Data asset quality evaluation method based on hybrid multi-criterion decision and storage medium
Through a hybrid multi-criteria decision-making method, combined with DEMATEL, BWM and Fuzzy-TOPSIS, a data quality evaluation index system was built, which solved the problem of data quality evaluation complexity in the context of data assetization, and achieved comprehensive and accurate evaluation of data asset quality and enterprise data Improvement of asset quality.
Patent Information
- Application Number
- CN202411076593.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the context of data assetization, it is difficult for the existing technology to effectively comprehensively consider the weights of traditional and new data quality dimensions and quantify the correlation, resulting in complex data quality evaluation and lack of simple and effective methods.
A method based on hybrid multi-criteria decision-making is adopted, combined with DEMATEL, BWM and Fuzzy-TOPSIS methods, a data quality evaluation index system is constructed, key indicators are identified, their weights are determined, and expert evaluation language is used to describe triangular fuzzy numbers to sort data asset quality.
It has achieved a comprehensive and accurate evaluation of the quality of data assets, simplified the evaluation process, improved the accuracy and reliability of evaluation, helped enterprises improve the quality level of data assets, and promoted the development of data assetization.
Smart Images

Figure CN119090127A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and more specifically to a data asset quality evaluation method and storage medium based on hybrid multi-criteria decision-making. Background Art
[0002] With the rapid development of information technology and the rise of digital transformation, data is becoming an important resource and strategic asset for enterprises. Data assetization has also become an important path for enterprises to achieve digital transformation and intelligent development. Data assetization aims to transform data into valuable assets and apply them to all aspects of business decision-making, business processes and business innovation. It is worth noting that the dynamic influx of a large amount of raw data collected from physical manufacturing sites or generated by various related information systems also poses challenges to the data governance capabilities of enterprises, especially the data quality issues in the big data environment. As a key factor affecting the value of data, data asset quality has also attracted much attention. Both theoretical and practical literature show that whether data can be analyzed depends largely on its quality level. When data is analyzed and mined with quality defects, the results may not only be worthless, but may even affect reasonable business decisions. In order to cope with the challenges brought by data assetization, data quality evaluation has become an important part of the process.
[0003] Data quality evaluation is a complex process involving multiple levels and indicators. Its goal is to determine whether the data is suitable for the specific needs of a particular enterprise and to provide a reliable data foundation to support business decision-making and innovation.
[0004] Existing studies have evaluated data quality from different perspectives using different evaluation methods, making important contributions to understanding data quality. With the development of data assetization and the continuous improvement of the data factor market, how to evaluate and improve the quality of data assets has become a new and important topic in the new context. Specifically, first, how to comprehensively consider traditional data quality dimensions and new data quality dimensions? Traditional data quality dimensions include accuracy, completeness, consistency, credibility, and timeliness. With the development of data assetization, the evaluation object is changing from data sets to data assets. In the context of data assetization, new data quality dimensions include cost-effectiveness and profitability. How to combine these dimensions and determine their weights has become a new discussion issue. Secondly. There are interactions and influences between different dimensions. How to quantify the correlation and conduct a comprehensive analysis. For example, the integrity and consistency of data may affect the credibility and profitability of data. How to take the correlation into account. Thirdly, given the complexity and diversity of data, it is necessary to develop a simple and effective method to describe the dimensions and evaluation process of quality evaluation. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide a data asset quality evaluation method and storage medium based on hybrid multi-criteria decision-making, so as to improve the quality level of enterprise data assets and promote the development of enterprise data assets.
[0006] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.
[0007] The data asset quality evaluation method based on hybrid multi-criteria decision-making includes the following steps:
[0008] A. Combine the asset attributes of data to build a new data quality evaluation index system in the context of data assetization to evaluate the quality of enterprise data assets;
[0009] B. During the model building process, the DEMATEL method was used to identify key indicators;
[0010] C. Introduce the BWM method to determine the weights of key indicators;
[0011] D. Use triangular fuzzy numbers to describe the expert evaluation language, and apply the Fuzzy-TOPSIS method to rank the data quality status of different enterprises.
[0012] To further optimize the technical solution, in step A, the quality evaluation indicators include data integrity, consistency, uniqueness, timeliness, relevance, accuracy, ease of operation, comprehensibility and accessibility.
[0013] To further optimize the technical solution, in step A, the data quality evaluation system includes three aspects: quality cost indicators, profitability indicators and risk indicators.
[0014] To further optimize the technical solution, in step B, the method of identifying key indicators using the DEMATEL method includes the following steps:
[0015] B1. Decision makers evaluate the direct relationship between any two indicators and determine the included indicators C = {C1, C2, ..., C n}, establish expert group P k (k=1, 2, ..., l) Evaluate the mutual influence relationship and influence degree of two indicators;
[0016] B2. Aggregate the expert evaluation matrix to obtain the initial direct relationship matrix;
[0017] Each expert P k The evaluation of the impact relationship between indicators can obtain the corresponding direct impact matrix, in which all main diagonal elements are equal to 0. Indicates expert, P k Evaluation index, Ci For indicator C j The initial direct impact matrix Z = [z ij ] n×n By combining the opinions of each expert, a matrix reflecting the influence relationship and influence degree between the two indicators is obtained, which can be obtained by aggregation through the following formula:
[0018]
[0019] B3. Establish a standardized impact matrix;
[0020] According to the initial direct impact matrix, it is standardized to obtain the standardized impact matrix:
[0021] X=λZ,
[0022] in,
[0023]
[0024] B4. Establish a comprehensive impact matrix;
[0025] The standardized impact matrix is processed to the limit to obtain the comprehensive impact matrix, which represents the final direct and indirect impact relationships between indicators. The comprehensive impact matrix T can be calculated by the following formula:
[0026]
[0027] Where I is the identity matrix;
[0028] B5. Calculate the centrality and causality of each indicator;
[0029] Centrality is the sum of influence and influence, indicating the role of the indicator in the system; causality is the difference between influence and influence, also known as net influence. If causality is greater than 0, it represents a cause indicator, and if it is less than 0, it represents a result indicator. j , the calculation formula of its centrality is as follows:
[0030]
[0031] The calculation formula of the cause degree is as follows:
[0032]
[0033] To further optimize the technical solution, in step C, the method for calculating the weight of key indicators using the BWM method includes the following steps:
[0034] C1. Each expert determines the most important and least important indicators in the evaluation system;
[0035] Expert P k Determine the most important indicator among n indicators for a specific scenario problem and the least important indicator Then, determine For the importance of all indicators, the higher the importance, the larger the value. Similarly, the importance of other indicators is determined. The importance of can be obtained by comparison:
[0036]
[0037] In the formula, Indicates P k The most important evaluation indicator For other indicators of importance, and Indicates P k Evaluate other indicators For the least important indicator of importance, and
[0038] C2. Calculate the optimal weight of the indicator;
[0039] Assume that based on the BWM method, P k The weight obtained by evaluation is In order to obtain the optimal weight of the indicator, and The value of j=1,2,...,n should be minimized, and the optimal weight of each indicator can be obtained by building a model;
[0040] minξ
[0041] st
[0042]
[0043] When there are multiple experts, based on the expert weight information, the most effective indicator weight obtained by aggregating each expert is as follows:
[0044]
[0045] In the formula, YesP k The weight of the jth indicator given; λ k Expert P k The weight of .
[0046] To further optimize the technical solution, in step D, the method for ranking the data quality status of different enterprises includes the following steps:
[0047] D1. Construct expert evaluation matrix;
[0048] Constructing triangular fuzzy numbers where x L ≦x M ≦x U , x L , x M , x U They are The lower bound element, main element and upper bound element of L =x M =x U When It degenerates into ordinary real numbers; suppose there are m asset sets to be evaluated, denoted by A = {A1, A2, ..., A m}; The indicator weights determined by the BWM method are α = {α1, α2, ..., α n}; i=1,2,...,m;j=1,2,...,n, represents the evaluation value of the jth indicator of the i-th data asset set; P k The matrix for evaluating n indicators of m data assets is as follows:
[0049]
[0050] When there are multiple experts, the evaluation of l experts is k=1, 2, ..., l for aggregation, and the comprehensive evaluation matrix is as follows:
[0051]
[0052] D2. Construct a weighted evaluation matrix;
[0053] The indicator weights determined by the BWM method are α = {α1, α2, ..., α n}, the evaluation matrix after weighting the indicators is as follows:
[0054]
[0055] D3. According to formula (2-14) and (1-15), determine PIS and NIS;
[0056]
[0057] in, j=1, 2, ..., n is the fuzzy maximal set corresponding to the weighted index value of index j; j=1, 2, ..., n is the fuzzy minimum set corresponding to the weighted index value of index j;
[0058] D4. Calculate A according to formula (1-16) and (1-17) i Euclidean distance to PIS and NIS Then calculate the closeness D according to formula (1-18) i , get the final quality ranking, D i The larger the value, the higher the corresponding data asset quality.
[0059]
[0060] in,
[0061]
[0062] in,
[0063]
[0064] A storage medium stores a computer program or instruction, which implements the content of any of the above methods when executed by a processor.
[0065] Due to the adoption of the above technical scheme, the technical progress achieved by the present invention is as follows.
[0066] The data asset quality evaluation method and storage medium based on hybrid multi-criteria decision-making provided by the present invention realize a comprehensive and accurate evaluation of data asset quality through the proposed framework by taking into account the interdependence and uncertainty among various standards. The proposed evaluation index system and method provide a new methodological reference for data quality evaluation in the context of data assetization, which is beneficial for enterprises to improve the quality level of data assets and promote the development of data assets in enterprises. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0068] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0069] Data asset quality evaluation method based on hybrid multi-criteria decision making, combined with Figure 1 As shown, it includes the following steps:
[0070] A. Combine the asset attributes of data to build a new data quality evaluation index system in the context of data assetization to evaluate the quality of enterprise data assets;
[0071] Quality evaluation refers to the completeness, consistency, uniqueness, timeliness, relevance, accuracy, ease of operation, comprehensibility and accessibility of data.
[0072] The data quality evaluation system includes three aspects: quality cost indicators, profitability indicators and risk indicators. Quality cost indicators consider data asset quality from an economic perspective, profitability indicators are used to measure the ability of data quality to increase business operations, and risk indicators consider potential losses due to poor data quality. High-quality data can provide accurate, timely and relevant information to help companies make more informed decisions and improve business benefits. On the other hand, risk indicators consider potential losses due to poor data quality. Low-quality data may mislead corporate decisions and increase business risks.
[0073] B. During the model building process, the DEMATEL method was used to identify key indicators;
[0074] The DEMATEL method can be used to determine the relatively important indicators within the data asset quality evaluation framework. This method is a matrix-based analysis method that reveals key driving factors and dependencies by analyzing the cause-and-effect relationships between elements within the system. The quality evaluation dimensions and indicator system of data assets are shown in the following table:
[0075] Evaluation indicators Indicator Description <![CDATA[Integrity C1]]> The completeness of the data in the dataset <![CDATA[Consistency C2]]> The degree of consistency in data recording formats <![CDATA[Uniqueness C3]]> The degree of non-redundancy of data records <![CDATA[Accuracy C4]]> The accuracy of the data <![CDATA[Relevance C5]]> The degree to which the data meets the business accuracy requirements <![CDATA[Timeliness C6]]> The degree to which the data meets timeliness requirements <![CDATA[Ease of operation C7]]> How easy is it to manipulate the data? <![CDATA[Comprehensibility C8]]> The comprehensibility of the data <![CDATA[Availability C9]]> The degree of accessibility and availability of data <![CDATA[Cost nature C 10 > The cost of implementing data quality governance <![CDATA[Profitability C 11 > Benefits of data-related services to enterprises <![CDATA[Risky C 12 > Potential losses caused by low-quality data
[0076] The application of the DEMATEL method can provide an in-depth and detailed analysis of the data asset quality evaluation framework. By revealing the causal relationship and importance between indicators, this method can help researchers and decision makers better understand the complexity of data asset quality evaluation, thereby developing more targeted and effective improvement strategies, which include the following steps:
[0077] B1. Decision makers evaluate the direct relationship between any two indicators and determine the included indicators C = {C1, C2, ..., C n}, establish expert group P k (k=1, 2, ..., l) The mutual influence relationship and influence degree of two indicators are evaluated, and a 5-point method is used, namely "no influence (0)", "low influence (1)", "medium influence (2)", "high influence (3)" and "very high influence (4)" to evaluate the direct influence relationship between every two indicators.
[0078] B2. Aggregate the experts’ evaluation matrix to obtain the initial direct relationship matrix.
[0079] Each expert P kThe evaluation of the impact relationship between indicators can obtain the corresponding direct impact matrix, in which all main diagonal elements are equal to 0. Indicates expert, P k Evaluation index, C i For indicator C j The initial direct impact matrix Z = [z ij ] n×n By combining the opinions of each expert, a matrix reflecting the influence relationship and influence degree between the two indicators is obtained, which can be obtained by aggregation through the following formula:
[0080]
[0081] B3. Establish a standardized impact matrix.
[0082] According to the initial direct impact matrix, it is standardized to obtain the standardized impact matrix:
[0083] X=λZ, (1-2)
[0084] in,
[0085]
[0086] B4. Establish a comprehensive impact matrix;
[0087] The standardized impact matrix is processed to the limit to obtain the comprehensive impact matrix, which represents the final direct and indirect impact relationships between indicators. The comprehensive impact matrix T can be calculated by the following formula:
[0088]
[0089] Where I is the identity matrix;
[0090] B5. Calculate the centrality and causality of each indicator;
[0091] The sum of the rows of the comprehensive influence matrix represents the influence of each indicator, that is, the comprehensive influence of the indicator on other indicators; the sum of the arrays of the comprehensive influence matrix represents the influence of each indicator, that is, the comprehensive influence of other indicators on this indicator. This step mainly calculates the centrality and causality of each indicator. The centrality is the sum of the influence and the influence, indicating the role of the indicator in the system; the causality is the difference between the influence and the influence, also known as the net influence. If the causality is greater than 0, it represents a cause indicator, and if it is less than 0, it represents a result indicator. For indicator C j , the calculation formulas of centrality and causality are as follows:
[0092]
[0093] The calculation formula of the cause degree is as follows:
[0094]
[0095] C. Introduce the BWM method to determine the weights of key indicators;
[0096] By building a linear programming model, the BWM method can calculate the weight of each indicator based on the expert's evaluation results. The model takes into account the relative importance of indicators and the balance between weights, making the weight calculation more accurate and reliable. By solving the model, the optimal weight of each indicator can be obtained for comprehensive evaluation in the data asset quality evaluation framework.
[0097] The application of the BWM method makes the evaluation process more simplified and operational, reduces the requirements for experts and evaluation data, and not only improves the accuracy of the evaluation, but also reduces the subjectivity and uncertainty in the evaluation process. Therefore, the BWM method plays a key role in determining the comprehensive weights of important indicators in the data asset quality evaluation framework, providing decision makers with reliable evaluation results and decision support. The specific steps are as follows:
[0098] C1. Each expert determines the most important and least important indicators in the evaluation system.
[0099] Expert P k Determine the most important indicator among n indicators for a specific scenario problem and the least important indicator Then, determine For the importance of all indicators, the higher the importance, the larger the value. Similarly, the importance of other indicators is determined. The importance of can be obtained by comparison:
[0100]
[0101] In the formula, Indicates P k The most important evaluation indicator For other indicators of importance, and Indicates P k Evaluate other indicators For the least important indicator of importance, and
[0102] C2. Calculate the optimal weights of the indicators.
[0103] Assume that based on the BWM method, P k The weight obtained by evaluation is In order to obtain the optimal weight of the indicator, and The value of j=1,2,...,n should be minimized, and the optimal weight of each indicator can be obtained by constructing model (1-9);
[0104] minξ
[0105] st
[0106]
[0107] When there are multiple experts, based on the expert weight information, the most effective indicator weight obtained by aggregating each expert is as follows:
[0108]
[0109] In the formula, YesP k The weight of the jth indicator given; λ k Expert P k The weight of .
[0110] D. Use triangular fuzzy numbers to describe the expert evaluation language, and apply the Fuzzy-TOPSIS method to rank the data quality status of different enterprises.
[0111] In order to determine the ranking of indicators, an expert evaluation matrix is constructed based on the evaluation information obtained from each expert. The matrix contains the collective expertise and opinions of the experts. Subsequently, the TOPSIS method is used to determine the ranking of data asset quality to obtain the final evaluation results. The specific steps are as follows:
[0112] D1. Construct an expert evaluation matrix.
[0113] Constructing triangular fuzzy numbers where x L ≦x M ≦x U , x L , x M , x U They are The lower bound element, main element and upper bound element of L =x M =x U When It degenerates into ordinary real numbers; suppose there are m asset sets to be evaluated, denoted by A = {A1, A2, ..., A m}; The indicator weights determined by the BWM method are α = {α1, α2, ..., α n}; i=1,2,...,m;j=1,2,...,n, represents the evaluation value of the jth indicator of the i-th data asset set; P k The matrix for evaluating n indicators of m data assets is as follows:
[0114]
[0115] When there are multiple experts, the evaluation of l experts is k=1, 2, ..., l for aggregation, and the comprehensive evaluation matrix is as follows:
[0116]
[0117] D2. Construct a weighted evaluation matrix.
[0118] The indicator weights determined by the BWM method are α = {α1, α2, ..., α n}, the evaluation matrix after weighting the indicators is as follows:
[0119]
[0120] D3. Determine PIS and NIS according to formulas (2-14) and (1-15).
[0121]
[0122] in, j=1, 2, ..., n is the fuzzy maximal set corresponding to the weighted index value of index j; j=1, 2, ..., n is the fuzzy minimum set corresponding to the weighted index value of index j.
[0123] D4. Calculate A according to formula (1-16) and (1-17) i Euclidean distance to PIS and NIS Then calculate the closeness D according to formula (1-18) i , get the final quality ranking, D i The larger the value, the higher the corresponding data asset quality.
[0124]
[0125] in,
[0126]
[0127] in,
[0128]
[0129] A storage medium stores a computer program or instruction, which implements the content described in the method when executed by a processor.
[0130] This paper combines DEMATEL, BWM and TOPSIS methods to propose a hybrid MCDM method for evaluating data asset quality. By combining related methods, this method provides a more effective method for evaluating the quality of data assets. By evaluating the quality of data assets, it helps to improve the quality management practices of enterprise data assets and promote the development of organizational data assets.
Claims
1. A data asset quality evaluation method based on hybrid multi-criteria decision making, characterized by: It includes the following steps: A. Combine the asset attributes of data to build a new data quality evaluation index system in the context of data assetization to evaluate the quality of enterprise data assets; B. During the model building process, the DEMATEL method was used to identify key indicators; C. Introduce the BWM method to determine the weights of key indicators; D. Use triangular fuzzy numbers to describe the expert evaluation language, and apply the Fuzzy-TOPSIS method to rank the data quality status of different enterprises.
2. The data asset quality evaluation method based on hybrid multi-criteria decision-making according to claim 1 is characterized by: In step A, the quality evaluation indicators include data integrity, consistency, uniqueness, timeliness, relevance, accuracy, ease of operation, comprehensibility and accessibility.
3. The data asset quality evaluation method based on hybrid multi-criteria decision-making according to claim 1 is characterized by: In step A, the data quality evaluation system includes three aspects: quality cost index, profitability index and risk index.
4. The data asset quality evaluation method based on hybrid multi-criteria decision making according to claim 1 is characterized by: In step B, the method of identifying key indicators using the DEMATEL method includes the following steps: B1. Decision makers evaluate the direct relationship between any two indicators and determine the included indicators C = {C1, C2, ..., C n }, establish expert group P k (k=1, 2, ..., l) Evaluate the mutual influence relationship and influence degree of two indicators; B2. Aggregate the expert evaluation matrix to obtain the initial direct relationship matrix; Each expert P k The evaluation of the impact relationship between indicators can obtain the corresponding direct impact matrix, in which all main diagonal elements are equal to 0. Indicates expert, P k Evaluation index, C i For indicator C j The initial direct impact matrix Z = [z ij ] n×n By combining the opinions of each expert, a matrix reflecting the influence relationship and influence degree between the two indicators is obtained, which can be obtained by aggregation through the following formula: B3. Establish a standardized impact matrix; According to the initial direct impact matrix, it is standardized to obtain the standardized impact matrix: X=λZ, (1-2)where, B4. Establish a comprehensive impact matrix; The standardized impact matrix is processed to the limit to obtain the comprehensive impact matrix, which represents the final direct and indirect impact relationships between indicators. The comprehensive impact matrix T can be calculated by the following formula: Where I is the identity matrix; B5. Calculate the centrality and causality of each indicator; Centrality is the sum of influence and influence, indicating the role of the indicator in the system; causality is the difference between influence and influence, also known as net influence. If causality is greater than 0, it represents a cause indicator, and if it is less than 0, it represents a result indicator. j , the calculation formula of its centrality is as follows: The calculation formula of the cause degree is as follows:
5. The data asset quality evaluation method based on hybrid multi-criteria decision making according to claim 1 is characterized by: In step C, the method for calculating the weight of key indicators using the BWM method includes the following steps: C1. Each expert determines the most important and least important indicators in the evaluation system; Expert P k Determine the most important indicator among n indicators for a specific scenario problem and the least important indicator Then, determine For the importance of all indicators, the higher the importance, the larger the value. Similarly, the importance of other indicators is determined. The importance of can be obtained by comparison: In the formula, Indicates P k The most important evaluation indicator For other indicators of importance, and Indicates P k Evaluate other indicators For the least important indicator of importance, and C2. Calculate the optimal weight of the indicator; Assume that based on the BWM method, P k The weight obtained by evaluation is In order to obtain the optimal weight of the indicator, and The value should be minimized, and the optimal weight of each indicator can be obtained by constructing models (1-9); When there are multiple experts, based on the expert weight information, the most effective indicator weight obtained by aggregating each expert is as follows: In the formula, YesP k The weight of the jth indicator given; λ k Expert P k The weight of .
6. The data asset quality evaluation method based on hybrid multi-criteria decision-making according to claim 1 is characterized by: In step D, the method for ranking the data quality status of different enterprises includes the following steps: D1. Construct expert evaluation matrix; Constructing triangular fuzzy numbers where x L ≦x M ≦x U , x L , x M , x U They are The lower bound element, main element and upper bound element of L =x M =x U When It degenerates into ordinary real numbers; suppose there are m asset sets to be evaluated, denoted by A = {A1, A2, ..., A m }; The indicator weight determined by the BWM method is α = {α1, α2, ..., α n }; i=1,2,...,m;j=1,2,...,n, represents the evaluation value of the jth indicator of the i-th data asset set; P k The matrix for evaluating n indicators of m data assets is as follows: When there are multiple experts, the evaluation of l experts is k=1, 2, ..., l for aggregation, and the comprehensive evaluation matrix is as follows: D2. Construct a weighted evaluation matrix; The indicator weights determined by the BWM method are α = {α1, α2, ..., α n }, the evaluation matrix after weighting the indicators is as follows: D3. According to formula (2-14) and (1-15), determine PIS and NIS; in, j=1, 2, ..., n is the fuzzy maximal set corresponding to the weighted index value of index j; j=1, 2, ..., n is the fuzzy minimum set corresponding to the weighted index value of index j; D4. Calculate A according to formula (1-16) and (1-17) i Euclidean distance to PIS and NIS Then calculate the closeness D according to formula (1-18) i , get the final quality ranking, D i The larger the value, the higher the corresponding data asset quality. in, in, 7. A storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the content of any method of claims 1 to 6 is implemented.
