INTRODUCTION
Supply chain has become a key element in the global economy. Currently, the supply chain is changing rapidly, resulting in a business environment with uncertainty (Xiao et al., 2012; Wu and Olson, 2008). The global competition is intensifying, making supply chain has become more complex (Tummala and Schoenherr, 2011). The supply chain is the global network used to deliver products and services, since the raw materials sourcing process through the information flow process and physical distributions (Cooper and Ellram, 1993). Supply chain consists of management systems, operations, assembly, purchasing, production scheduling, inventory management, processing, storage and transportation. Thus, supply chain management is the integration of activity over the relation of the supply chain to achieve sustainable competitive advantage (Monczka et al., 2008; Rossetti and Dooley, 2010). The increased competition resulted in a manufacturing organization to meet the needs of customers (Vinodh et al., 2011). Supply chain is a network of suppliers (Sahin and Robinson, 2002; Wu and Olson, 2008). Therefore, the supplier selection is the beginning of supply chain process. Thus, the appropriate supplier selection is one of the most important issues for any organization (Amindoust et al., 2012). When changes in the competitive environment, supplier selection is not only based on price and quality but various risk factors should be take into consideration as well. The risk is an inherently subjective construct that deals with the possibility of loss (Yates and Stone, 1992). The risk in supply chain related to the supplier selection. Risk suppliers are important parts of the supply chain risk (Xiao et al., 2012). As a result, the companies need to consider the supplier risks because the supplier risk in the supply chain is the probability of disruption and impact on performance of any firm (Zsidisin and Ritchie, 2008). In order to generate the sustainable maximize profit, the company needs to achieve the goal of low cost, high flexible quality and greater customer satisfaction (Setak et al., 2012). The supplier selection process is a key role which helps the business to be success in a global arena.
The purpose of this study, was to develop a supplier selection model based on supply chain risk factors. An electrical and electronic industry in Thailand is selected as a case study to investigate supply chain risk factors and to develop a model for helping decision makers in supplier selection decision. The first part of the study is to identify and explore supply chain risk factors in a supplier selection and the second part is to use the fuzzy analytic network process model to calculate the local weights of the factors and sub-factors, inner dependence weights, global weights and then calculate supply chain risk value of each supplier. The results of the application are discussed and main findings and contributions are drawn and future developments are suggested.
FACTOR ANALYSIS
Factor analysis is a statistical technique commonly used in many areas of science whenever it is intended to extract the relevant information about a data set, usually presented in the form of a table whose rows are the cases and the columns are the variables that characterize these cases (De Sa et al., 2014). The basic idea of factor analysis is based on correlation to group the original variables, to make variable correlation in the same group high and in the different groups less (Niu et al., 2011). The variables in the same group have a high correlation. In contrast, the variables with low correlation are arranged apart into different groups (Jun et al., 2011).
The statistical method of studying the internal interdependent relation of the related matrix of the index variable Xi(i = 1, 2,..., p) into a few factor Fj(i = 1,..., m, m<, p) and the factor Fj(i = 1, 2,..., m) reappear (Gao, 2005), can be expressed as follows:
Among which, X = (X1, X2,..., Xp)T is the original index F = (F1, F2,..., Fm)T, is the factor of the index X, A = (aij)pxm(m<p) is the factor load matrix and, ε is th special factor.
The basic steps to draw major factors with the factor analysis method are show as follow (Wang, 2004):
| • | To study out the related correlation coefficient matrix of the original data |
| • |
To study out the feature value λi(i = 1, 2,..., p) on the basis of the matrix and define its feature vector’s matrix A meets F = A'X, F, is the major factor matrix |
| • |
To fix the number of the major factors by calculating the information contribution and accumulating contribution rates of the feature root λi. The cobtribution rate: |
| |
reflects the percentage of how much the i-th major factor’s degree of variation occupies in the degree of the total variation. The higher the rate is, the more important the factor. The accumulating contribution rate: |
| |
is the important basic to selecting the number m of the major factors which should generally be a higher percentage |
| • |
The factor can be rotated to make the loads of m major factors on the variable Xi(i = 1, 2,..., p) have noticeable discrepany in order to get the clear meaning of the major factor |
Fuzzy analytic network process (FANP): Fuzzy ANP technique uses both interdependence of criteria and inner dependence of criteria with pairwise comparison matrix. Chang (1996) extent analysis method is used to evaluate fuzzy pairwise comparisons. It is very useful in situations where there is a high degree of interdependence between various attributes of the alternatives. The membership functions of triangular fuzzy number used to represent pairwise comparison of decision variables from “Very bad” to “Excellent” and the middle preference values between them. The membership functions, Mi = (mi1, mi2, mi3), where, i =1,2,..., n and mi1, mi2, mi3 are the lower, middle and upper values of the fuzzy number Mi, respectively. According to the concept of extent analysis, each object is taken and extent analysis for each goal gi is performed, respectively. Therefore, the m extent analysis values for each object are obtained as the following signs:
where, are triangular fuzzy numbers.
The steps of Chang (1996) extent analysis can be given as the following:
The value of fuzzy synthetic extent with respect to the i-th object is defined as:
To obtain perform the fuzzy addition operation of m extent analysis values for a particular matrix such that:
And to obtain:
perform the fuzzy addition operation of values such that:
Then compute the inverse of the vector in Eq. 5 as follows:
The degree of possibility of M2 = (l2, m2, u2)≥M1 = (l1, m1, u1) is defined as:
and can be equivalently expressed as follows:
The degree possibility for a convex fuzzy number to be greater than k convex fuzzy numbers Mi (i = 1, 2,..., k) can be defined by:
Assume that:
Then calculate the weights vector is given by:
Via normalization, the normalized weight vectors are:
where, W is a nonfuzzy number, and Ai = (i = 1, 2,..., n) are n elements.
MATERIALS AND METHODS
This study proposes the supplier selection modeling based on supply chain risk factors which consists of two main stages. These stages can be divided into 10 steps for using to evaluate suppliers. The first stage is to identify supply chain risk factors for establishing group of factors and sub-factors from practical point of views. The second stage is to employ fuzzy ANP method for calculating the weights of the factors and sub-factors and calculate supply chain risk value of each supplier for the alternative. The model steps of this study are demonstrated in Fig. 1:
| Step 1: |
Identify supply chain risk factors from related literature reviews which included a variety of supply risk sources such as price, financial, material, performance, capability and so on |
| Step 2: |
Integrate expert opinions by interviewing key persons about risk and relevant factors in supplier selection process |
| Step 3: |
Conduct a questionnaire survey for exploring supply chain risk factors in practices |
| Step 4: |
Group supply risk factors and sub-factors through factor analysis |
| Step 5: |
Determine the fuzzy scale for importance weight of factors and sub-factors from a decision maker who takes responsibility in supplier selection process in a company. Choose the appropriate linguistic variables for the relative weights of the factors which are given in Fig. 2 and Table 1, then make a pairwise comparison with respect to each factor from expert opinion by using the linguistic scale |
| Step 6: |
Calculate the local weights of the factors and sub-factors |
| Step 7: |
Calculate the inner dependence weights of the factors, the inner dependence matrix of each factor with respect to the other factors. This inner dependence matrix is multiplied with the local weights of the factors in order to compute the interdependent weights of the factors |
|
| Fig. 1: |
Step of proposed methodology |
|
| Fig. 2: |
Linguistic scale for relative importance |
| Table 1: |
Linguistic expression for fuzzy scale of relative weights of factors |
|
| Table 2: |
Linguistic values and mean of fuzzy numbers |
|
| Step 8: |
Calculate the global weights of sub-factors by multiplying local weight of the sub-factors with the interdependent weights of the factors |
| Step 9: |
Determine the membership functions of these linguistic variables by expert opinion as shown in Fig. 3 and the average value related with this variables are shown in Table 2. While using this evaluation scale, the linguistic variables can take different values depending on the structure of the sub-factor. For example, in the evaluation risk of a sub-factor which affects quality product risk is the “Low (L)” linguistic variable takes 0.25 fuzzy average fuzzy average values in the evaluation of a sub-factor which affects quality product risk |
| Step10: |
Calculate supply chain risk value of sub-factors each supplier by multiplying global weights of the sub-factors with scale value of sub-factors each supplier, then prioritize risk value in order to select the best supplier |
|
| Fig. 3: |
Membership functions of linguistic values for sub-factors rating |
RESULTS
An application of this model is implemented in the electrical and electronic industry in Thailand. There are two stages of this process. The first part (step 1-4) is to explore supply chain risk factors of supplier selection in practices by using interviewing and a questionnaire survey. The second part (step 5-10) is to employ fuzzy ANP model to calculate supply chain risk value of each supplier in an illustrated example. The case study of a large Printed Circuit Board Assembly (PCBA) company in Thailand which has three main suppliers is applied. The proposed model will be explained step by step together with the results as follows:
| Step 1: |
Identify supply chain risk factors from literature review to find as mentioned in Table 3 |
| Step 2: |
Interview key expert opinions from five electrical and electronic companies in Thailand in different departments working in the company’s purchasing team such as purchasing, quality control as well as Research and Development (R and D). Then, develop a questionnaire by integrating the relevant factors from interviewing and literature survey which found significant in supplier selection process |
| Step 3: |
The questionnaire was distributed to 225 samples who take responsibility in supplier selection process in electrical and electronic manufacturing companies in Thailand. Respondents were asked to rate question under a five-point Likert scale (i.e., 1 = Very low importance, 2 = Low importance, 3 = Medium importance, 4 = High importance, 5 = Very high importance). Results from the survey indicated that the most of respondents (43.11%) who answered questionnaires are in a manager position. Most of them (64%) are working in large companies and 48% of those companies have their own brand products (Table 4) |
| Step 4: |
Group supply risk factors and sub-factors through factor analysis which detailed are shown in Table 5 using SPSS. Factor analysis using principal component method revealed six factors that collectively described managements’ perspective on their supplier selection These six factors have been extracted from 30 sub-factors by using the cut-off initial eigen value of 1.00 which account for 16.042, 12.712, 10.280, 10.278, 9.880 and 7.906% of the variance explained after rotating maximum value as displayed in Table 6. Factor 1 (or R1) is entitled “External risk factor” comprising eight variables including policies, politics, natural disasters, infrastructure, environment, legal, labor disputes and economic. Factor 2 (or R2) is entitled “Quality control risk factor” comprising five variables including industrial standards, technology, safety, collaborative and R and D. Factor 3 (or R3) is entitled “Delivery risk factor” comprising four variables including on-time, quantity, quality delivery and return. Factor 4 (or R4) is entitled “Material control risk factor” comprising six variables including performance, material, purchase order, inventory, information and organization. Factor 5 (or R5) is entitled “Production risk factor” comprising four variables including productivity, quality product, capability and production process. And, Factor 6 (or R6) is entitled “Cost and financial risk factor” comprising three variables including price, financial and forecast as shown in Fig. 4. Extracting factors are shown in Table 7 |
| Table 3: |
Literature review on supply chain risk factors |
|
| Table 4: |
Summary of the demographic characteristics |
|
| Table 5: |
Total variance explained |
|
| Table 6: |
Rotated component matrix |
|
| Table 7: |
Extracting factors |
|
| Step 5: |
This decision is converted to the hierarchical structure to transform the factors, sub-factors and alternative as the schematic structure shown in Fig. 5. The ultimate goal is to choose the best supplier which will be placed in the first level. Main factors (external risk, quality control risk, delivery risk, material control risk, production risk, cost and financial risk) and their sub-factors are located in the second and the third level, respectively. Then a decision maker is asked to determine the fuzzy scale for importance weight of factors and sub-factors. Choose the appropriate linguistic variables for the relative weights of the factors which is given in Fig. 3 and Table 3. Then the decision maker continues to make a pairwise comparison with respect to each factor using the linguistic scale. For example “What do you think of scale importance between quality control risk factors when it is compared with delivery risk factors?” If the answer is “Quality control risk is Weakly More Important (WMI) than delivery risk”, so the linguistic scale is (1, 3/2, 2) as details appeared in Table 3 |
|
| Fig. 4: |
Model for supplier selection |
|
| Fig. 5: |
Supply chain risk weight of factors result by suppliers |
| Step 6: |
Calculate the fuzzy evaluation matrix and the local weights as shown in Table 8-14 |
| Step 7: |
Calculate the inner dependence weights of the factors are calculated and the dependencies among the factors are considered of main factors are shown in Table 15 |
| Step 8: |
Calculate the interdependent weights of factors by multiplying inner dependence matrix in Table 16 with the local weights of the factors provided in Table 8: |
| |
|
| Table 8: |
Local weights and pair-wise comparison matrix of “Main factors” |
|
| Table 9: |
Local weights and pair-wise comparison matrix of “External risk” sub-factors |
|
| Table 10: |
Local weights and pair-wise comparison matrix of “Quality control risk” sub-factors |
|
| Table 11: |
Local weights and pair-wise comparison matrix of “Delivery risk” sub-factors |
|
| Table 12: |
Local weights and pair-wise comparison matrix of “Material control” risk sub-factors |
|
| Table 13: |
Local weights and pair-wise comparison matrix of “Production risk” sub-factors |
|
| Table 14: |
Local weights and pair-wise comparison matrix of “Price and financial risk” sub-factors |
|
| Table 15: |
Computed global weights of sub-factors |
|
| |
Then, calculate the global weights of sub-factors by multiplying local weight of the sub-factors with the interdependent weight of factors. Computed values are shown in Table 16 |
| Step 9: |
The membership functions of these linguistic variables as shown in Fig. 4 for eachsupplier are determined by the same decision maker and the average value related with this variables are shown in Table 2. Therefore, the scale value each supplier are shown in Table 17 |
| Table 16: |
Inner dependence weights of the factors |
|
| Table 17: |
Supplier selection based on supply risk factor |
|
| Step 10: |
Calculate supply chain risk value of sub-factors each supplier by multiplying the globalweights of the sub-factors with scale value of sub-factors each supplier. The risk value results of each supplier are shown in Table 17. The table displays that a total risk value of supplier A = 0.359, Supplier B = 0.246 and Supplier C = 0.366, respectively. Graphicall presentation of results of each supplier are demonstrated in Fig. 5 |
From the graphical representation, it can be seen that supplier C has the highest risk on external criteria, quality control and delivery, whereas supplier B has the highest risk on material control. The production risk of every supplier is not different. Supplier A has the highest risk on cost and financial issues, followed by supplier C and supplier B, respectively. As such, in overall it can be concluded that supplier B is the best supplier for this company which has the lowest priority from the other alternatives.
DISCUSSION
A decision making of the best supplier selection can increase the efficiency of supply chain operations. Factor analysis and fuzzy ANP are used in the integrated way to supplement the supplier selection process. This study explores supply chain risk factors for supplier selection applied to an actual case study. The results reveal that there are six factors which electrical and electronic companies in Thailand are taken into consideration i.e. risks from external, quality control, delivery, material control, production, cost and financial which include thirty sub-factors. quality control has the highest ranking, followed by production and material control, respectively. This proved that this study has validity in practical scenario as quality is always considered as the first priority to evaluation supplier (Dickson, 1966; Ho et al., 2007; Deshmukh and Sunnapwar, 2013). If qualities of raw material or product do not meet the standards or industrial regulations, it will lead to ineffective supply chain and decrease the value of products. Consequently, products cannot be exported or launched to the market in time. Furthermore, production risk was also ranked as the second importance for the electrical and electronic case company. It is pretty much true as production process also will have directly affected to the efficiency of any company. When production process is not flexible or inefficient, it will decrease the performance of the organization. For the risk of material control like material shortages, inventory management low efficiency will affect to inventory holding cost, error on forecasting resulting in incapability to meet customer’ due date. When considering of the important weights of sub-factors risk, it found that production process, price and productivity are the top three highest ranked. These three sub-factors are also linked to the importance of main factors since they have directed effect to the quality of product, ability to compete on price and flexibility to be able to change quickly to support customers. According to this study, supplier B has found to be the best supplier because overall score was the lowest among all suppliers.
CONCLUSION
The study employed the fuzzy ANP approach to determine the fuzzy scale for importance weights of factors can application of the proposed method will offer relevant companies for more precise and accuracy analysis and help them to be more flexible in making a decision to evaluate the suppliers. The model can include both subjective and objective criteria which gain more realistic in making a decision. Moreover, it is easy to modify the concepts with other firms who need to find suitable tool in selecting the right alternative. However, there are some imitations in this research. The study applies fuzzy ANP model with a small number of expert, therefore, future study should include a group of expert opinion for determining fuzzy number in making pairwise comparison. in addition, further research can be extended by combining the FANP model to Technique for Order Performance by Similarity to Ideal Solution (TOPSIS) method to rank supplier.