to integrate all resources into the transformation process, and to create some thing valuable. Efficient routines in knowledge transformation improve productivity. KT is greatly related to entrepreneurial activities. Entrepreneurs organize resources and find a solution to satisfy customers’ needs. The solution is an integration of capital goods and knowledge. The accesses to resources, such as market information, customers, and skilled workers decide the specific model of entrepreneurial activities. Transaction structure that associated with the supporting networks determines business model. Entrepreneurship and competition are the most important dynamics in this module.
- Value realization and ex-post selection of innovation (RS). Entrepreneurs realize the value of their solutions in market. Market is a vital mechanism of ex-post selection of innovations (Ziman, 2000). When innovation realizes its value, it would be imitated by competitors, thus effective competition in SI assists KD module to adjust knowledge circulation. Access to market is determined by the bargain among innovators, competitors, users and dealers, which is affected by the supporting network. Integration of useful solution and proper access leads profit and competitive advantage.
- Financial support to other parts of the SI (FS). Investment is indispensable in every function module in SI. This sub-system are the foundation upon which the entire process of capital movement in SI. Knowledge to govern the portfolio of innovation investments is one of the most important knowledge in SI. So, this sub-system also promotes the KT sub-system to transform knowledge to value. Comparatively, to get more capital to finance innovation is another aspect of this module. Access and amount of financial support to innovation determine innovation behavior of the system as a whole.
4.2. System dynamics: sub-dynamics and inter-functional dynamics
System dynamics in the metaphorical model of SI include both micro-level dynamics and macro-level ones. Sub-dynamics enable modules to generate and reinforce itself to be competent to internal coordination and outside competition. They are the foundation of SI, while the inter-functional dynamics link the system as a whole. It should be noted that this list of functions and tentative explanations on system mechanism in SI require further revisions as and when research on SI dynamics provides new insights.
4.2.1. Sub-dynamics in sub-systems
Just as the whole system is in a harmonious matching process, every part of the system dynamics should be in an evolutionary balance. Thus it is significant to describe these micro-dynamics of the system clearly and roundly.
The most important issue to incent KC module is to keep the balance between the creation of public knowledge and private ones. Government has to keep a suitable scale of public knowledge creators for two purposes: on one hand, to continuously supply public knowledge which market is failed to incent private ones to do; on the other hand, to prevent private ones from depending on public supply of competitive knowledge.
Knowledge diffuses through market and non-market approaches. It also changes the balance between public knowledge bases and private ones. A proper legitimation on IPR protection could incent the innovators to create more knowledge and facilitate knowledge diffusio. Orientation of education also affects knowledge diffusion. Continual reforms on education system are key constituents of sub-dynamics in KD module.
To keep a sustainable transforming capability of SI, it is remarkable to cultivate an entrepreneur-friendly environment. Emerging enterprises are less inertial to keep old routines which are based on outdated knowledge, thus the main drivers for the emergence of new routines and models in transformation. The successful ones in them introduce new routines into whole economy. Thus to keep multiple springs of competing transforming routines by encouraging entrepreneurship is in the core of KT module.
Competition is a crucial part of evolutionary dynamics in RS module, but not the only driver of this sub-system. In an effective market, interactions between producers and users are co-evolutionary process. The co-evolution results in mutual enlightenments about customers’ needs, by which transformation is oriented. Customers’ cognition about their needs is sometimes the outcome of a recursive explanatory process, which adjusts the ex-post selection and is determinant to the uncertainty in innovation.
Commitment is determined by strategic intent and resources. Contradictions between financial needs and existing resources cause the introduction of external resources and the knowledge about efficiency of innovation investment. The former leads to strategic alliances and utilization of capital market. Permanent improvement on the latter causes dynamic optimization of innovation investment portfolio and capital sources portfolio.
4.2.2. Inter-functional dynamics in ideal state
Harmony emerges when competent sub-systems match with each other closely by interactions. Tentative clarifications of harmonious Mutual Control Cycle in the metaphorized model are given as following.
KC controlling KT: Application of new knowledge in transformation could prevent KT from rigidity which is the byproduct of increasing returns of knowledge bases.
KD controlling RS: diffusion of knowledge erode the appropriablity of innovators’. Thus, followers’ imitations incent innovation by less lucre and more furious competition.
KT controlling FS: competent routines in transformation utilize resources properly, and have decisive effect to prevent excessive thirst in investment and overflow of capital.
RS controlling KC: enterprises try their best to exploit the potential of existing knowledge bases. Market propels knowledge in the orientation decided by customers’ need to restrain the Yang aspect of KC from ascending excessively.
FS controlling KD: the development of FS sub-system is part of the solution to excessive imitation by those SMEs who are short of capital to innovate indigenously.

Fig 5.sequence of self-coordination by mutual control in Five Phases system
Any problem will lead to disharmony. If changes are moderate, SI can adjust by self-coordination. Otherwise, it could not be adjusted spontaneouslyand result in disharmonies. Fig.5 is the general sequence of self-coordination by mutual control.
4.2.3. Inter-functional dynamics in disharmonious systems
Harmony would be destroyed when balances among functional modules are broken or the supporting networks are not expedite. Concretely as following: ①KC not producing KD: supporting networks are insufficient to deliver knowledge to the diffusion network. ②Fire not producing Earth: 1) KD diffuse improper knowledge to KT, e.g. students could not learn the knowledge they will need in their occupation; and 2) the FS Yang could not help KT utilizes investment appropriately, which causes waste even disorder in KT. ③KT not producing RS: access for innovation to market would be severed for many reasons. In developing countries, a large broker and symptom of “foreign is better”(FIB) may be blocks for entrepreneurs. ④RS not producing FS: producers not entrepreneurs invest their profit to innovation-unrelated realms. ⑤FS not producing KC: once both capital and knowledge base are insufficient, it is difficult to create knowledge especially in those emerging areas which have a high threshold in scientific infrastructure. Above five items are the disharmonies in Mutual Producing Cycle. There also exist disharmonies in Mutual Controlling Cycle, while we do not discuss these here because of length limitation.
Obviously, disharmonies here are similar to “system failure” in evolutionary economics. It is necessary to clarify that situations described above are just basic forms of disharmonies of the system. Realities of deviated SI are usually certain combination of basic forms. Thus some more analyses are needed to develop proper innovation policy.
5. Policy Implications of Metaphorical Model: Some Issues about China’s NIS
Because the introduction of methodology in TCM which originated more than two thousand years ago to research CAS of human body, the metaphorical model may seem to be complex and remote to common SI research. Thus, it is useful to discuss its policy implications. Some issues analyzed following are related to the situation of NIS in China.
(1) Is the idea of general model inconsistent with the common sense in SI research? Traditional SI research emphasize the path-dependence, thus it seems popular that there is no ideal model in SI. This argument can be discussed from two perspectives. On one hand, harmony is realized through dynamic balances between knowledge and capital, and among sub-systems what are linked by supporting networks. So the structure of SI is changing all the time motivated by internal self-adjustment. Thus the ideal model corresponds to a set of states but not an isolated point. That means the model is so robust that it has many specific forms in reality which are highly related to environment and initial conditions. On the other hand, those failed systems which could not incent innovations properly are deviated from the ideal model. Any deviation is determined by path-dependence, thus deviation is highly specific bases on history. Remedies to the deviations should be corresponding to the heterogeneous disharmonies. So there could exist a framework of anticipatory institutional changes (AIC, Galli and Teubal, 1997) in the transition of SIs, but this framework should not be a simple graft of “best practices”; reversely a treatment derived from specific deviation.
(2) Holistic and generative view on policy-making. This is very important to SIs in transition. The ideal model surveys SI in a holistic to reflect the systemic nature of innovation. It tells the policy makers that a proper AIC should take all aspects into consideration. Innovation policies in developing countries like China are usually developed without anticipatory design on system level. For example, one defect in the evolution of China’s SI is the separate reform courses on enterprises and S&T system since 1980s. There exists even no coordination between these two reforms. So both the reform on S&T system and the reform on enterprises always get half the result with twice the effort (Lu, 2006). Furthermore, the ideal model holds a generative perspective on SI. System dynamics determine the transferring order and corresponding results of deviation among sub-systems. It implies that sequence of policy implementation is so important that it shapes new path-dependence of system in future. Once there is some disarrangement in policy sequence, the efficiency of system transition would be debased, and paths of transition in next phase would be altered. In the reform of S&T system in China in 1990s, lots of public research institutes were reformed into S&T enterprises. This changed the transaction structure between knowledge suppliers and users. While most of the enterprises which have limited absorptive capabilities and inefficient competition, have not realized the significance of knowledge and innovation. Furthermore, the old institutional junctions of both sides are destroyed with the repeal of industrial ministries in earlier time. Thus, this reform was not effective as expected.
(3) Multi-causation or one-to-one relationship? Dynamics in the CAS deny the one-to-one relations between phenomena and essence. There exist multiple causations in SI. Therefore, policy makers have to make sure the essence of disharmonies in SI. A false judgment may leads to inefficient, even opposite outcome. For example, Chinese government is glad to see the surge of R&D expenditure in enterprises. But there are several possibilities for a surge in enterprises R&D. It can be resulted in the shortage of knowledge on governing innovation investment portfolio (Deficiency in FS Yang). It also can be derived from the incompetence of enterprises to integrate resources in the process of transformation (KT is humiliated by FS.). It is also related to the lack of knowledge bases (to prevent FS not producing KC) and inadequacy in education to play its role in KD module (FS insults KD). Thus, a proper investment arrangement on R&D should be shaped based on a clear understanding of multiple causations in the whole system.
6. Conclusion
The paper has carried out an experiment to build an ideal model of systems of innovation based on a selection of basic approach in the research of SI and the help of the basic thought of TCM. Although some concrete explanations to system dynamics are still not supported by existing research, such as the situation of FS humiliated by KD, this paper analyzes SI with some unique methodologies. The main conclusions of this paper on the methodology in SI research can be summarized as follow:
(1) Compared to the actor-defined approach, FDA may be more helpful to future research in SI in a systemic manner. This is decided by the inconsistency between the systemic nature of innovation and the component-based reductionism.
(2) As a CAS, dynamic mechanism in SI should be analyzed from both micro and macro level. This is the most crucial suggestion from TCM. In FDA model, evolution of SI is seemed as the synthetic outcome of both sub-dynamics and inter-module dynamics.
(3) Innovation policies intervened to the CAS have to be developed on the basis of a thorough consideration So a holistic and generative policy framework should be formed in view of systemic heterogeneity and multi-causations among sub-systems.
Reference:
Bergek, A., Jacobsson, S., Carlsson, B., Lindmark, S., and Rickne, A. (2008), “Analyzing the functional dynamics of technological innovation systems: A scheme of analysis”, Research Policy 37:407–429.
Breschi, S., and Malerba, F. (1997), Sectoral innovation systems: technological regimes, Schumpeterian dynamics, and spatial boundaries, in Edquist 1997a: 130-156;
Carlsson, B., and Stankiewitz, R. (1995), On the nature, function and composition of technological systems, in Calsson 1995: 21-56.
Cooke, P., Gomez, M.U. and Etxebarria, G. (1997), “Regional innovation systems: Institutional and organisational dimensions”, Research Policy 26(4/5): 475-491.
Dosi, G., Freeman,C., Nelson, R. G. and Silverberg, L. S. (1988), Technical change and economic theory, London: Pinter.
Edquist, C. (1997a), System of innovation: technologies, institutions and organizations, London: Printer.
Edquist, C. (1997b), System of innovation approaches-their emergence and characteristics, in Edquist 1997a: 1-35.
Edquist, C. (2005), System of innovation: perspectives and challenges, in Fagerberg et al. 2005: 181-208.
Eisenhardt, K. M. and Martin, J. (2000), “Dynamic Capabilities: what are they”, Strategic Management Journal 21(Special issue):1105-1121.
Fagerberg, J., Mowery, D. C., Nelson, R. (2005), The Oxford handbook of innovation, Oxford: OUP.
Freeman, Christopher (1987), Technology, institution and economic performance: lessons from Japan, London: Pinter.
Galli, R. and Teubal, M.(1997), Paradigmatic Shifts in National Innovation Systems, in Edquist 1997a: 342-70.
Hekkert, M.P., Suurs, R.A.A., Negro, S.O., Kuhlmann, S. and Smits, R.E.H.M. (2007), “Functions of innovation systems: A new approach for analyzing technological change”, Technological Forecasting & Social Change 74:413–32.
Jacobsson, S., and Bergek, A. (2004), “Transforming the energy sector: the evolution of technological systems in renewable energy technology”. Industrial and Corporate Change 13: 815–849.
Johnson, A. (2001), Functions in innovation system approaches, DRUID’s Nelson-Winter Conference, Aalborg, Denmark.
Kaptchuk, T. J. (2000), The web that has no weaver: understanding Chinese medicine, New York: McGraw-Hill.
Liu, X. and White, S. (2001),”Comparing innovation systems: a framework and application to China’s transitional context”, Research Policy30(7):1091–1114.
Lu, Feng (2006), The paths to indigenous innovation, Guilin: Guangxi Normal University Press.
Lundvall, B. A. (1992), National systems of innovation: towards a theory of innovation and interactive learning, London: Pinter.
Morone1, P. and Taylor, R. (2004), “Knowledge diffusion dynamics and network properties of face-to-face interactions”, J Evol Econ 14:327-351.
Needham, Joseph, Lu, Gwei-djen and Sivin, N. (2000), Science and civilization in China, Volume VI, Part VI :Medicine, CUP.
Nelson, R.(1988), “Institutions supporting technical change in the United States”, in Dosi, G. et al. (1988):.
Nelson, R.(1993),National innovation systems: a comparative analysis, New York: OUP.
OECD (1997). National innovation systems [EB/OL]. www.oecd.org.
Sorenson, O. and Stuart, T. E. (2001), “Syndication networks and the spatial distribution of venture capital investments”, American Journal of Sociology 106 (6): 1546-1588.
Suurs, R. A. A., Marko P. Hekkert (2009), “Cumulative causation in the formation of a technological innovation system: The case of biofuels in the Netherlands”, Technological Forecasting & Social Change: 2-38.
Todtling, F., and Trippl, M. (2005), “One size fits all? Towards a differentiated regional innovation policy approach”, Research Policy 34:1203–1219.
Ziman, J. (2000), Technological innovation as an evolutionary process, New York: CUP.

STUDY ON THE METHODS OF IDENTIFICATION

STUDY ON THE METHODS OF IDENTIFICATION AND JUDGEMENT FOR OPINION LEADERS IN PUBLIC OPINION
LIU YIJUN
Institute of Policy and Management, Chinese Academy of Sciences, Beijing 100190, China
Center for Interdisciplinary Studies of Natural and Social Sciences, Chinese Academy of Science, Beijing 100190, China
E-mail: yijunliu@casipm.ac.cn
TANG XIJIN GU JIFA
Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190
Center for Interdisciplinary Studies of Natural and Social Sciences, Chinese Academy of Science, Beijing 100190, China
E-mail: jfgu@amss.ac.cn
During the process of opinion evolution, the individuals look for emotional support and depend on opinion leaders complying with “psychological balance” principle and “Emotional resonance” principle. That is the root cause of generation of opinion leaders. This paper adopts meta-synthetic approach (MSA) and social network analysis (SNA) to identify and judge the opinion leaders and master their behaviors and traces for further exploring the nature of opinion and then effectively controlling and guiding opinion.
1. Introduction
Opinion leaders refer to the individuals who can informally influence other people's attitudes or change their behaviors. The mass media often influence audience and then change their attitudes and behaviors through interpersonal relations. Opinion is basically transmitted from the mass media to opinion leaders, and in turn to people who the leaders want to influence, which is well-known as secondary communication theory (Chen, 1999). Opinion leaders can be treated as audience and also leaders to influence audience. The special position of opinion leaders during opinion evolution process builds their enormous force.
The main effects of opinion leaders include analysis of social problems by critical thinking mode, integration of different public awareness and the dispersive views and awaking people's consciousness. This evaluation not only pointed out the direction of ethos but also judged the different views to overcome some misconception of the public. Once the public summarized their views, they will be infected by faith and passion and become followers of a certain opinion (Liu, 2002). Therefore, opinion leader is always viewed as announcer, persuader and prover.
In general, opinion leaders have the following characteristics, belonging to the same group as general members with many same attributes even the leaders owned the special status because of their knowledge and capacity, keeping in more touch with various sources of information and outside environment than other audiences and acting as sources of information and leading role of other members in a certain area.
Due to the "psychological balance" and "Emotional resonance" principles (Liu and Gu, 2008; Liu, Niu and Gu, 2009) in the social behavior entropy theory (Niu, 2001), the public continually look for the emotional support and depend on opinion leaders. That is the root cause of generation of opinion leaders.
Hegselmann etc al. (2002) figured out that opinion can be formed in a group as small as a few experts or as large as in the whole society. Based on this viewpoint, Section II of this paper will use meta-synthetic approach (MSA) and expert mining (EM) to identify and judge “expert leaders” during the process of experts argumentation. In section III, social network analysis (SNA) will be involved to find out the “opinion leaders” during the opinion formation and evolution over network. Conclusion and the future works will be proposed in Section IV.
2. Identify Opinion Leaders based on MSA
Social public opinion is a complex system. Identifying the opinion leaders emerged in the individuals mass behaviors during the evolution of public opinion is further a complex systems engineering. So this paper adopts meta-synthetic approach as a directional and advanced way to guide the identification of opinion leaders.
2.1. Meta-synthetic Approach and Expert Mining
MSA, proposed by a Chinese system scientist Qian Xuesen (Tsien HsueShen), is one of the system methodologies to tackle with open complex giant system (OCGS) problems from the view of systems in the early 1990s (Qian, Yu and Dai, 1990). Here, we regarded OCGS problems such as social public opinion as ill-structured or wicked problems. This approach expects to unite organically the expert group, data, all sorts of information, the computer technology, and even scientific theory of various disciplines and human experience and knowledge for proposing hypothesis and quantitative validating. Later it is evolved into Hall of Workshop for Meta-Synthetic Engineering (HWMSE) which emphasizes to make full use of breaking advances in information technologies (Gu and Tang, 2003; Gu and Tang, 2005).
Expert mining (EM), as a new mining method, is put forward based on the meta-synthetic approach (Gu, 2006; Gu, Song and Zhu, 2008). This method emphasizes expert experience, ideas and wisdom mining. It is not built on the basis of mass data but in a smaller group of samples based on the thinking of experts to conduct in-depth experience in mining. This method is also different from those based on artificial intelligence-based expert system because it focuses more on people - machine, human-oriented to people's wisdom and the wisdom of the main groups. Mining expert system methodology, which combines science, scientific thinking and knowledge of scientific theories and makes full use of modern computer technology, is the development of the former theory and technology.
This section tries to identify and judge expert leaders by expert leader judgment module with guidance of MSA and EM.
2.2. Hall for Workshop of Expert Argumentation and Expert Leader Judgement Module
Based on MSA, expert mining method and knowledge creation model, the Hall for Workshop of Expert Argumentation is to provide a distributed computer platform. On which, participants bring out new ideas and knowledge through communication and collaboration (Tang and Liu, 2004; Liu and Tang, 2005). The Hall integrates proposals and views from experts to build solution and compute quantitatively degree of centralization and consensus.
Aiming to the discussion topic, the Hall for Workshop of Expert Argumentation expresses the registered ID (shown in rectangular box) and keywords (shown in ellipse box) as a visualized two-dimensional map, as shown in Figure 1, The experts owning high degree of concerns will be centralized. This provides a new way to share knowledge and solve unstructured problems.
Discussion space is a joint thinking space for the participants. Via the 2-dimensional space, the idea association process to stimulate participants’ thinking, idea generation, tacit knowledge surfacing and even wisdom emergence is exhibited based on the utterances and keywords from participants. The global structure and relationships between participants and their utterances are shared by all participants in the session. It helps the user acquire a general impression about each participant’s contributions toward the discussing topic, and understand the relationships of each thinking structure about the topic between participants.
The expert leader judgement module of the Hall for Workshop of Expert Argumentation constructs the consistent matrix based on the sameness and difference of keywords from all participants. The largest eigenvector will be computed to achieve sort of speaker. The sort can also be used to exhibit contribution of each participant (Tang and Liu, 2004). The matrix A can be expressed as,
and . (1)
Where, represents the set of keywords from no. i participant.
After discussion, participants will be evaluated to help analyze quantitatively discussion result and try to find out key speaker based on effects on group from each participant. Those key speakers are “opinion leader”.
2.3. Example analysis
The Xiangshan Science Conference (XSSC, www.xssc.ac.cn), which is initiated in 1993 in similar to Gordon Research Conferences and denotes as the general designation of a series of small-scale academic meetings which bring together a group of scientists working at the frontier of research of a particular area and enable them to discuss in depth all aspects of the most recent advances in the field and to stimulate new directions for research, is a top-level science forum for interdisciplinary and cutting-edge studies and can be viewed as a platform for knowledge sharing and creation in China. Next we apply our tool to analyze Xiangshan Science Conference.
Figure 1 shows the process and result map of analyzing "the brain, consciousness and intelligence" topic with experts meeting system (Liu and Tang, 2005). Detail of design and development of the system will not be explained here. Figure 1(b) is different from Figure 1 because one new expert (“Pan”) is added into the discussion. But the two maps own the same character that the expert "Wang Yunjiu" locates at the center of the discussion. That indicates that he actively involved in the "brain" research and relative meeting. This result can be verified by the record in text mode from Xiangshan Conference.












Figure1 (a) Figure1 (b)
Figure 1. Two-dimensional Distribution of Participants and Keywords
Table 1 lists the evaluation of participation based on agreement and discrepancy matrixes. It is shown that user holds highest rank based on both eigenvectors, which may be justified by his active role as one of chairpersons or plenary speech contributors among those conferences, which furthermore exposes his big influence in neuroscience field in China.
Table 1. Evaluction of 9 Participantions
Maximum eigenvector of agreement matrix: (0.3761, 1.0914, 0.3082, 0.6179, 0.2522, 0.3618, 0.3125, 0.1937, 0.1092 )
Rank of the top five participants: > > > >

Due to less staff and simple content, Prof. Guo can not be defined as “opinion leader”. Instead, “leader expert” is better. However, such a new idea builds an important basis for research of identifying “opinion leader”.
The social network analysis proposed in the following section of this article can be used to identify “opinion leaders" from a large scale of participants.
3. Detect opinion leader based on SNA
3.1. SNA
Social network analysis (SNA), as a new paradigm for sociological research (Liu, 2004; Luo, 2005), is proposed in 1930s and enhanced in 1970s. This article intends to detect the “opinion leaders” by this method. In fact, the opinion leaders are those special individuals who appear during the formation of opinion from microcosmic individual actions to macroscopical group behavior.
“Social network” refers to the social actors and the collection of the relationship between different actors. That is, a social network is a collection of a number of points (social actors) and the connection between the points (the relationship between actors). “Social network” emphasizes that each actor has a certain extent relationship with other actors. Social network analysis build models for these relationships, try to describe the structure of relations between group members and study the effect on group and individual from this structure.
Social network analysis can be used to identify quantitatively the “opinion leaders” because this approach has exactly described the relationship between the subjects of opinion in a very good way. In which, the social network position refers to a series of individual actors who have the similar characters in social activities, relationship and interaction located in the same relationship network, network factor refers to combination of relations to link the social positions and mode of the relation between the actors or positions.
Some other concepts such as point, edge, degree, betweenness, cutpoint, component, subgroup and centralization and so on are involved in SNA. During the formation and evolution of opinion, this article particularly concerns the "cutpoint".
3.2. Cutpoint
In graph theory, the only one point connecting two sub-graphs is called as cutpoint. The cutpoint is very important because its absence will divide network into independent segments named after block. Such a point is important to not only network but also the other point That is, cutpoint plays the "opinion leaders" role among the subjects of opinion.
3.3. Example analysis
A series of serious terrorist attacks occurred in the in the eastern part of United States at September 11, 2001. With this incident, World Trade Center in New York, the Pentagon where U.S. Department of Defense locates in Washington and some other important buildings had been attacked and heavy casualties were caused. By the later survey, this is an organized and purposeful terrorist activity against the interests of the people, the U.S. security and even world peace. After that, not only the United States governments but also experts around the world analyze this incident in-depth for getting more meaningful and valuable information and forecasting such terrorism. Figure 2 (http://www.orgnet.com/tnet.html) shows the social network analysis of key man of 9 • 11 terrorist events.























Figure 2 Social network analysis of participants of 9 • 11 terrorist events

This case is involved here to indicate that social network analysis is a good method and technique to identify the “key persons”. Analogously, opinion leader can be easily identified in a war of opinion through the "cut point" algorithm if the network topology of opinion subjects had been built out.
Nie et al (2005) have analyzed the relationship between scientific collaborators (385 articles and 192 authors produced from different NSFC major research) through social network analysis and found out the "expert leaders" of the network of scientific cooperation with cut-point algorithm.
4. Conclusions
Due to the “psychological balance” and “Emotional resonance” principles, the public continually look for the emotional support and depend on opinion leaders. That is the root cause of generation of opinion leaders. In this paper, MSA and SNA are involved to identify the opinion leaders and master their behaviors and traces for further exploring the nature of opinion and then effectively controlling and guiding opinion.
Opinion leaders play the key roles in the process of guiding opinion. Trend of opinion would be obviously affected through intervening and controlling opinion leaders. In future, intervening of opinion leaders during formation and evolution of opinion by soft control theory will be studied in-depth.
References
Chen, L. D. (1999) “Public Opinion”, Beijing: China radio and television press. (in Chinese)
Gu, J. F. (2006) “Expert mining for discussing the social complex problems”, MCS2006, Beijing, September 22.
Gu, J. F., Song, W. Q., Zhu, Z. X. (2008) “Expert Mining and TCM knowledge”, KSS 2008, Guangzhou， December11-13.
Gu, J. F., Tang X. J. (2005) “Meta-synthesis approach to Complex System Modeling”, European Journal of Operational Research, 166(33): 597-614.
Gu, J. F., Tang,X. J. (2003) “Some Developments in the Studies of Meta-Synthesis System Approach”, Journal of Systems Science and Systems Engineering, 12(2): 171-189.
Hegselmann, R., Krause, U. (2002) “Opinion Dynamics and Bounded Confidence Models, Analysis, and Simulation”, Journal of Artifical Societies and Social Simulation, 5(3):1-33.
http://www.orgnet.com/tnet.html
http://www.xssc.ac.cn.
Liu, J. (2004) “An Introduction to Social Network Analysis”, Beijing: Social Sciences Academic Press.
Liu, J. M. (2002) “Principles of public opinion”, Beijing: Huaxia Publishing Co., Ltd (in Chinese)
Liu, Y. J., Gu, J. F. (2008) “Systems Analysis and Modeling of Opinion Infection”, IEEE International Conference on Systems, Man and Cybernetics: 484-488.
Liu, Y. J., Niu, W. Y., Gu, J. F. (2009) “Study on Public Opinion Based on Social Physics”, Proceedings of the 20th International Conference on Multiple Criteria Decision Making : 318-324.
Liu, Y. J., Tang, X. J. (2005) “A Preliminary Analysis of Xiangshan Science Conference as Transdisciplinary Argumentation”, New Development of Management Science and Systematic Science : 35-40.
Liu, Y. J., Tang, X. J. (2005) “The Introduction of Some Mental Models and Tools for Creativity Support”, Systems Engineering –Theory & Practice, 5（2）: 56-61.
Luo, J. D. (2005) “Social Network Analysis”, Beijing: Social Sciences Academic Press.
Nie, K., Tang, X. J., Gu, J. F. (2005) “An Analysis of a Practical .Scientific Collaboration Network”, New Development of Management Science and Systematic Science: 261-267.
Niu, W.Y. (2001) “Social physics: significance of the discipline’s value and its application”, Science, forum. 54（3）: 32-35. (in Chinese)
Qian, X. S., Yu, J. Y., Dai, R. W. (1993) “A new Discipline of Science - the Study of Open Complex Giant Systems and its Methodology”, Chinese Journal of Systems Engineering & Electronics, 4(2): 2-12.
Tang, X. J., Liu, Y. J. (2004) “Computerized Support for Idea Generation During Knowledge Creating Process”, Proceedings of Second International Conference on KEST, Tsinghua University Press: 81-88.
Tang, X.J., Liu, Y.J. (2004) “Exploring Computerized Support for Group Argumentation for Idea Generation”, Proceedings of 5th Knowledge and System Sciences: 296-302.