Evaluation of information service in smart scenic area based on “SEM-ANP” synthetic evaluation mode

2.1.1 Mathematical structure of structural equation modeling

Widely used in economics, psychology and other social sciences, structural equation modeling (SEM) is a form of causal modeling proposed in 1970 s by Karlg Joreskog, a Swedish statistician [19]. During its practical use, SEM often appears as a system built upon a linear equation, which is derived from the causal relation between observable variables. The equation system can be divided into two parts: 1) measurement model and 2) structural model. ( 1 ) Y = Λ Y η + ɛ ( 2 ) X = Λ X ξ + δ ( 3 ) η = B η + Γ ξ + ς

where, formula 1 and 2 are the measurement model equation; and formula 3 is the structural model equation. Definitions of variables are as follows:

Y = Exogenous observation variable, X = Endogenous observation variable;

η= Exogenous latent variable, ξ= Endogenous latent variable;

Λy= Relationship between exogenous observation variables and exogenous latent variables;

Λx= Relationship between endogenous observation variables and endogenous latent variables;

B = Path coefficient between endogenous latent variables;

Γ= Path coefficient between exogenous latent variables and endogenous latent variables;

ɛ= Measurement error of Y, δ= Measurement error of X, ζ= Residual term of the structural equation.

With exploratory factor analysis on the questionnaire, it is concluded that four internal and external latent variables can affect respondents’ evaluation of “smart tourism” in the region. They are 1) policy implementation and infrastructure, 2) intelligent transportation, 3) enterprise and terminal, and 4) supervision and database. Thus, a structural equation is established for the evaluation index system. It has combined 28 observation variables, including network coverage and laws and regulations. The path coefficient among latent variables is also calculated.

2.1.2 Empirical analysis of structural equation model

This journal takes Jiangxi Province, China, as the case study that evaluating the capability of constructing smart tourism system and exploring the evaluation index system for Jiangxi smart touring system. In the questionnaire survey, it takes the practitioners of scenic spots in Jiangxi as the target survey object, received 529 valid questionnaires. Then through SPSS-Amos software, it successfully draws the structural equation model of Jiangxi smart tourism construction capability (Fig. 1), model goodness of fit table (Table 1) and related path coefficient tables (Table 2) were obtained respectively.

OPEN IN VIEWER

Fig. 1

Structural equation model of objective index system for evaluating the development level of Jiangxi “smart tourism”.

The exploratory factor analysis is used to preprocess the questionnaire data. Four secondary indicators are selected: 1) level of policy implementation and infrastructure, intelligent transportation, scenic area and terminal condition, and supervision database. They serve as latent variables for the evaluation system by the investigated groups, which measure the main influencing factors of information service at Jiangxi scenic areas. Next, 28 third-level indexes such as WL1, WJ1, etc. are derived as obvious variables (observed variables) of the four latent variables to jointly build a structural equation model. In this way, investigated groups’ evaluation model for information service of Jiangxi scenic areas has been established.

Table 1 shows that the chi-square freedom ratio (CMIN / DF), RMR, IFI, PGFI, PNFI and CFI of this structural equation model of objective evaluation index system satisfy ideal standards. It indicates that the hypothesis has a good overall fitting effect on the model of factors affecting consumers’ purchase intention. Values of each goodness-of-fit index meet the requirements. The structural equation has a good fitting effect on the data.

According to the table, policy implementation and infrastructure are latent variable that indirectly affect the evaluation index. Their influence is exerted through affecting three latent variables of the evaluation index, including 1) intelligent transportation (with relevant path coefficient of 0.77), 2) enterprise terminal (with relevant path coefficient of 0.93), and 3) supervision and data base (with relevant path coefficient of 0.92). Moreover, policy implementation and infrastructure has the biggest influence, which is as high as 0.8545, on the evaluation index. Latent variables rank after it include supervision and database construction (with correlation path coefficient of 0.53), enterprise terminal construction level (with correlation path coefficient of 0.32) and intelligent transportation construction level (c with correlation path coefficient of 0.09). The order of these latent variables are based on their influence on the evaluation index.

2.2.1 Mathematical structure of analysis network process

In the way of meeting discussion and experts filling in forms, the problems to be decided are systematically analyzed, control levels are established, decision objectives are defined, and relative objective weights of each criterion are obtained based on ANP method.

1. Constructing judgment matrix

Let the element _aij in the judgment matrix A = (a_ij)m*n represent the ratio of relative importance degree of element i to element j. The numbers 1–9 and their reciprocals are used as weight scale, among which 1-3-5-7-9 (odd numbers) successively represent the equally important –slightly important –relatively important –very important –extremely important, and 2-4-6-8 (even numbers) represent the intermediate state of the relative position. Elements of control hierarchy Us (S = 1) to 4are taken as the criteria, and elements in U_sj(j = 1,2,3,4, ...  ...  ,n) are sub-criteria, thus the judgment matrix is obtained through direct comparison and indirect comparison.

2. Consistency test of judgment matrix

Calculate M and normalize the result:

Mi = Π j = 1 n aij 1 / n , i = 1,2,3,...,n; w_i = Mi/Mj, i = 1,2,3, ...  ...  ,n.

Calculate λ_max and the internal consistency index CI: λ max = 1 n ∑ i = 1 n ( AW ) i W i CI = λ mac - n n - 1

Let the maximum eigenvalue of judgment matrix A = (aij)m*n beλmax and n be the order of matrix, and we get: λ max = 1 n ∑ i = 1 n ( AW ) i W i CI = λ mac - n n - 1

Calculate random consistency ratio CR:

Let RI be random consistency index, then random consistency ratio CR is: CR = CI RI

If CR value is less than 0.1, the judgment matrix A = (a_ij)m*n meets the consistency requirements. Only by meeting the consistency requirements can the following steps be carried out. Otherwise, the adjustment test shall repeat, until CR value is less than 0.1.

3. Establishment of super matrix and weights of each element group

According to the judgment matrix, the ranking vector can be calculated by the eigenvalue method.

W_ij is denoted as: W ij = [ W j 1 i 1 W j 2 i 1 . . . W jn i i 1 W j 1 i 2 W j 2 i 2 . . . W jn i i 2 . . . . . . . . . . . . W j 1 in i W j 1 in i . . . W j 1 in i ]

4. Calculation of weighted super matrix

Based on W_ij, every column of the unweighted super matrix is normalized and further weighted into a weighted super matrix. After that, all the normalized eigenvectors are combined to obtain the weighted super matrix W, with the column sum of 1.

W is denoted as: W ij = [ W 11 . . . . . . W 1 N . . . . . . . . . W jn i i 2 . . . . . . . . . . . . W N 1 . . . . . . W NN ]

5. Calculation of limit super matrix

The direct relationship between elements i and j is reflected by the weighted super matrix W_ij, and its complicated indirect relationship is reflected by the iteration of super matrix. Elements’ weight is established by constructing the limit super matrix among elements. Each element’s weight index vector depends on the column vector of the limit super matrix.

2.2.2 Empirical analysis of analysis network process

Based on the demand of objective evaluating the capability of constructing smart information tourism system, considering the current situations of Chinese smart tourism construction, this study invites 9 veteran practitioners to carry out in-depth interviews.

Before the ANP analysis, this study firstly adopts Kappa coefficient to verify the consistency of different interviewees evaluating the capability of constructing the smart tourism in the 4 dimensions of ANP. Generally speaking, when the Kappa coefficient between the various dimensions involved in the analysis is 0.4–0.6, it indicates that the consistency between the dimensions is general; when the Kappa coefficient is 0.6–0.8, it indicates that there is a high consistency between the dimensions; the Kappa coefficient is greater than 0.8, it indicates that the consistency between the dimensions is very good. In addition, the dimensions participating in the Kappa symmetry measurement should also pass the progressive significance test.

In this study, the 9 groups of evaluation data after Kappa symmetrical measurement, the Kappa value is 0.612, and the progressive significance is.000. The 9 invited staff who are deeply involved in this field have good evaluation consistency and passed statistics, the following research and analysis can be carried out.

Through the software of Super decision, it draw out a subjective evaluation index system of Jiangxi smart tourism construction capability based on ANP. Figure 2 is the development path model of the subjective evaluation system of Jiangxi scenic spots information service construction under the smart tourism vision.

According to Fig. 2, from the perspective of “smart tourism”, the development path of Jiangxi scenic area information service follows a four-dimensional structure: There are four dimensions of criteria layers: 1) Policy & infrastructure construction, 2) Intelligent transportation construction, 3) Enterprise and terminal construction, and 4) Supervision and database construction. In the network layer that corresponds to criteria layer: Policy & infrastructure construction implementation includes eight elements: WL1, WJ1, WJ2, WJ3, QR1, QR2, QR3 and FG1; Intelligent transportation construction includes three elements: ZJ1, ZJ2 and ZJ3; Enterprise and terminal construction includes eight elements: DP1, DP2, DP3, WP1, WP2, WP3, XY3 and XY4;Supervision and database construction include nine elements: AX1, AX2, AX3, AT1, AT2, AT3, LJ1, LJ2 and LJ3.

OPEN IN VIEWER

Fig. 2

Development path model of subjective evaluation index system of Jiangxi “smart tourism” construction level.

Development path model of information service at Jiangxi scenic areas are studied by software calculation and information summary. A relevant weighted super matrix is established. This is to obtain the judgment matrix of each influential dimension and weight table of dimension elements from the perspective of “smart tourism”, as shown in Tables 3 and 4. Through the observation on the inconsistency values of each element set, the CR of all dimensions and dimension elements in each variable set are all less than 0.1, and the judgment matrix and dimension element weight table pass the consistency test.

By looking into the element weight table of the development path model of information service construction of Jiangxi scenic areas under the vision of smart tourism, it can be seen that weight of element FG1 (construction level of relevant laws and regulations) is 0.2557, which is significantly higher than that of other elements. At the same time, weight of other elements generally varies between 0.05–0.15, which is close to each other. This result also coincides with the previous conclusion that the observers were easily affected by many factors when appraising the information service construction level of Jiangxi scenic areas under the vision of smart tourism.

2.3.1 Mathematical structure of “SEM-ANP”

Combining the objective evaluation (SEM) of smart tourism evaluation indicators obtained in section 2.1 above and the subjective evaluation (ANP) in section 2.2 above, in order to minimize the errors of each evaluation system. the next step that it combining the above 2 evaluation systems by constructing an objective function that minimizes the dispersion.

Through the interviewing data and literatures analysis, the latent variables of the smart tourism construction are the following 4 types, these are policy implementation and infrastructure construction, smart transportation, digital platform construction and database administration and its construction. The structural equation model of the evaluation index system of tourism construction level is obtained, and the Γ and B values of each latent variable to the evaluation index system are obtained. Combined with the Amos model, the final path coefficient of each latent variable to the evaluation index system is further obtained, assuming that each path coefficient is A. Through normalization, a normalized SEM objective evaluation index system is finally obtained.

Denoted as B_i, specifically as: B i = A i ∑ i = 1 4 A i

On this basis, suppose that there are a total of n evaluation indicators in the comprehensive evaluation problem of the evaluation index system, one evaluation methods is an objective evaluation method based on the SEM structural equation model; the other evaluation method is a subjective evaluation method based on the network analytic hierarchy process. Suppose the weight variables of the various evaluation index systems of these two methods are W1 and W2, then W1 = Bi; W2 = Wij (the ANP weighted super matrix calculated above), and also suppose the combined evaluation between SEM and ANP method, the vector is W0 = (w1⁰, w2⁰, . . . , wn⁰)  ^T , among this, and the optimization model of the combined evaluation index system based on the SEM-ANP method to minimize the dispersion is: { W 0 = min ∥ W 0 - W K ∥ 2 = min ∑ k = 1 2 ∑ k = 1 n ( W i 0 - W i k ) { s . t ∑ i = 1 2 W i 0 = 1

2.3.2 Empirical analysis of “SEM-ANP”

Based on the mathematical logic of the SEM-ANP combined evaluation method proposed in 3.5 above, ignoring the influence of residuals, the path coefficients of the structural equation model of the objective evaluation index system of the Jiangxi smart tourism construction capability in Table 2 are normalized, and the path coefficient table is shown in Table 5.

According to Table 6, the normalization effects of the latent variables of the Jiangxi smart tourism construction level on the evaluation indicators are: policy implementation and infrastructure construction level (0.476), smart transportation construction level (0.050), enterprise and terminal platform construction level (0.178), information supervision and database construction (0.296).

Based on the optimization model of the combined evaluation index system with the minimization of dispersion constructed in section 2.3.1 above, the normalized path coefficients of the objective evaluation index system and the weight coefficients of the development path model of the subjective evaluation index system are sequentially imported.

After calculation, the weight table of the combined evaluation index system of Jiangxi smart tourism construction level is shown in Table 6.

Through the analyzing the weights of the combined evaluation index system, it shows that compared with the subjective evaluation index system obtained by the ANP analysis method, the weights of the various dimensions of the combined evaluation system have been modified to a large extent, which in turn affects the set of elements contained in it. The weight to the overall. Among the secondary indicators, the weights of policy implementation and smart information infrastructure construction levels have dropped significantly, the weights of the other two secondary indicators, which are the level of enterprise and information terminal construction, and the other level of supervision and database construction have been effectively strengthened. Compared with the traditional subjective evaluation standard, it effectively corrects the systematic errors, which caused by the expert’s external environmental influence or internal subjective influence.