4.1.3. Data deblurring process
In order to reflect the overall influence degree of each index on the governance system and to clarify the specific values of the expert scoring results, the integrated fuzzy values need to be defuzzified. The defuzzification process takes the distribution pattern, shape and height of the fuzzy numbers into consideration. According to Table 2, the results of the above expert survey scoring table are transformed into triangular fuzzy numbers, and the comprehensive fuzzy value of the degree of influence is obtained based on equations (1)-(3). Where, indicates the number of experts who participated in the questionnaire; indicates the serial number of experts who provided data.
Since the CFCS (Converting Fuzzy data into Crisp Scores) method can effectively distinguish two fuzzy numbers with the same exact value (Lina, Guotao & Jue, 2021). Therefore, in this study, CFCS method is used for defuzzification, and the combined fuzzy values are regarded as the weight values of indexes for the governance system, as follows.
Step 1 . Fuzzy number normalization.
Step 2. The upper limit value is normalized to the lower limit value.
Step 3. Calculate the weight value.
Among them, .
4.1.4. Direct relation matrix
According to the initial fuzzy weights of indexes for supply chain data governance optimization, the DEMATEL model is introduced to optimize the initial index weights, so as to weaken the subjective determination of data results caused by expert scoring.
In view of the wide scope of supply chain data governance under the industrial Internet, the complexity of the correlation between relevant stakeholders, the interference of the overlapping influence of indexes when experts score, and the difficulty of ensuring the independence of indexes, the study optimizes the data sources based on the original DEMATEL method, and points the experts’ scoring to a single evaluation target, which circumvents the tedious operational problems caused by experts due to the importance of multiple indexes for two-by-two comparison. It also reduces the chance of misjudgment due to the lack of independence among the elements. The direct relation matrix is set as to characterize the influence of the indexon, and the data source is the ratio of the initial fuzzy weights of each index, i.e.. Where,[1,24]. Therefore, the final direct relation matrix is established (Table A1).
4.1.5. Integrated relationship matrix
The direct relation matrix is normalized according to Equation (10) to obtain the normalization matrix .
Where,denotes the number of indexes. In order to represent the degree of direct and indirect influence among the indexes of supply chain data governance, Matlab software is further used to calculate the integrated relationship matrix of based on Equation(11) (Table A2). Where is the unit matrix.
4.1.6. Influence index ranking analysis
After determining the integrated relationship matrix , the influence degree of each factor, the influenced degree , the centrality degree (+), and the reason degree (-) are calculated according to equations(12)-(13), where the influence degree indicates the comprehensive influence degree of each index in the comprehensive influence matrix on all other indexes; the influenced degree indicates the comprehensive influence degree of each column of indexes inon other indexes; the centrality degree indicates the importance degree of indexes in the whole evaluation system; the reason degree is used to portray the contribution degree of an index to the formation of the evaluation system, that is, the degree of interrelationship between the index and other indexes.
where is the element of the row and column of the integrated relationship matrix.