A hierarchical structure weight calculation method and system based on time series data

Abstract

This invention relates to the field of data processing and decision analysis technology, specifically providing a hierarchical weight calculation method and system based on time-series data. The method includes: constructing a three-layer evaluation structure comprising a top layer, a middle layer, and a bottom layer; acquiring time-series data of bottom-layer indicator node elements within a preset time window, calculating the initial importance index of the bottom-layer indicator node elements, and correcting it to obtain the bottom-layer importance index; calculating the initial importance index of the middle layer and correcting it to obtain the middle-layer importance index; calculating the top-layer importance index of the top-layer target node elements based on a subjective weighting method; and aggregating and normalizing the bottom-layer importance index, the middle-layer importance index, and the top-layer importance index to obtain the final weight of each bottom-layer indicator node element relative to the top-layer target node element. The three-layer isolation ensures independent processing of subjective and objective information, and the final output weights can be directly applied to various comprehensive evaluation scenarios.

Description

Technical Field

[0001] This invention belongs to the field of data processing and decision analysis technology, and in particular relates to a method and system for calculating hierarchical structure weights based on time series data. Background Technology

[0002] In the field of multi-attribute decision-making and comprehensive evaluation, determining the weight coefficients is a core step in determining the scientific validity and rationality of the evaluation results. Based on the different sources and methods of determining the weight information, existing methods are mainly divided into subjective weighting methods, objective weighting methods, and combined weighting methods. Subjective weighting methods rely on the prior knowledge and experience of decision-makers or experts, and can better reflect the strategic intentions and preferences of decision-makers; however, the results of this method have poor stability and are easily affected by the subjective factors and cognitive biases of experts, raising questions about its objectivity and repeatability. Objective weighting methods are based on the observed data of the evaluation objects, determining weights by mining the inherent differences in the data, and have good mathematical objectivity and interpretability. However, purely objective weighting cannot incorporate the subjective preferences of decision-makers, and its calculation results are easily interfered with by the structural characteristics of the evaluation index system itself, potentially leading to weight results that contradict actual importance. To combine the advantages of both methods, combined weighting methods have become the current main approach, employing subjective or objective methods to assign weights at different levels, and then synthesizing the final weights through linear weighting or other aggregation strategies.

[0003] However, this method has inherent limitations in complex evaluation systems with hierarchical structures: First, existing combined weighting methods typically mix the objective weights of lower-level indicators with the subjective weights of upper-level criteria in the final synthesis stage. This approach ignores the fundamental differences between subjective preferences and objective laws in terms of information nature. Existing technologies directly add or multiply them at the numerical level, resulting in a final weight that neither purely represents the inherent differences in the data nor accurately reflects the true intentions of decision-makers. This creates a mixed metric lacking clear semantic direction, leading to information distortion.

[0004] Secondly, in reality, evaluation systems often have uneven tree-like topologies, with different criterion nodes potentially having vastly different numbers of indicators. Existing methods largely implicitly assume a uniform structural distribution in their mathematical derivations. When this assumption is not met, branches with a larger number of indicators receive higher cumulative weights due to size rather than importance. This systematic bias introduced by the inherent structural heterogeneity of the system is not identified or corrected in existing methods, resulting in a lack of horizontal comparability of evaluation results across different branches.

[0005] In summary, existing technologies have not yet solved the technical challenge of how to remove structural interference and achieve deep integration of subjective and objective information in complex, heterogeneous, multi-level evaluation scenarios. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method for calculating the weights of a hierarchical structure based on time-series data, comprising the following steps: Step S1: Construct a three-layer evaluation structure consisting of a top layer, a middle layer, and a bottom layer, wherein each upper-layer node element contains at least one lower-layer node element as its child node element, and the upper-layer node element is the parent node element of the child node element. Step S2: Obtain the time series data of the underlying indicator node elements within a preset time window, calculate the underlying initial importance index of the underlying indicator node elements based on the time series data, and correct the underlying initial importance index based on the number of underlying indicator node elements contained in the parent node element of the underlying indicator node element to obtain the underlying importance index. Step S3: Obtain the underlying indicator node elements contained in the intermediate layer criterion node elements; calculate the intermediate layer initial importance index of the intermediate layer criterion node element based on the underlying indicator node elements and their underlying importance index; correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node element to obtain the intermediate layer importance index. Step S4: Based on the decision preference information input from the outside, calculate the top importance index of the top target node element, where the decision preference information input from the outside is a judgment matrix; Step S5: Perform nonlinear multiplication and aggregation of the bottom layer importance index, the middle layer importance index, and the top layer importance index, and normalize the aggregation result to obtain the final weight of each bottom layer indicator node element relative to the top layer target node element.

[0007] Based on the above scheme, the correction of the initial importance index in step S2 includes: Obtain the first structure correction coefficient, which is the ratio of the number of bottom index node elements contained in the intermediate criterion node element to which the bottom index node element belongs to to the average number of bottom index node elements contained in the intermediate criterion node element to which the bottom index node element belongs to. The product of the first structural correction coefficient and the initial importance index of the bottom layer is used as the corrected bottom layer importance index.

[0008] Preferably, the method for calculating the initial importance index of the underlying layer is as follows: The time series data is subjected to longitudinal standardization to obtain standardized time series data; Based on the standardized time-series data, the initial importance index of the underlying indicator node elements is calculated using the information entropy method.

[0009] Based on the above scheme, the correction of the initial importance index of the intermediate layer in step S3 includes: Obtain the second structure correction coefficient, which is the ratio of the number of intermediate layer criterion node elements contained in the top-level target node element to which the intermediate layer criterion node element belongs to the average number of intermediate layer criterion node elements contained in each of the top-level target node elements. The product of the second structural correction coefficient and the initial importance index of the intermediate layer is used as the corrected importance index of the intermediate layer.

[0010] Preferably, the method for calculating the initial importance index of the intermediate layer is as follows: Obtain the underlying importance index of the underlying indicator node elements contained in the intermediate layer criterion node elements; The weighted result is obtained by multiplying the underlying importance index with the standardized values ​​of its corresponding underlying indicator node elements in a weighted manner. The weighted results are summed to generate a sequence of representative values ​​for the intermediate layer criterion node elements; Calculate the coefficient of variation of the representative value sequence, and use it as the initial importance index of the intermediate layer criterion node element.

[0011] Preferably, the method for calculating the top-level importance index is as follows: Based on the judgment matrix, the single sorting technique in the analytic hierarchy process is used to calculate the top-level importance index of each top-level target node element.

[0012] Preferably, the longitudinal standardization process is the range method: For efficiency-type indicators: , For cost-related indicators: , in, For the first i Each underlying indicator node element t The original observations at time [time]. For the first i Each underlying indicator node element t Standardized value of time, For the first i The maximum value of each underlying indicator node element across the entire time series dataset. For the first i The minimum value of each underlying indicator node element across the entire time series dataset.

[0013] Preferably, the first or second structural correction coefficient is subjected to a logarithmic or polynomial transformation to correct the initial importance index.

[0014] Based on the above scheme, the step of using the single sorting technique in the analytic hierarchy process to calculate the top-level importance index of each top-level target node element further includes: Based on the judgment matrix, calculate the consistency index and the relative consistency index; When the relative consistency index is greater than a preset threshold, the judgment matrix is ​​corrected and recalculated until the relative consistency index is less than or equal to the preset threshold.

[0015] On the other hand, the present invention provides a hierarchical weight calculation system based on time-series data, the system comprising: Architecture building module, used to build a three-tier evaluation structure consisting of a top layer, a middle layer, and a bottom layer; The underlying processing module is used to acquire time-series data of underlying indicator node elements within a preset time window, calculate the underlying initial importance index of each underlying indicator node element based on the time-series data, correct the initial importance index according to the number of underlying indicator node elements under the intermediate layer criterion node element to which each indicator node element belongs, and output the underlying importance index of each underlying indicator node element. The intermediate layer processing module is used to obtain the underlying indicator node elements contained in the intermediate layer criterion node elements, calculate the intermediate layer initial importance index of the intermediate layer criterion node elements based on the underlying indicator node elements and their underlying importance indices, and correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node elements to obtain the intermediate layer importance index. The top-level processing module is used to calculate the top-level importance index of each target node element based on the decision preference information input from external sources. The aggregation output module is used to aggregate the bottom-level importance index, the middle-level importance index, and the top-level importance index, and to normalize the aggregation result, outputting the final weight of each indicator node element relative to the top-level target node element.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves deep integration of subjective and objective weights at the system architecture level through hierarchical isolation and structural correction. The resulting weights not only objectively reflect the dynamic patterns contained in time-series data, but also accurately implement the strategic intentions of decision-makers. At the same time, it eliminates the systematic biases introduced by the imbalance of hierarchical structure, providing more scientific and reliable weight support for complex multi-attribute decision-making. 2. By introducing a structural correction coefficient, the ratio of the number of child nodes under a parent node to the average value of the same layer is used as a correction factor to compensate for the initial importance index. This fundamentally eliminates the systematic bias introduced by the unbalanced structure of the evaluation model and significantly improves the horizontal fairness and comparability of the evaluation results across different branches and elements. 3. The importance index is calculated at different levels. At the bottom level, the time entropy method is used to capture the micro-time uncertainty of the indicators. At the middle level, the coefficient of variation method is used to measure the macro-time evolution sensitivity of information elements. At the top level, the single ranking technique in the analytic hierarchy process is used to reflect the subjective preferences of decision-makers. This maximizes the mining and utilization of information from different dimensions in the data, and the model has strong explanatory power. 4. The outputs of each level and the system aggregation stage are all non-normalized intensity scales, which break away from the linear constraints of traditional weights and can more realistically depict the unbalanced importance distribution of each element in a complex system; the differences between the indices directly reflect the actual differences in importance between elements, providing decision-makers with a richer and more reliable information foundation. Attached Figure Description

[0017] Figure 1 This is a flowchart of the overall method for weight calculation in this invention; Figure 2 This is a flowchart illustrating the specific method for weight calculation in this invention; Figure 3 This is a schematic diagram of the weight calculation system of the present invention. Detailed Implementation

[0018] The invention will be further described below with reference to specific embodiments. It should be understood that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0019] This embodiment uses the evaluation of sustainable development capacity in a certain region as an example to provide a detailed explanation of the hierarchical structure weight calculation method based on time-series data described in this invention. Figure 1 and Figure 2 As shown, this embodiment provides a method for calculating the weights of a hierarchical structure based on time-series data, including the following steps: Step S1: Construct a three-layer evaluation structure consisting of a top layer, a middle layer, and a bottom layer, wherein each upper-layer node element contains at least one lower-layer node element as its child node element, and the upper-layer node element is the parent node element of the child node element. As shown in Table 1, this embodiment constructs a three-layer regional sustainable development capacity assessment system comprising a target layer, a criterion layer, and an indicator layer; Table 1 Evaluation Index System for Regional Sustainable Development Capacity

[0020] The evaluation system includes: Top-level T: Includes top-level target node elements of natural ecological environment system T1 and socio-economic system T2; The intermediate layer M: The sub-node elements under T1 include: environmental governance M1, climate change response M2, and ecological protection M3 intermediate layer node elements; the sub-node elements under T2 include: economic development M4, innovation-driven development M5, people's livelihood and well-being M6, and security guarantee M7 intermediate layer node elements.

[0021] Bottom layer B: Bottom layer indicator node elements include: The indicator nodes belonging to M1 include: fine particulate matter (PM2.5) concentration B1, comprehensive ambient air quality index B2, proportion of nearshore sea water quality with good (Class I and II) B3, proportion of rivers flowing into the sea with water quality reaching or better than Class III water bodies B4, total nitrogen oxide emissions B5, total chemical oxygen demand emissions B6, and total ammonia nitrogen emissions B7. The indicator nodes belonging to M2 include: energy consumption per unit of GDP (B8), industrial carbon emission intensity based on energy consumption (B9), and water consumption per 10,000 yuan of GDP (B10). The indicator nodes belonging to M3 include: per capita arable land area B11, the proportion of ecological water use to total water use B12, fertilizer application intensity B13, spring seawater inflow B14, and green coverage rate of built-up areas B15. The indicator nodes belonging to M4 include: per capita GDP (B16), total labor productivity (B17), and urbanization rate of permanent residents (B18). The indicator nodes belonging to M5 include: the intensity of R&D investment in the whole society (B19) and the efficiency of effective invention patent output per unit of human resources in large-scale industrial enterprises (B20). The indicator nodes belonging to M6 include: per capita disposable income of residents (B21), total population employment coefficient (B22), number of practicing (assistant) physicians per thousand people (B23), and basic old-age insurance participation rate (B24). The indicator node elements belonging to M7 include: comprehensive grain production capacity B25 and comprehensive energy production capacity B26.

[0022] Step S2: Obtain the time series data of the underlying indicator node elements within a preset time window, calculate the underlying initial importance index of the underlying indicator node elements based on the time series data, and correct the underlying initial importance index based on the number of underlying indicator node elements contained in the parent node element of the underlying indicator node element to obtain the underlying importance index. Step S2 specifically includes: S21: For each underlying indicator node element, obtain its time series data within a preset time window, perform vertical standardization on the time series data, and obtain standardized time series data. In this embodiment, the vertical standardization process is the range method: For benefit-related indicators, calculate the positive standardized value: , For cost-related indicators, calculate the negative standardized value: , in, For the first i Each underlying indicator node element t The original observations at time [time]. For the first i Each underlying indicator node element t Standardized value of time, For the first i The maximum value of each underlying indicator node element across the entire time series dataset. For the first i The minimum value of each underlying indicator node element across the entire time series dataset.

[0023] S22: Perform a small shift on the standardized time series data to obtain non-zero standardized values: ;in, ; S23: Perform vertical normalization on the non-zero standardized values ​​to obtain the normalized values: ; S24: Calculate the initial importance index of the underlying indicator node elements using the information entropy method: First, the information entropy value of each underlying indicator node element is calculated based on the normalized value: ;in, n is the number of samples within the preset time window; Secondly, based on the information entropy value, the first... i The information utility value of each underlying indicator node element: ; Based on information utility value, the first i The initial importance index of each underlying indicator node element: .

[0024] S25: Correct the initial importance index of the bottom layer: Obtain the first structural correction coefficient: ;in, This represents the number of underlying indicator node elements contained in the intermediate-level criterion node element to which the underlying indicator node element belongs. This is the average number of underlying indicator node elements contained in each of the intermediate criterion node elements under the top-level target node element to which the underlying indicator node element belongs. First structural correction coefficient relative to the initial importance index of the underlying layer The product of these is used as the corrected underlying importance index: .

[0025] This embodiment obtains the importance index of the bottom-level indicator node elements in the evaluation index system through the above step S2, as shown in Table 2; Table 2. Overview of Importance Indicators of Node Elements in the Bottom-Level Indicator System

[0026] Further, step S3 is executed to obtain the underlying indicator node elements contained in the intermediate layer criterion node elements, calculate the intermediate layer initial importance index of the intermediate layer criterion node element based on the underlying indicator node elements and their underlying importance index, and correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node element to obtain the intermediate layer importance index. Step S3 in this embodiment specifically includes: S31: Obtain the underlying importance index of the underlying indicator node elements contained in the intermediate layer criterion node elements. ; The weighted result is obtained by multiplying the underlying importance index by the standardized values ​​of its corresponding underlying indicator node elements; the weighted result is then summed to generate the representative value sequence of the intermediate layer criterion node elements. ;in, It is the set of all bottom-level indicator nodes that belong to the intermediate-level criterion node element j.

[0027] S32: Calculate the coefficient of variation of the representative value sequence, which serves as the initial importance index of the intermediate layer criterion node elements: ;in, For intermediate layer criterion node elements j The standard deviation of the representative value sequence within a preset time window The average value of the representative value sequence of intermediate layer criterion node element j within a preset time window; S33: Obtain the second structural correction coefficient: ;in, This represents the number of intermediate-level criterion node elements subordinate to the top-level target node element to which the intermediate-level criterion node element belongs. This is the average number of intermediate-level criterion node elements under all top-level target node elements; S34: Adjust the second structural correction factor Initial importance index of intermediate layer The product of these is used as the corrected intermediate layer importance index: .

[0028] In this embodiment, the importance index of each criterion node element in the intermediate layer of the evaluation index system is obtained through step S3, as shown in Table 3.

[0029] Table 3. Overview of Importance Indices of Node Elements in the Intermediate Layer of the Evaluation Index System

[0030] Step S4: Calculate the top-level importance index of the top-level target node element based on the decision preference information input from the external source; Specifically, the calculation method for the top-level importance index is as follows: S41: Construct the judgment matrix: Use the pairwise comparison method to perform pairwise comparison and scoring on the elements of the top-level target node. The scoring scale adopts the quartile relative importance scale (as shown in Table 4) to obtain the pairwise comparison judgment matrix. Table 4. Ninth percentile scale

[0031] This embodiment assumes that k experts participate in determining the relative importance scale of the indicators; let the x-th expert consider the relative importance of the i-th indicator to the j-th indicator to be... (x=1, 2, 3, ..., k), then the judgment matrix obtained based on the evaluation result of the x-th expert is: (n is the number of indicators); in, , ;when hour, The geometric mean of the expert judgments: .

[0032] When the consensus of the judgment matrices given by the experts is poor, that is, when there is a large discrepancy in the judgments on the relative importance of a certain two indicators, the experts can be reorganized based on the Delphi method to conduct a new round of consultation and judgment until the consensus of the experts on the relative importance of all indicators is good, so as to improve the accuracy and consistency of the experts' judgments.

[0033] S42: Calculate the eigenvectors of the judgment matrix Then, normalize the vector to obtain the importance index vector of the top-level target node elements: ; S43: Calculate the largest eigenvalue based on the judgment matrix A and the importance index vector W: ; S44: Calculate the consistency index CI and the relative consistency index CR, and verify the consistency of the judgment matrix A: ; ; where RI is the random consistency index, which can be obtained by looking up a table.

[0034] S45: If the consistency requirement is not met, the relevant elements in the original judgment matrix A are corrected until CR ≤ the preset threshold. Since the smaller the CR, the better the consistency of the judgment matrix, it is generally considered that when CR ≤ 0.1, that is, when the preset threshold is 0.1, the judgment matrix has satisfactory consistency.

[0035] S46: Based on the judgment matrix that meets the consistency requirements, calculate the importance index of the top-level target node elements. .

[0036] In this embodiment, the importance index of each top-level target node element in the evaluation index system is obtained through step S4, as shown in Table 5.

[0037] Table 5. Overview of the Importance Index of Top-Level Target Node Elements in the Evaluation Index System

[0038] Step S5, assign the underlying importance index Importance index of intermediate layer and the top-level importance index Perform nonlinear product aggregation: The aggregation results are then normalized. This yields the final weights of each underlying indicator node element relative to the top-level target node element.

[0039] In this embodiment, the final weights of each underlying indicator node element relative to the decision target node element are obtained through step S5, as shown in Table 6.

[0040] Table 6. A summary of the final weights of each bottom-level indicator node element relative to the top-level target node element.

[0041] According to another preferred embodiment, in order to obtain a better correction effect, the first structural correction coefficient or the second structural correction coefficient is mathematically transformed according to actual needs, such as logarithmic transformation or polynomial transformation, and the transformed correction coefficient is used to correct the initial importance index.

[0042] The method of this invention provides a weight that objectively reflects the information content and differences of each indicator in the time-series evolution, accurately reflects the strategic preferences of decision-makers for the natural ecological environment and socio-economic system, and eliminates the systematic bias caused by the unbalanced structure of the evaluation system itself, thus providing reliable weight support for the scientific evaluation of regional sustainable development capacity.

[0043] The weights calculated using the method of this invention have significantly lower weight deviations between different branches compared to existing methods. This solves the technical problem that existing combined weighting methods simply mix the objective weights of the underlying indicators with the subjective weights of the upper-level criteria in the final synthesis stage, leading to mutual interference and coupling distortion between the two types of information with different properties.

[0044] Example 2 Based on the same technical concept, this invention also provides a hierarchical weight calculation system based on time-series data, such as... Figure 3 As shown, the system includes: Architecture building module, used to build a three-tier evaluation structure consisting of a top layer, a middle layer, and a bottom layer; The underlying processing module is used to obtain the time series data of the underlying indicator node elements within a preset time window, calculate the underlying initial importance index of each underlying indicator node element based on the time series data, correct the initial importance index according to the number of underlying indicator node elements under the intermediate layer criterion node element to which each indicator node element belongs, and output the underlying importance index of each underlying indicator node element. Specifically, the underlying processing module includes: The underlying initial importance index calculation module is used to calculate the initial importance index of the underlying indicator node elements based on the time series data of all underlying indicator node elements within a preset time window. The first structure correction coefficient calculation module is used to determine the first structure correction coefficient based on the number of underlying indicator node elements contained in the parent node element of the underlying indicator node element, and to use the first structure correction coefficient to correct the initial importance index of all underlying indicator node elements under the corresponding intermediate layer criterion node element to obtain the underlying importance index. The intermediate layer processing module is used to obtain the underlying indicator node elements contained in the intermediate layer criterion node elements, calculate the intermediate layer initial importance index of the intermediate layer criterion node elements based on the underlying indicator node elements and their underlying importance indices, and correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node elements to obtain the intermediate layer importance index. The intermediate layer processing module specifically includes: The intermediate layer representative value sequence generation module is used to aggregate and generate a representative value sequence of all intermediate layer criterion node elements based on the underlying importance index of all underlying indicator node elements under each intermediate layer criterion node element and the normalized time series value corresponding to the underlying indicator node element. The intermediate layer initial importance index calculation module is used to calculate the intermediate layer initial importance index of all intermediate layer criterion node elements based on the representative value sequence of each intermediate layer criterion node element. The second structure correction coefficient calculation module is used to determine the second structure correction coefficient based on the number of intermediate layer criterion node elements contained in the parent node element of the intermediate layer criterion node element, and to use the second structure correction coefficient to correct the initial importance index of the intermediate layer of all intermediate layer criterion node elements under the corresponding top-level target node element, so as to obtain the intermediate layer importance index. The top-level processing module is used to calculate the top-level importance index of each target node element based on the decision preference information input from the external source; wherein, the decision preference information input from the external source is the judgment matrix obtained by the analytic hierarchy process.

[0045] The aggregation output module is used to aggregate the bottom-level importance index, the middle-level importance index, and the top-level importance index, and to normalize the aggregation result, outputting the final weight of each indicator node element.

[0046] The specific implementation of the hierarchical weight calculation system based on time-series data in this embodiment is the same as that in Embodiment 1, and will not be repeated here.

[0047] In general, various exemplary embodiments of the present invention can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of the present invention are illustrated or described as block diagrams, flowcharts, or represented using certain other images, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or certain combinations thereof.

[0048] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0049] Although the specific embodiments of the present invention have been described above, they are not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions of the present invention, various modifications or variations that can be made by those skilled in the art without creative effort are still within the scope of protection of the present invention.

Claims

1. A method for calculating the weights of a hierarchical structure based on time-series data, characterized in that, Includes the following steps: Step S1: Construct a three-layer evaluation structure consisting of a top layer, a middle layer, and a bottom layer, wherein each upper-layer node element contains at least one lower-layer node element as its child node element, and the upper-layer node element is the parent node element of the child node element. Step S2: Obtain the time series data of the underlying indicator node elements within a preset time window, calculate the underlying initial importance index of the underlying indicator node elements based on the time series data, and correct the underlying initial importance index based on the number of underlying indicator node elements contained in the parent node element of the underlying indicator node element to obtain the underlying importance index. Step S3: Obtain the underlying indicator node elements contained in the intermediate layer criterion node elements; calculate the intermediate layer initial importance index of the intermediate layer criterion node element based on the underlying indicator node elements and their underlying importance index; correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node element to obtain the intermediate layer importance index. Step S4: Based on the decision preference information input from the outside, calculate the top-level importance index of the top-level target node element, where the decision preference information input from the outside is a judgment matrix; Step S5: Perform nonlinear multiplication and aggregation of the bottom layer importance index, the middle layer importance index, and the top layer importance index, and normalize the aggregation result to obtain the final weight of each bottom layer indicator node element relative to the top layer target node element.

2. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 1, characterized in that, The step S2 of correcting the initial importance index of the underlying layer includes: Obtain the first structure correction coefficient, which is the ratio of the number of bottom index node elements contained in the intermediate criterion node element to which the bottom index node element belongs to to the average number of bottom index node elements contained in the intermediate criterion node element to which the top target node element belongs to. The product of the first structural correction coefficient and the initial importance index of the bottom layer is used as the corrected bottom layer importance index.

3. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 1, characterized in that, The method for calculating the initial importance index at the bottom layer is as follows: The time series data is subjected to longitudinal standardization to obtain standardized time series data; Based on the standardized time-series data, the initial importance index of the underlying indicator node elements is calculated using the information entropy method.

4. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 1, characterized in that, The step S3 of correcting the initial importance index of the intermediate layer includes: Obtain the second structure correction coefficient, which is the ratio of the number of intermediate layer criterion node elements contained in the top-level target node element to which the intermediate layer criterion node element belongs to the average number of intermediate layer criterion node elements contained in each of the top-level target node elements. The product of the second structural correction coefficient and the initial importance index of the intermediate layer is used as the corrected importance index of the intermediate layer.

5. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 1, characterized in that, The method for calculating the initial importance index of the intermediate layer is as follows: Obtain the underlying importance index of the underlying indicator node elements contained in the intermediate layer criterion node elements; The weighted result is obtained by multiplying the underlying importance index with the standardized values ​​of its corresponding underlying indicator node elements in a weighted manner. The weighted results are summed to generate a sequence of representative values ​​for the intermediate layer criterion node elements; Calculate the coefficient of variation of the representative value sequence, and use it as the initial importance index of the intermediate layer criterion node element.

6. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 1, characterized in that, The calculation method for the top-level importance index is as follows: Based on the judgment matrix, the single sorting technique in the analytic hierarchy process is used to calculate the top-level importance index of each top-level target node element.

7. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 3, characterized in that, The longitudinal standardization process is the range method: For efficiency-type indicators: , For cost-related indicators: , in, For the first i Each underlying indicator node element t The original observations at time [time]. For the first i Each underlying indicator node element t Standardized value of time, For the first i The maximum value of each underlying indicator node element across the entire time series dataset. For the first i The minimum value of each underlying indicator node element across the entire time series dataset.

8. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 2, characterized in that, The first structural correction coefficients are subjected to logarithmic or polynomial transformation, and the initial importance index is corrected using the transformed correction coefficients.

9. The method for calculating the weights of a hierarchical structure based on time-series data according to claim 6, characterized in that, The method of calculating the top-level importance index of each top-level target node element using the single sorting technique in the analytic hierarchy process also includes: Based on the judgment matrix, calculate the consistency index and the relative consistency index; When the relative consistency index is greater than a preset threshold, the judgment matrix is ​​corrected and recalculated until the relative consistency index is less than or equal to the preset threshold.

10. A hierarchical weight calculation system based on time-series data, characterized in that, The system includes: Architecture building module, used to build a three-tier evaluation structure consisting of a top layer, a middle layer, and a bottom layer; The underlying processing module is used to acquire time-series data of underlying indicator node elements within a preset time window, calculate the underlying initial importance index of each underlying indicator node element based on the time-series data, correct the initial importance index according to the number of underlying indicator node elements under the intermediate layer criterion node element to which each indicator node element belongs, and output the underlying importance index of each underlying indicator node element. The intermediate layer processing module is used to obtain the underlying indicator node elements contained in the intermediate layer criterion node elements, calculate the intermediate layer initial importance index of the intermediate layer criterion node elements based on the underlying indicator node elements and their underlying importance indices, and correct the intermediate layer initial importance index based on the number of intermediate layer criterion node elements contained in the parent node elements of the intermediate layer criterion node elements to obtain the intermediate layer importance index. The top-level processing module is used to calculate the top-level importance index of each target node element based on the decision preference information input from external sources. The aggregation output module is used to aggregate the bottom-level importance index, the middle-level importance index, and the top-level importance index, and to normalize the aggregation result, outputting the final weight of each indicator node element relative to the top-level target node element.

Images