What is network analysis
Network analysis method (ANP) is a decision making method adapted to non-independent hierarchical structure proposed by Professor T.L.Saaty of University of Pittsburgh in 1996. It is a new practical decision-making method based on Analytic Hierarchy Process (AHP).
As a decision-making process, AHP provides a basic method to express the measurement of decision-making factors. This method adopts the form of relative scale and makes full use of human experience and judgment. Under the hierarchical structure, it compares the relative importance of relevant elements in the same level in pairs according to the relative scale — proportion scale, and depends on the judgment of decision-makers, and synthesizes the measure of the scheme to the decision-making goal from top to bottom. Although this hierarchical structure brings convenience to system problems, it also limits its application in complex decision problems. In many practical problems, the internal elements of each level are often dependent on the low-level elements of C and dominate the high-level elements, i.e. there is feedback. The structure of the system is more similar to the network structure. Network analysis is a systematic decision-making method developed by AHP to meet this need.
ANP firstly divides the system elements into two parts: the first part is called the control factor layer, which includes problem objectives and decision criteria. All decision criteria are considered independent of each other and governed only by the target element. There may be no decision criteria in the control factor, but at least one goal. The weight of each criterion in the control layer can be obtained by AHP method. The second part is the network layer, it is all controlled by the control layer of group C is its internal network structure influence each other, it is composed of all controlled by the control layer of elements, element, interdependence, mutual control between elements and levels between internal is not independent, pass in the class hierarchy of each criterion is not a simple internal independent elements, It’s an interdependent, feedback network. The control layer and network layer constitute a typical ANP hierarchy, as shown in the figure below.
Characteristics of network analysis method
AHP analyzes a series of factors affecting the target, compares their relative importance, and finally selects the scheme with the highest score as the optimal scheme. Harker and Vargas once evaluated AHP as follows: “AHP is a complex evaluation system. When we make multi-objective, multi-criterion and multi-judge decisions, AHP can be used to solve various quantitative and non-quantitative, rational and irrational decision-making problems in the face of numerous alternative schemes. AHP is easy to use, and its rigorous theoretical foundation determines that it can solve all kinds of practical problems. The AHP model connects decision levels and derives cross-level relationships. The top layer of the model is the overall goal of the enterprise, and then it is decomposed into various specific criteria and sub-criteria layer by layer until the manager can quantify the relative weight of each sub-criteria.
Analytic hierarchy process (AHP) can solve all kinds of complex system problems for decision-makers, but it also has some shortcomings. For example, AHP fails to take into account the interaction between different decision levels or the same level. AHP model only emphasizes the one-way hierarchical relationship between decision levels, that is, the influence of the next level on the previous level. However, in practical work, when we decompose the total target layer layer by layer, we often encounter the situation of cross action of various factors. For example, different research stages of a project have different weights for each judge; Similarly, the scores of the judges on the evaluation indicators will also change at different stages of project research. At this point, the AHP model appears to be a little powerless.
The characteristic of network analysis method is that, on the basis of analytic hierarchy process (AHP), it takes into account the mutual influence of each factor or adjacent level, and makes comprehensive analysis of each interacting and influencing factor by using “super matrix” to get its mixed weight. However, ANP model does not require a strict hierarchical relationship like AHP model. There are interactions between each decision level or the same level, and the interaction between levels is represented by double arrows. Interactions in the same layer are represented by double-circular arrows. The factors pointed at by the arrow affect the decision factors at the end of the arrow. Based on this characteristic, ANP is more and more favored by decision makers and becomes an effective tool for enterprises to make decisions on many complex problems. The determination of relative importance index of each factor in ANP is basically the same as that in AHP. The relative importance index (scale) of each factor is obtained through a questionnaire survey of decision makers, but sometimes there are some inconsistent phenomena (for example, the ratio of I to H, scale is 3; J to K ratio, scale 5; The ratio of I to K is 6).
Case study of network analysis method
- Case: Risk analysis model of hydropower project based on ANP
- 1. Identification of risk factors of hydropower projects
Because many hydropower project breakdown project and project construction period is longer than the general, all kinds of construction risk will also vary, the overall risk of hydropower project risk identification will have certain difficulty, and is likely to miss the important risk factors, so it is necessary to the overall project before in identifying risks properly subdivisional work division, Then carry on the risk identification to each sub-project. At the same time, due to the diversification of risk factors, it is necessary to decompose risks according to certain risk principles. Therefore, this paper adopts the method of combining project decomposition structure (WBS) and risk decomposition structure (RBS) to identify risks. In addition, using this method for risk identification will also be beneficial to the establishment and solution of risk factor ANP structure model.
- 2. Hierarchical structure model of engineering projects
When establishing the network analysis model structure of the whole project risk factors, the working structure model of the project should be established first. Since each sub-project has its corresponding engineering control objectives: cost, schedule, quality and safety, and each sub-project must have different important degrees of influence on the overall project objectives. Therefore, when establishing the hierarchical structure of engineering projects, the importance of each sub-project should be judged according to the engineering objectives as a judgment criterion. The importance model of each sub-project established on the basis of WBS is an AHP structure, as shown in the following figure.
- 3. Network structure model of risk factors
According to the sources of risk, risk factors can be divided into five categories: natural risk, technological risk, economic risk, organizational risk and social policy risk. According to these five categories of risk, specific risk factors are divided.
Traditional risk analysis that risk probability and loss of two attributes, but this definition clearly can’t comprehensively reflects the nature of the risk of, zhang construction will therefore predictability, controllability, transferability is attribute to the risk, the risk as a multidimensional characteristics of these five attributes object is described. Is multidimensional properties of risk is described from different angles can more fully reflect the characteristics of risk factors, but the transferability and predictability can reflect on the controllability, so only will be estimated controllability can comprehensively reflect the characteristics of risk, according to the needs of research, analysis the risk estimates generally is to estimate the negative impact, On the basis of the traditional two-dimensional attribute, the “uncontrollable” is introduced to evaluate the risk of hydropower projects.
In the process of risk identification, only risk factors are identified, and the interaction between risk factors must be studied in order to establish ANP model. The influence relationship of risk factors can be obtained through expert investigation or group discussion, as shown in the table below.
According to the influence relation table, ANP structural model is established based on the occurrence probability, loss and ircontrollability of risk factors, as shown in Figure 3.
It is necessary to establish the corresponding risk factor ANP structural model for the risk factors of each sub-engineering project, so as to obtain the overall structural model of the engineering project. The established ANP structural model of the overall risk factors is a multi-criteria and multi-level model.
- 4. Analysis of risk analysis model of hydropower project based on ANP
(1) Determination of self-importance of sub-project items.
The importance of each sub-project is calculated. Since the structural model established based on WBS is in the form of AHP, it is easy to solve the importance of the model in the traditional way.
(2) Weight vector and ranking of risk factors of sub-projects.
The weight vector of interrelated risk factors under each subengineering project is determined as ANP structural model of risk factors of subengineering project in FIG. 3.
The key step of risk factor ranking of the whole project is also the core of risk analysis using ANP. According to the ANP structural model in Figure 3 and the influence relationship of risk factors in Table 1, the weight calculation of risk factors under sub-engineering projects is carried out as follows:
1) Calculate the weight of risk attributes. Compare the importance of probabilities, losses, and uncontrollability that describe the magnitude of the risk. These three attributes are regarded as criteria for evaluating risk factors, so the traditional AHP method can be used to determine their weight.
2) Calculate the weight of each risk factor under single criteria. Since this model is a multi-criteria problem, the interrelated risk factors should be compared and judged respectively under the three criteria. Now, the probability criterion is used to study the risk factors, and the process can be divided into the following steps:
A. Build the hypermatrix. The relative importance of other risk factors was compared with probability as the main criterion and one of the risk factors as the secondary criterion. That is, the importance of other risk factors was compared with the degree of influence on the probability of occurrence of this risk factor. Since not every other risk factor has an impact on it, not all elements should be compared under this criterion. Other risk factors that influence this risk factor can be found in the influence relationship table. Then, each risk category group is taken as a unit to calculate its feature vectors, that is, corresponding local weight vectors. After comparative judgment and calculation with each element as sub-criterion, the super matrix is established according to Formula (1).
Among them,
B. Establish the weight matrix. With probability as the main criterion and risk category R** I as the secondary criterion, a judgment matrix is constructed for comparison and judgment of all categories, that is, the degree of influence of each risk category on the occurrence probability of R** I risk category is compared. The comparison of judgment includes the comparison of R** I’s influence on itself with that of other categories. Because the influence degree of each risk factor is compared and judged in each risk category, the column vectors in the super matrix composed of multiple matrices are not normalized, that is, the sum of column vectors is not 1, so it is impossible to compare the influence degree of elements in different categories on a factor which is a sub-criterion. In addition, the power method cannot be used to solve the relative weight vector of limit for the unweighted super matrix, so the importance of each risk category should be compared and judged. After comparison and judgment with each category as the secondary criterion, five judgment matrices are obtained, and feature vectors are calculated. Finally, the weight matrix as shown in Equation (2) can be obtained.
C. Establish and solve the weighted hypermatrix. By weighting the super matrix according to Formula (3), the weighted super matrix can be obtained. The column vector element size in the weighted super matrix is the influence of each risk factor on the factors in this column. If a risk factor has no influence on this factor, the corresponding value is zero. In this case, the power method or other methods can be used to solve the relative ranking vector of the weighted super matrix, and the final relative ranking vector is the weight of each risk factor under the probability criterion.
3) Calculate the weight of multi-criteria risk factors. The weight vector of each risk factor was solved according to the criterion of loss and ircontrollability in turn, and then the weight obtained in step (1) was synthesized for the weight of each single criterion, so as to obtain the risk ranking of risk factors in sub-engineering projects.
(3) Overall project risk factors ranking
By solving the weight vector of the risk factors of each sub-project, we can synthesize the weight and calculate the total ranking of the risk factors of the whole project.
The weight of the risk factors of each sub-project corresponds to all the risk factors of the overall project, and the weight of the risk factors that do not affect the sub-project is set to zero. The weight of each sub-project in the overall project can be obtained from the calculation of the importance degree of the above engineering project. Therefore, the order of each risk factor in the upper engineering project can be obtained through the weighted sum of the weight of risk factors under the sub-project of each level. Finally, the total ranking of risk factors of the whole project can be obtained.
From the total ranking results, it is easy to find the biggest and most critical risk factors faced by engineering projects. As the mutual influence of risk factors is considered, the final results will reflect the actual situation more objectively and truly.
According to the above research, the flow chart of risk analysis of hydropower projects based on ANP is obtained, as shown in the figure below.
reference
- Tang Xiaoli, Feng Junwen, Wang Xuerong. Project risk management based on network analysis method. Statistics and Decision, no.16,2005
- Huang Benxiao, Peng Yumei. Selecting R&D projects using network analysis (J). Science of Science and Management of Science and Technology, no.2,2003
- Zhong Denghua CAI Shaokuan Li Yuqin. Risk analysis and application of hydropower project based on Network Analysis method (ANP). Journal of Hydroelectric Engineering. Vol. 27 no. 1, February 2008