Training applied talent in artificial intelligence multi-cloud scheduling is a system of complex factors. When studying the training of applied talent, we need to comprehensively consider various factors and consider the interaction between the factors in the system. In this way, the calculated results can achieve the purpose of treating both symptoms and root causes and continuous improvement. Traditional analytic hierarchy processes, factor analysis, principal and component analysis cannot consider the relationship between complex factors in the system. In recent years, some experts and scholars have integrated and applied a variety of mathematical models to avoid the above problems. Among them, the method of combining the DEMATEL method and the ISM method is currently developing rapidly and has been frequently used in recent years. These methods can consider the interaction between indicators in the process of practical application.
Model selection
Huang argue that ISM is one of the important methods in educational technology research. Based on hacker and anti-hacker thinking, he proposed five improved ISM methods and elaborated on them in detail, bringing enlightenment to the application and expansion of the ISM model [51]. Zhou and Zhang proposed the integration of DEMATEL/ISM to construct a system hierarchy and provided the theoretical basis and algorithm of method integration. Effective integration of the DEMATEL/ISM method can reduce the computational amount and complexity of reachable matrices, providing a new idea for the analysis and decision-making of complex systems [52]. Xie first proposed the adversarial interpretation structural model method, the AISM method. This study uses the TOPSIS-AISM to analyze 8 evaluation objects and 19 dimensions of indicators and finally uses four sets of hierarchical topological maps to indicate the entire evaluation process [53].
We found that in the previous application of the ISM method and the AISM method, only the multilevel hierarchical structural diagram of the subject index is obtained. Through the structural diagram, we can only see the hierarchical structure and correlation of the index, but we cannot see the impact of the correlation between the indices, resulting in an unclear display structure of the model. Based on the above method, this study adds the comprehensive influence value, namely, the TAISM method, on the basis of AISM. The TAISM method combines the comprehensive influence value between indicators with the extraction results of confrontation levels and displays the results of the improved method by drawing a group of directed topological hierarchical structural diagrams with comprehensive influence values. This structural diagram enables the relevance and influence degree of each indicator in the influencing factor system to be more clearly displayed, which is conducive to the analysis of the model’s results.
Based on the above influencing factor system of applied talent training of artificial intelligence multi-cloud scheduling, we first calculate the matrix through the DEMATEL method to obtain the degree of influence, degree of being influenced, degree of cause and center of each factor. Combined with the causal relationship diagram of factors, the key influencing factors of artificial intelligence multi-cloud scheduling applied talent training are identified. The comprehensive influence matrix calculated by the DEMATEL method can directly calculate the reachable matrix. According to the reachable matrix and the TAISM method, a set of adversarial directed topological hierarchy graphs with comprehensive influence values can be obtained. Therefore, the combined use of the DEMATEL-TAISM model can identify and evaluate key factors in complex systems and clarify the structural levels of factors in the system.
DEMATEL method
Build the original matrix O
Through questionnaires and interviews, experts are asked to quantify and score the mutual influence relationship among the factors in the multidimensional index system affecting the training of artificial intelligence multi-cloud scheduling applied talent, where 0 means that there is no influence, 1 means that there is a weak influence, 2 means there is a general influence, 3 means that there is a strong influence and 4 means that there is a stronger influence. The research objects involved in expert fields include 3 professors and associate professors of artificial intelligence in applied universities, 2 personnel and senior executives of artificial intelligence companies, and 1 practitioner in the artificial intelligence industry. After all the expert scoring forms are recovered, the scores are summarized, and the sum of quantitative influence relationships in the six scoring forms is calculated. Determining the relationship between factors and the relationship between the direct influence degree Si(i = 1, 2…18) and Sj(j = 1, 2…18) is represented by oij, oij is the strength of the influence of the i factor on factor j, and O(oij) 18 × 18 is the direct influence matrix.
Calculate the composite influence matrix T
We normalize the direct influence matrix O by the row sum maximum method to obtain the normative influence matrix N, and the calculation process is shown in Formula 1 as follows:
$$N={left(frac{o_{ij}}{Maxvar}right)}_{18times 18}$$
(1)
where (Maxvar=mathit{max}left(sum_{j=1}^n{o}_{ij}right)).
Using the obtained standardized influence matrix and Formula (2), the comprehensive influence matrix T of the influencing factors of artificial intelligence multi-cloud scheduling applied talent training is obtained, that is, T = (tij)18 × 18.
$$T=N+{N}^2+{N}^3+cdots +{N}^k=sum_{k=1}^{infty }{N}^kto T=N{left(I-Nright)}^{-1}$$
(2)
where I is the identity matrix and A is the inverse matrix of B. From this, the composite influence matrix T can be seen in Table 2.
Calculate the degree of influence Di and the degree of being influenced Ci
Based on the comprehensive influence matrix T, the degree of influence Di can be obtained by accumulating the values of the rows, and the degree of being influenced Ci of each factor can be obtained by accumulating the values of the columns as follows:
$${D}_i={sum}_{j=1}^n{t}_{ij},1le ile n$$
(3)
$${C}_i={sum}_{j=1}^n{t}_{ji},1le ile n$$
(4)
Calculate the centrality degree Mi and the reason degree Ri
According to Formula (5), the centrality degree Mi is obtained by adding the degree of influence Di and the degree of being influenced Ci of each influencing factor. According to Formula (6), the reason degree Ri is obtained by subtracting the degree of being influenced Ci of each influencing factor from the degree of influence Di as follows (Table 3):
$${M}_i={D}_i+{C}_i$$
(5)
$${R}_i={D}_i-{C}_i$$
(6)
TAISM method
Build an adjacency matrix A
Introducing the intercept A, the intercept value in this paper is obtained by statistics with the matrix T, and its calculation formula is as follows:
$$lambda =overline{x}+sigma$$
(7)
where A is the mean of the matrix values in the T matrix and B is the sample standard deviation (Table 4).
$$sigma =sqrt{frac{sum_{i=1}^{n^2}{left({x}_i-overline{x}right)}^2}{n^2-1}}$$
(8)
$${O}_{ij}=left{begin{array}{c}1,{t}_{ij}ge lambda =0.16501710625713\ {}0,{t}_{ij}<lambda =0.16501710625713end{array}right.left(i,j=1,2,3,…mathrm{18}right)$$
Build the reachable matrix R
For any adjacency matrix A, the reachable matrix R is calculated as follows:
where B is the multiplication matrix and I is the identity matrix, that is, a Boolean square matrix with only the diagonal value of 1. Then, B is multiplied consecutively as follows:
$${B}^{k-1}ne {B}^k={B}^{k+1}=R$$
(10)
From this, the reachable matrix R is obtained, and the result is as follows (Table 5):
Build a general skeleton matrix S
The reduction point is carried out by the reachable matrix R; the loop in the reachable matrix is regarded as a point, which is called the reduction point. After reduction of the point, the reduction points reachable matrix R′ is obtained, and then the edge reduction calculation is carried out. The essence of the edge reduction operation is to delete the repeated paths. The method is as follows:
$${S}^{prime }={R}^{prime }-{left({R}^{prime }-Iright)}^2-I$$
(11)
The skeleton matrix S′ is obtained by reducing the edge of R′, and the general skeleton matrix S is obtained by substituting the loop elements.
Build the matrix with influence values WS
The value of 1 in the general skeleton matrix S is replaced with the comprehensive influence value, that is, the corresponding value in T is replaced to obtain the influence value skeleton matrix TS. We mark the directed edge inside the loop chain with 1 to obtain the matrix with influence value WS. WS is the matrix with influence values shown in Table 6.
Extraction of adversarial hierarchy
For the reachable matrix, there is reachable set R, prior set Q, and common set T, where T = R ∩ Q. For example, for the adjacency matrix A, the reachable set of ei is denoted as R(ei), that is, all elements with a row value that corresponds to 1. The prior set of ei is denoted as Q(ei), that is, all the elements with a column value that corresponds to 1. The common set of ei is denoted as T(ei), that is, R(ei) ∩ Q(ei).
For the UP-type topology diagram, the results are prioritized for hierarchical division, and the extraction rules are as follows: T(ei) = R(ei). As long as the reachable set is the same as the common set, the relevant elements are extracted. The extracted features are placed above each time, and the extracted features are placed in order from top to bottom.
For the DOWN-type topology diagram, the reasons are prioritized for hierarchical division, and the extraction rules are as follows: T(ei) = Q(ei). The extracted features are placed below each time, and the extracted features are placed in order from bottom to top.
Extracted according to the above method, the results are displayed in Table 7:
Draw an adversarial topology hierarchy structure model diagram
According to the relationship between elements and the results of the confrontation level extraction, a directed topological hierarchy graph can be drawn. There is a reachable relationship among factors in the system, which is represented by a directed line segment and a two-way arrow to form a loop; they are reachable relationships with each other. At the same time, the lower layer indicates that the influencing factors are rooted, and the higher layer indicates that the influencing factors are direct. The UP-type and DOWN-type topological hierarchical structural model diagrams are depicted in Figs. 3 and 4.
UP-type topological hierarchical structural model diagram
DOWN-type topological hierarchical structural model diagram
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Disclaimer:
This article is autogenerated using RSS feeds and has not been created or edited by OA JF.
Click here for Source link (https://www.springeropen.com/)
