Project performance evaluation method, system and equipment based on weight optimization method and medium

Abstract

The invention discloses a project performance evaluation method, system and device based on a weight optimization method, and a medium. The method comprises the following steps: S1, inputting an initial project performance evaluation index system; s2, constructing a comparison judgment matrix, and calculating a consistency ratio CR, if CRlt; 0.1, if so, passing the consistency test, and performing the step S4; if CR is greater than or equal to 0.1, adjusting the comparison judgment matrix, and performing the step S3; s3, automatically optimizing the comparison judgment matrix, and recalculating CR until CRlt; 0.1, optimization is stopped, and the step S4 is carried out after consistency check; and S4, outputting the effective weight value of each index if the consistency test is passed, and obtaining a final performance evaluation index system. The method has the advantages that the defects that when a classical analytic hierarchy process is used for performance evaluation, the workload is large, dynamic optimization adjustment is difficult and the like are overcome, efficient data processing and analysis are achieved, especially when the number of orders of a comparison judgment matrix is large, the adjustment efficiency can be improved, the number of iterations can be reduced, and the weight accuracy can be improved. And the scientificity and practicability of the analytic hierarchy process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a project performance evaluation method, system, device, and medium based on a weight optimization method. Background Art

[0002] Project performance evaluation plays an extremely important role in project management, and can help managers understand the current execution situation of the project and the resource utilization efficiency. The evaluation results can provide an important basis for whether the project should continue, whether it needs to adjust the direction, or whether it should be terminated. The analytic hierarchy process (AHP) is a decision-making analysis method that combines qualitative and quantitative methods. This method can transform the subjective judgments of experts into quantitative data, avoiding evaluation biases caused by subjective factors. At the same time, this method verifies the effectiveness of the judgment matrix through a consistency test to ensure the scientificity and rationality of the evaluation process. However, in the prior art, the consistency adjustment of the AHP often relies on manual repeated modification, which is not only inefficient but also easily affected by subjective factors. Therefore, it is very necessary to automatically optimize the effectiveness of the judgment matrix of the project performance evaluation system by writing a program.

[0003] Chinese Patent Publication No. CN114723277A discloses a method, device, equipment, medium, and program product for performance evaluation, including: obtaining historical relevant data related to performance evaluation, where the historical relevant data includes multiple indicators used by multiple users during the performance evaluation; determining multiple target indicators for performance evaluation based on the historical relevant data; using the multiple target indicators to build a multi-layer indicator model, where the multi-layer indicator model includes a target layer, an intermediate layer, and a scheme layer, the intermediate layer includes multiple levels of indicators associated with the target layer, and the scheme layer includes at least one third-level indicator associated with the intermediate layer; using the AHP to configure weights for each indicator in the built multi-layer indicator model, where the weight of each indicator is the relative weight between the indicator and the associated upper-level indicator; performing dimension normalization processing on at least one third-level indicator according to a preset rule to obtain the normalized dimension of each third-level indicator; and determining the performance information of the target layer based on the weight of each indicator and the normalized dimension of at least one third-level indicator. This patent mentions that when the CR of the judgment matrix < 0.1 or when λ max = n and CI = 0, it can be considered that the judgment matrix has satisfactory consistency, otherwise the elements in the judgment matrix need to be adjusted to make it have satisfactory consistency. When the total ranking consistency CR < 0.1, it means that the total ranking consistency test is passed, otherwise the model needs to be reconsidered or those judgment matrices with a larger consistency ratio CR need to be reconstructed. However, it does not mention how to adjust the elements in the judgment matrix or how to adjust the model.

[0004] Chinese Invention Patent Publication No. CN118780634A discloses a method for route selection of long-distance pipelines based on the AHP-TOPSIS method. By constructing a hierarchical model (goal layer - criterion layer - scheme layer), a judgment matrix is constructed by pairwise comparison of factor indicators. Based on the Analytic Hierarchy Process (AHP), the index weights are calculated, and the comprehensive evaluation of each scheme is quantified by the TOPSIS method to achieve the scientific and objective route selection. This patent also does not disclose how to handle it when CR≥0.1. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that the consistency adjustment of the current Analytic Hierarchy Process often relies on repeated manual modification, which is not only inefficient but also easily affected by subjective factors. Therefore, a performance evaluation method, system, device and medium based on the weight optimization method are provided.

[0006] The technical solution of the present invention is: A project performance evaluation method based on the weight optimization method, including the following steps:

[0007] S1: Input the initial project performance evaluation index system;

[0008] S2: Construct a comparison judgment matrix, calculate the consistency ratio CR. If CR < 0.1, pass the consistency test and proceed to step S4; if CR≥0.1, adjust the comparison judgment matrix and proceed to step S3;

[0009] S3: Automatically optimize the comparison judgment matrix, recalculate CR, and stop optimizing until CR < 0.1. Pass the consistency test and proceed to step S4; The automatic optimization includes: ① Oscillation detection: Calculate the variance of the CR value changes in N consecutive iterations; ② Construct a dynamic decay function of the learning rate α: When it is detected that the variance of the CR value changes in N consecutive iterations is greater than the threshold T, automatically reduce the adjustment learning rate α according to the exponential decay law; ③ Lower limit protection of the learning rate: During the automatic optimization process, set the minimum learning rate αmin to ensure that the learning rate is not lower than αmin; ④ Periodic reset of the learning rate: When CR < 0.1 is not achieved for K consecutive times, reset the learning rate to the initial learning rate;

[0010] S4: If the consistency test is passed, output the effective weight values of each index to obtain the final performance evaluation index system.

[0011] In the above solution, the oscillation detection adopts the sliding window variance calculation method, the window size N ∈ [5, 15], and the variance threshold T ∈ [0.005, 0.03].

[0012] In the above solution, the learning rate decay execution module performs exponential decay according to α←α×μ, where μ ∈ (0, 1) is a preset decay coefficient.

[0013] In the above solution, the attenuation coefficient μ ∈ (0.5, 0.9).

[0014] In the above solution, the lower limit αmin of the minimum learning rate ∈ [0.05, 0.2].

[0015] In the above solution, the reset threshold K for learning rate cycle reset ∈ [8, 15].

[0016] A project performance evaluation system based on the weight optimization method adopts the project performance evaluation method described in any one of the above. This system includes the following modules:

[0017] (1) A module for constructing the performance evaluation index system of the project;

[0018] (2) A module for constructing a comparison judgment matrix, calculating the eigenvector and the maximum eigenvalue λ of this judgment matrix, obtaining the consistency ratio CR; if the consistency ratio CR < 0.1, then the consistency test is passed; if CR ≥ 0.1, then the judgment matrix is adjusted;

[0019] (3) A module for automatically optimizing the comparison judgment matrix, including an oscillation detection module, a learning rate attenuation execution module, a learning rate lower limit protection module, and a learning rate cycle reset module;

[0020] (4) A module for outputting the performance evaluation index system, after passing the consistency test, outputting the effective weight values of each index, and obtaining the final performance evaluation index system.

[0021] A computer device based on the weight optimization method, the computer device includes: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the project performance evaluation method described in any one of the above.

[0022] A computer storage medium based on the weight optimization method, on which a computer program is stored, and when this program is executed by a processor, it implements the project performance evaluation method described in any one of the above.

[0023] The beneficial effect of the present invention is to overcome the defects of the classical analytic hierarchy process for performance evaluation, such as large workload and difficult dynamic optimization and adjustment, to achieve efficient data processing and analysis. Especially when the order of the comparison judgment matrix is relatively large, it can improve the adjustment efficiency, reduce the number of iterations, and improve the weight accuracy, etc., thereby enhancing the scientificity and practicality of the analytic hierarchy process. Description of the Drawings

[0024] Figure 1 It is a flow schematic diagram of the project performance evaluation method based on the weight optimization method of the present invention;

[0025] Figure 2It is part of the interface 1 for automatic optimization implemented in Python;

[0026] Figure 3 It is part of the interface 2 for automatic optimization implemented in Python. Specific implementation manner

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] (1) Input the initial project performance evaluation index system, as shown in Table 1.

[0029] Table 1 Project performance evaluation index system table

[0030]

[0031] (2) Through the expert consultation method, a comparison judgment matrix was constructed for the third-level indicators, as shown in Table 2. Calculate the eigenvector and the maximum eigenvalue λ of this judgment matrix, and obtain the consistency ratio CR of 0.1483. Since CR ≥ 0.1, the judgment matrix needs to be adjusted.

[0032]

[0033] (3) Automatically optimize the comparison judgment matrix. The present invention uses a Python program based on Pandas and NumPy for the analytic hierarchy process (AHP) to automatically optimize the judgment matrix. This program iteratively adjusts the matrix elements to gradually reduce the consistency ratio (CR) until the consistency requirement is met.

[0034] The present invention realizes the oscillation detection and the dynamic adjustment of the learning rate. By defining a sliding window to store the recent CR values and calculating the variance. When the variance exceeds the threshold, the learning rate decay is triggered. At the same time, a counter is maintained to track the number of consecutive non-improved iterations to decide whether to reset the learning rate. In addition, the adjustment of the matrix needs to ensure reciprocity, that is, a_ij = 1 / a_ji, so it is necessary to check and correct when adjusting each element.

[0035] The present invention includes the following steps:

[0036] Step 1, oscillation detection: Calculate the variance of the CR value changes in N consecutive iterations. When the variance exceeds the threshold T, the learning rate decay is triggered. The sliding window variance method is used to monitor the CR value fluctuations in real time. When the variance within the detection window exceeds the threshold (default 0.01), the learning rate decay is triggered.

[0037] Step 2, dynamic decay of the learning rate: When oscillation is detected, the learning rate α is adjusted by exponential decay, according to the exponential decay of α ← α × μ (μ ∈ (0, 1)), and the decay coefficient μ is between 0.5 and 0.9. At the same time, a learning rate history record is provided for analysis.

[0038] Step 3, Lower Bound Protection of Learning Rate: Ensure that the learning rate is not lower than α min , with a range between 0.05 and 0.2.

[0039] Step 4, Learning Rate Period Reset: If the CR < 0.1 is not achieved in 10 consecutive iterations, reset the learning rate to the initial value to prevent getting stuck in a local optimal solution.

[0040] At the same time, perform linear interpolation based on the theoretical values of the eigenvectors to forcibly maintain matrix reciprocity and scale constraints (1 / 9 ≤ a ij ≤ 9).

[0041] The partial interface for automatic optimization implemented using Python is as Figure 2-3 shown.

[0042] The verification of the optimization effect is shown in Table 3.

[0043] Table 3 Optimization Effect Table

[0044] Number of iterations CR value Learning rate Adjustment action 1 0.1483 0.6 Initial state 5 0.1352 0.6 Normal adjustment 8 0.1297 0.42 First decay 12 0.1135 0.29 Second decay 15 0.1128 0.6 Learning rate reset 18 0.0957 0.6 Reached convergence

[0045] The recommended parameter table is shown in Table 4.

[0046] Table 4 Recommended Parameter Table

[0047]

[0048] Through the dynamic parameter adjustment mechanism, the present invention optimizes the CR value from 0.1483 to 0.0957 within 18 iterations, meeting the requirements of consistency test.

[0049] (4) Having passed the consistency test, the python program automatically outputs the effective weight values of each index, and the final performance evaluation index system is shown in Table 5.

[0050] Table 5 Optimized Project Performance Evaluation Index System Table

[0051]

[0052] Project managers can conduct performance evaluations on projects based on the project performance evaluation index system table shown in Table 5. This index system is scientific and reasonable, and is conducive to improving the efficiency of project resource allocation and utilization benefits.

[0053] The present invention can also be applied to efficiently and scientifically determine the weights of other evaluation indicators in the case where the fixed weights of certain evaluation indicators are given, so as to obtain the required project performance evaluation index system.

[0054] The present invention uses a Python program to achieve the automatic optimization of the analytic hierarchy process (AHP) comparison judgment matrix. This program iteratively adjusts the matrix elements to gradually reduce the consistency ratio (CR) until the consistency requirement is met. Compared with the prior art, the improvements of the present invention include: first, achieving the automatic optimization of the analytic hierarchy process (AHP) comparison judgment matrix until the consistency ratio CR < 0.1 and meeting the consistency requirement; second, constructing a complete adjustment chain of shock detection - attenuation - amplitude limiting - resetting to avoid the limitations of a single adjustment method; third, adopting a method of dynamically adjusting the learning rate to automatically adjust the learning rate value according to the change of the CR value to balance the convergence speed and accuracy. Fourth, adopting shock-sensitive detection, through the sliding window variance method instead of simple difference, can accurately identify the shock mode. Fifth, adopting an exponential decay strategy compared with linear decay to retain a larger adjustment space in the later stage of iteration. Sixth, adopting the learning rate periodic reset to avoid falling into local optimum and improve the optimization success rate of complex matrices.

[0055] It should be noted that in this article, terms such as "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or device.

[0056] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A project performance evaluation method based on a weight optimization method, characterized in that, It includes the following steps: S1: Input the initial project performance evaluation index system; S2: Construct a comparison judgment matrix, calculate the consistency ratio CR. If CR < 0.1, pass the consistency test and proceed to step S4; If CR ≥ 0.1, adjust the comparison judgment matrix and proceed to step S3; S3: Automatically optimize the comparison judgment matrix, recalculate CR, and stop optimizing until CR < 0.

1. Pass the consistency test and proceed to step S4; The automatic optimization includes: ① Oscillation detection: Calculate the variance of the CR value changes in N consecutive iterations; ② Construct a dynamic decay function of the learning rate α: When the variance of the CR value changes in N consecutive iterations is detected to be greater than the threshold T, automatically reduce and adjust the learning rate α according to the exponential decay law; ③ Lower limit protection of the learning rate: During the automatic optimization process, set the minimum learning rate αmin to ensure that the learning rate is not lower than αmin; ④ Periodic reset of the learning rate: When CR < 0.1 is not achieved for K consecutive times, reset the learning rate to the initial learning rate; S4: If the consistency test is passed, output the effective weight values of each index and obtain the final performance evaluation index system.

2. The project performance evaluation method based on the weight optimization method according to claim 1, characterized in that: The oscillation detection adopts the sliding window variance calculation method, the window size N ∈ [5, 15], and the variance threshold T ∈ [0.005, 0.03].

3. The project performance evaluation method based on the weight optimization method according to claim 2, characterized in that: The learning rate decay execution module performs exponential decay according to α ← α × μ, where μ ∈ (0, 1) is the preset decay coefficient.

4. The project performance evaluation method based on the weight optimization method according to claim 3, characterized in that: The decay coefficient μ ∈ (0.5, 0.9).

5. The project performance evaluation method based on the weight optimization method according to claim 4, characterized in that: The minimum learning rate lower limit αmin ∈ [0.05, 0.2].

6. The project performance evaluation method based on the weight optimization method according to claim 5, characterized in that: The reset threshold K of the learning rate periodic reset ∈ [8, 15].

7. The project performance evaluation system based on the weight optimization method is characterized in that: Adopt the project performance evaluation method described in any one of claims 1 - 6. The system includes the following modules: (1) Project performance evaluation index system construction module; (2) Comparison judgment matrix construction module, calculate the eigenvector and the maximum eigenvalue λ of the judgment matrix, and obtain the consistency ratio CR; If the consistency ratio CR < 0.1, pass the consistency test; if CR ≥ 0.1, adjust the judgment matrix; (3) Comparison judgment matrix automatic optimization module, including an oscillation detection module, a learning rate decay execution module, a learning rate lower limit protection module, and a learning rate periodic reset module; (4) Performance evaluation index system output module. After passing the consistency test, output the effective weight values of each index and obtain the final performance evaluation index system.

8. A computer device based on a weight optimization method, characterized in that: The computer device includes: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the project performance evaluation method described in any one of claims 1 - 6.

9. A computer storage medium based on a weight optimization method, on which a computer program is stored, characterized in that, When the program is executed by the processor, it implements the project performance evaluation method described in any one of claims 1 - 6.

Images