Engineering cost control strategy matching method and device oriented to multiple feature attributes, and storage medium

Abstract

The invention provides a multi-feature attribute-oriented engineering cost control strategy matching method and device and a storage medium, and relates to the technical field of engineering cost control. The engineering cost control strategy matching method comprises the following steps: firstly, constructing a case feature library containing standardized historical case feature values and a knowledge base containing production rules; calculating the closeness of historical cases and ideal schemes, and extracting typical feature expressions through micro-cluster and optimal cluster clustering; constructing a case feature cloud model, and calculating the similarity between the feature value of the case to be decided and the typical feature; then constructing a fuzzy Petri net based on a production rule, and inputting similarity reasoning to obtain a strategy matching degree; and finally, reasoning is stopped according to an iteration termination condition, and strategies meeting the condition are screened. According to the method, the multi-feature attribute and the case feature cloud model are integrated, the limitation of dependence on expert experience traditionally is broken, strategy matching standardization and intelligence are realized, and the strategy selection efficiency is improved.

Description

Technical Field

[0001] This invention mainly relates to the field of engineering cost control technology, specifically to a method, device, and storage medium for matching engineering cost control strategies with multiple feature attributes. Background Technology

[0002] Currently, the field of engineering cost control strategy matching relies heavily on expert experience for decision-making, lacking a standardized and scientific analysis and evaluation system. This makes strategy selection susceptible to subjective factors, resulting in low accuracy and efficiency. On the one hand, engineering cost control involves multiple characteristic attributes such as tower structure, foundation engineering, conductor parameters, and environmental conditions. Existing technologies struggle to effectively handle the ambiguity (e.g., characteristics like "high load" and "regular risk" lack clear quantitative boundaries) and randomness (e.g., characteristic parameters of similar projects exhibit objective fluctuations) inherent in these characteristics in actual projects, making it impossible to achieve precise correspondence between characteristics and strategies. On the other hand, traditional methods lack a systematic mechanism for reusing historical cases and rule-based reasoning, making it difficult to transform massive amounts of historical project experience into reusable matching criteria. Consequently, when facing new projects requiring decision-making, strategy matching lacks data support and fails to meet the needs of integrated project cost precision control. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method, device and storage medium for matching engineering cost control strategies with multiple feature attributes, in order to address the shortcomings of the prior art.

[0004] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A method for matching engineering cost control strategies oriented towards multiple feature attributes, comprising the following steps: Construct an engineering cost knowledge base, which includes a case feature library and a rule library. The case feature library includes standardized feature values ​​of multiple historical cases, and the rule library includes production rules. Based on the feature values ​​of each historical case in the case feature library, the closeness to the ideal solution is calculated, and the closeness corresponding to each historical case is obtained. Micro-clustering and optimal clustering are performed sequentially on the proximity and feature values ​​corresponding to all historical cases to obtain the historical cases corresponding to the optimal clustering, and typical feature performance is extracted from the historical cases. Construct a case feature cloud model, import the standardized feature values ​​of the cases to be decided, and calculate the feature similarity between the feature values ​​of the cases to be decided and the typical feature performance based on the case feature cloud model; A fuzzy Petri net is constructed based on the production rules, and the feature similarity is input into the fuzzy Petri net to perform inference from feature similarity to policy matching degree. The reasoning is stopped based on a preset iteration termination condition, and the reasoning results are obtained. From the reasoning results, strategies that meet the preset strategy selection conditions are selected as engineering cost control strategies.

[0005] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a matching device for engineering cost control strategies oriented towards multiple feature attributes, comprising: The knowledge base construction module is used to construct an engineering cost knowledge base, which includes a case feature library and a rule library. The case feature library includes standardized feature values ​​of multiple historical cases, and the rule library includes production rules. The proximity calculation module is used to calculate the proximity to the ideal solution based on the feature values ​​of each historical case in the case feature library, and obtain the proximity corresponding to each historical case. The clustering module is used to perform micro-clustering and optimal clustering on the proximity and feature values ​​corresponding to all historical cases in sequence to obtain the historical cases corresponding to the optimal clustering, and extract typical feature performance from the historical cases. The cloud model building module is used to build a case feature cloud model, import the standardized feature values ​​of the case to be decided, and calculate the feature similarity between the feature values ​​of the case to be decided and the typical feature performance based on the case feature cloud model. The inference module is used to construct a fuzzy Petri net based on the production rules, input the feature similarity into the fuzzy Petri net, and perform inference from feature similarity to policy matching degree. The strategy selection module is used to stop reasoning based on a preset iteration termination condition, obtain the reasoning result, and select the strategy that meets the preset strategy selection condition from the reasoning result as the engineering cost control strategy.

[0006] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a matching device for engineering cost control strategies oriented towards multiple feature attributes, characterized in that it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the engineering cost control strategy matching method oriented towards multiple feature attributes as described above.

[0007] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the engineering cost control strategy matching method oriented towards multiple feature attributes as described above.

[0008] The beneficial effects of this invention are as follows: by constructing a knowledge base, calculating proximity, extracting typical features through clustering, calculating similarity using a cloud model, performing fuzzy Petri net inference, and implementing a screening strategy, the invention integrates multiple feature attributes and case feature cloud models to complete the matching of engineering cost control strategies. This breaks through the limitations of traditional reliance on expert experience, achieving standardization and intelligence in strategy matching. At the same time, through a multi-step, progressive processing logic, it balances the comprehensiveness and relevance of matching, significantly improving the efficiency of strategy selection. Attached Figure Description

[0009] Figure 1 A flowchart illustrating the engineering cost control strategy matching method provided in an embodiment of the present invention; Figure 2 A functional block diagram of the engineering cost control strategy matching device provided in an embodiment of the present invention. Detailed Implementation

[0010] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0011] Example 1: As Figure 1 As shown, this embodiment of the invention provides a method for matching engineering cost control strategies based on multiple feature attributes, including the following steps: S1. Construct an engineering cost knowledge base, which includes a case feature base and a rule base. The case feature base includes standardized feature values ​​of multiple historical cases, and the rule base includes production rules. S2. Calculate the closeness to the ideal solution based on the feature values ​​of each historical case in the case feature library, and obtain the closeness to each historical case. S3. Perform micro-clustering and optimal clustering on the proximity and feature values ​​corresponding to all historical cases in sequence to obtain the historical cases corresponding to the optimal clustering, and extract typical feature performance from the historical cases. S4. Construct a case feature cloud model, import the standardized feature values ​​of the case to be decided, and calculate the feature similarity between the feature values ​​of the case to be decided and the typical feature performance based on the case feature cloud model. S5. Construct a fuzzy Petri net based on the production rules, input the feature similarity into the fuzzy Petri net, and perform inference from feature similarity to policy matching degree. S6. Stop the reasoning based on the preset iteration termination condition, obtain the reasoning result, and select the strategy that meets the preset strategy selection condition from the reasoning result as the engineering cost control strategy.

[0012] In the above embodiments, by constructing a knowledge base, calculating proximity, extracting typical features through clustering, calculating similarity using a cloud model, performing fuzzy Petri net inference, and implementing a screening strategy, the process integrates multiple feature attributes and case feature cloud models to complete the matching of engineering cost control strategies. This breaks through the limitations of traditional reliance on expert experience, achieving standardization and intelligence in strategy matching. At the same time, through a multi-step, progressive processing logic, the comprehensiveness and relevance of the matching are taken into account, significantly improving the efficiency of strategy selection.

[0013] Preferably, an engineering cost knowledge base is constructed, which includes a case feature base and a rule base, comprising: Based on key influencing factors of engineering cost control, historical cases of multiple engineering cost categories were collected. Then, the original attribute values ​​in each historical case were normalized using a normalization formula to obtain standardized feature values. These standardized feature values ​​were used to form a case feature library. The normalization formula is as follows: , in, These are the standardized eigenvalues. For the first i Case number 1 j Primitive values ​​of class attributes , The first j The minimum and maximum values ​​of class attributes are mapped to [0,1]. Based on expert experience and historical case results, the feature and policy matching rules described in natural language are transformed into production rules that include feature conditions, triggering logic of feature conditions and policy conclusions, feature importance, triggering threshold and rule strength. A rule base is formed through production rules.

[0014] Specifically, the rule base construction includes: Based on expert experience and the effectiveness of historical cases, the correspondence between "characteristic performance and control strategy" is analyzed. For example, "high conductor parameters (C≥0.8) and high comprehensive index (Z≥0.4) → high-risk control strategy Q1", and "low conductor parameters (C≤0.2) and low environmental conditions (E≤0.05) → conventional control strategy Q2".

[0015] Transforming natural language rules into production rules (such as "IF C=H∧Z=H THEN strategy=Q1") clarifies parameters such as feature weights and trigger thresholds, providing logical rule support for fuzzy Petri net inference.

[0016] In the above embodiments, when constructing the case feature library, a normalization formula is used to unify the range of historical case attribute data. When constructing the rule library, natural language rules are transformed into production rules containing multiple elements such as feature conditions and triggering logic. This ensures the comparability and consistency of case feature data, avoids analytical bias caused by differences in units, and makes the strategy matching rules more accurate and quantifiable, providing high-quality data and logical support for subsequent proximity calculation and inference matching.

[0017] Preferably, the degree of closeness to the ideal solution is calculated based on the feature values ​​of each historical case in the case feature library, resulting in the degree of closeness for each case, including: Combining the TOPSIS (Top-Side Distance) method, the closeness to the ideal solution is calculated based on the feature values ​​of each historical case in the case feature database. Specifically: The weight of the feature value of each historical case in the case feature database is determined based on the entropy weight method. The expression for the entropy weight method is as follows: , in, For the first j Entropy value of class features , n For the number of historical cases, For characteristic number, Standardized eigenvalues and their corresponding weights Perform a product operation to obtain the weighted standardized values ​​corresponding to each feature value. , The maximum value in the set-weighted standardized value is the positive ideal solution. The minimum value among the weighted standardized values ​​is the negative ideal solution. , Calculate the weighted standardized values ​​and the positive ideal solution. European distance , is represented as: , in, For the positive ideal solution The first in j One value, , Calculate each weighted standardized value and the negative ideal solution. European distance , is represented as: , in, negative ideal solution The first in j One value, ; Based on the proximity formula and the Euclidean distance corresponding to each historical case. and European distance Calculate the fit between historical cases and the ideal solution to obtain the closeness score for each historical case. The closeness score formula is as follows: , in, For the first i The degree of similarity between historical cases and the ideal solution, with a value range of [value range missing]. .

[0018] In the above embodiments, feature weights are determined based on the entropy weight method, and the weighted standardized value, the distance between positive and negative ideal solutions and the closeness are calculated by combining the TOPSIS method. This quantifies the degree of fit between historical cases and ideal solutions. The entropy weight method achieves objective assignment of feature weights, avoiding subjective experience bias. The TOPSIS method accurately quantifies the quality of cases, providing a scientific quantitative basis for subsequent cluster analysis and improving the accuracy of typical feature extraction.

[0019] Preferably, the proximity and feature values ​​corresponding to all historical cases are sequentially subjected to micro-cluster clustering and optimal cluster clustering, including: Calculate the Euclidean distance between the proximity and feature values ​​of each pair of historical cases to obtain the feature distance of multiple cases, and merge historical cases whose feature distance is less than the distance threshold into micro-clusters to obtain multiple micro-clusters; On a micro-cluster basis, calculate the number of cases, feature mean vectors, and Euclidean distance between the two micro-clusters to be merged. Based on the Bayesian information criterion, calculate the merging cost of the two micro-clusters. The expression for the Bayesian information criterion is as follows: , in, , Let A and B be the number of cases corresponding to the two microclusters A and B to be merged. , Let A and B be the feature mean vectors corresponding to the two microclusters A and B to be merged. The Euclidean distance between the mean vectors; When merger costs When the value is reduced to the minimum, the optimal cluster and its corresponding historical cases are obtained.

[0020] Specifically, calculate the mean of features within each cluster: calculate the mean of five types of features for each of the two clusters, and extract typical characteristics: Cluster 1 (22 cases): C mean 0.86, Z mean 0.43, defined as "high load - high risk segment", corresponding to strategy Q1; Cluster 2 (45 cases): C mean 0.18, E mean 0.02, defined as "equilibrium-normal segment", corresponding to strategy Q2. These two typical characteristics serve as the core sample data for constructing the cloud model.

[0021] In the above embodiments, similar cases are first merged using Euclidean distance to generate micro-clusters. Then, the cost of merging micro-clusters is calculated based on the Bayesian information criterion. The optimal clustering is determined and typical features are extracted. The accurate grouping of cases is completed automatically without the need for manual preset of the number of clusters. Representative feature patterns are efficiently extracted, providing core samples for cloud model construction, reducing subsequent computational redundancy, and improving the overall efficiency of the matching process.

[0022] Preferably, a case feature cloud model is constructed, the standardized feature values ​​of the cases to be decided are imported, and the feature similarity between the feature values ​​of the cases to be decided and the typical feature manifestations is calculated based on the case feature cloud model, including: Based on the reverse cloud algorithm, the case feature cloud model is defined by three core parameters, which include the expectation Ex, which characterizes the trend of feature center, the entropy En, which characterizes the degree of ambiguity, and the hyperentropy He, which characterizes the fluctuation of ambiguity. Based on the three core parameters of the case feature cloud model, and combined with the membership degree formula, the degree of fuzziness of the feature values ​​of the case to be decided belonging to the typical feature pattern is calculated. , is represented as: , , in, The random entropy that follows a normal distribution; Import the standardized feature values ​​of the cases to be decided, based on the degree of fuzziness. The feature similarity between the feature values ​​of the case to be decided and the typical feature performance is calculated using the cloud similarity formula, which is: , in, For feature similarity.

[0023] In the above embodiments, a cloud model containing three parameters—expectation, entropy, and hyperentropy—is constructed based on the reverse cloud algorithm. The degree of feature fuzziness is calculated using the membership formula, and the similarity between the case to be decided and the typical features is obtained using the cloud similarity formula. This effectively handles the fuzziness and randomness of engineering cost features, realizes flexible quantification of feature matching, avoids the error of "black and white" judgment, provides accurate similarity input for subsequent reasoning, and improves the accuracy of strategy matching.

[0024] Preferably, a fuzzy Petri net is constructed based on the production rules, and the feature similarity is input into the fuzzy Petri net to perform inference from feature similarity to policy matching degree, including: The core elements of inference for fuzzy Petri nets are constructed based on the feature conditions, triggering logic of feature conditions and policy conclusions, feature importance, triggering threshold, and rule strength in production rules. These core elements include input places, transitions, weights, thresholds, rule strength, intermediate places, and output places. For each transition, a transition trigger judgment is performed on the feature similarity received by the input library based on the transition triggering rule, including: a. Calculate the overall contribution of the feature similarity, expressed as: , in, For change The overall contribution of inputs, For input library Change The weights, For input library Credibility, like , λ For change The corresponding trigger threshold then changes Triggered if it is triggered, otherwise not triggered. b. For triggerable transitions, the credibility of the output locus is calculated based on the comprehensive contribution of the feature similarity and the rule strength, expressed as: , in, The updated credibility of the output repository or intermediate repository. To maintain credibility before the update, For change The corresponding rule strength; If there are multi-level inference, the credibility of the intermediate place will be used as the input for the next level of transition, repeating the process from a to b until the credibility propagation that can trigger the transition is completed. The credibility of the output place is then used as the policy matching degree.

[0025] In the above embodiments, a fuzzy Petri net containing core elements such as input locations and transitions is constructed based on production rules. The transition trigger is determined by calculating the comprehensive contribution, and the policy matching degree is obtained by propagating the credibility. The feature similarity and policy matching logic are deeply integrated to achieve accurate reasoning from fuzzy feature quantification values ​​to policy matching degree. It also supports multi-level reasoning to adapt to complex rules, further improving the logic and reliability of policy matching.

[0026] Preferably, the reasoning is stopped based on a preset iteration termination condition to obtain the reasoning result, and strategies that meet preset strategy selection conditions are selected from the reasoning result as engineering cost control strategies, including: When the changes in the credibility of all output locations obtained from two consecutive inferences satisfy the following conditions: At that time, stop reasoning. The preset change threshold is set to 0.001. From the strategy matching scores corresponding to the output library, select the strategy with the highest credibility that is greater than the preset strategy triggering standard (with a value of 0.6) as the final project cost control strategy.

[0027] In the above embodiments, the iteration termination condition is that the change in the credibility of the output library is less than a threshold for two consecutive times. The strategy with the highest credibility and greater than the triggering standard is selected as the final result to ensure that the reasoning process converges and the result is stable, avoiding excessive iteration or insufficient reasoning. At the same time, the optimal strategy is accurately locked through the dual screening rules to ensure the effectiveness and practicality of the output result, and finally realize the intelligent and accurate matching of engineering cost control strategies.

[0028] The implementation process of the engineering cost control strategy matching method of the present invention will be introduced through a specific example below.

[0029] The case feature attributes in step S1.1 include tower structure, foundation engineering, conductor parameters, environmental conditions, and comprehensive index. Step S1.2 uses the TOPSIS (Top-Solution Distance) method to calculate the feature attribute parameters of 67 cases, as shown in Table 1. Table 1 shows the TOPSIS evaluation calculation results.

[0030] Table 1 Step S1.3: Use a two-step clustering algorithm (i.e., micro-cluster clustering and optimal cluster clustering) to extract two types of feature representations, namely "high-load-high-risk segment" and "balanced-normal segment", as shown in Table 2. Table 2 shows the feature representations of typical cases and their strategy matching logic.

[0031] Table 2 Step S1.4: Construct a case feature cloud model based on the characteristic features of typical cases, as shown in Table 3, and draw a two-dimensional scatter plot. Table 3 shows the feature cloud model.

[0032] Table 3 Step S1.5: Design the FPN knowledge reasoning network, using feature similarity as input.

[0033] Step S1.6: Through validation set inference, the matching accuracy reached 100%, as shown in Table 4, realizing intelligent matching of engineering cost control strategies. Table 4 shows the validation set inference results.

[0034] Table 4 like Figure 2 As shown, this embodiment of the invention also provides a matching device for engineering cost control strategies oriented towards multiple feature attributes, including: The knowledge base construction module is used to construct an engineering cost knowledge base, which includes a case feature library and a rule library. The case feature library includes standardized feature values ​​of multiple historical cases, and the rule library includes production rules. The proximity calculation module is used to calculate the proximity to the ideal solution based on the feature values ​​of each historical case in the case feature library, and obtain the proximity corresponding to each historical case. The clustering module is used to perform micro-clustering and optimal clustering on the proximity and feature values ​​corresponding to all historical cases in sequence to obtain the historical cases corresponding to the optimal clustering, and extract typical feature performance from the historical cases. The cloud model building module is used to build a case feature cloud model, import the standardized feature values ​​of the case to be decided, and calculate the feature similarity between the feature values ​​of the case to be decided and the typical feature performance based on the case feature cloud model. The inference module is used to construct a fuzzy Petri net based on the production rules, input the feature similarity into the fuzzy Petri net, and perform inference from feature similarity to policy matching degree. The strategy selection module is used to stop reasoning based on a preset iteration termination condition, obtain the reasoning result, and select the strategy that meets the preset strategy selection condition from the reasoning result as the engineering cost control strategy.

[0035] This invention also provides a multi-feature attribute-oriented engineering cost control strategy matching device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multi-feature attribute-oriented engineering cost control strategy matching method as described above.

[0036] This invention also provides a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the engineering cost control strategy matching method oriented towards multiple feature attributes as described above.

[0037] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0038] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0039] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0040] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0041] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for matching engineering cost control strategies based on multiple feature attributes, characterized in that, Includes the following steps: Construct an engineering cost knowledge base, which includes a case feature library and a rule library. The case feature library includes standardized feature values ​​of multiple historical cases, and the rule library includes production rules. Based on the feature values ​​of each historical case in the case feature library, the closeness to the ideal solution is calculated, and the closeness corresponding to each historical case is obtained. Micro-clustering and optimal clustering are performed sequentially on the proximity and feature values ​​corresponding to all historical cases to obtain the historical cases corresponding to the optimal clustering, and typical feature performance is extracted from the historical cases. Construct a case feature cloud model, import the standardized feature values ​​of the cases to be decided, and calculate the feature similarity between the feature values ​​of the cases to be decided and the typical feature performance based on the case feature cloud model; A fuzzy Petri net is constructed based on the production rules, and the feature similarity is input into the fuzzy Petri net to perform inference from feature similarity to policy matching degree. The reasoning is stopped based on a preset iteration termination condition, and the reasoning results are obtained. From the reasoning results, strategies that meet the preset strategy selection conditions are selected as engineering cost control strategies.

2. The engineering cost control strategy matching method according to claim 1, characterized in that, Construct an engineering cost knowledge base, which includes a case feature base and a rule base, including: Based on key influencing factors of engineering cost control, historical cases of multiple engineering cost categories were collected. Then, the original attribute values ​​in each historical case were normalized using a normalization formula to obtain standardized feature values. These standardized feature values ​​were used to form a case feature library. The normalization formula is as follows: ， in, These are the standardized eigenvalues. For the first i Case number 1 j Primitive values ​​of class attributes , The first j The minimum and maximum values ​​of class attributes are mapped to [0,1]. Based on expert experience and historical case results, the feature and policy matching rules described in natural language are transformed into production rules that include feature conditions, triggering logic of feature conditions and policy conclusions, feature importance, triggering threshold and rule strength. A rule base is formed through production rules.

3. The engineering cost control strategy matching method according to claim 2, characterized in that, Based on the feature values ​​of each historical case in the case feature library, the closeness to the ideal solution is calculated, resulting in the closeness score for each case, including: The weight of the feature value of each historical case in the case feature database is determined based on the entropy weight method. The expression for the entropy weight method is as follows: ， in, For the first j Entropy value of class features , n For the number of historical cases, For characteristic number, Standardized eigenvalues and their corresponding weights Perform a product operation to obtain the weighted standardized values ​​corresponding to each feature value. ， The maximum value in the set-weighted standardized value is the positive ideal solution. The minimum value among the weighted standardized values ​​is the negative ideal solution. , Calculate the weighted standardized values ​​and the positive ideal solution. European distance , represented as: ， in, For the positive ideal solution The first in j One value, , Calculate each weighted standardized value and the negative ideal solution European distance , represented as: ， in, negative ideal solution The first in j One value, ; Based on the proximity formula and the Euclidean distance corresponding to each historical case. and European distance Calculate the fit between historical cases and the ideal solution to obtain the closeness score for each historical case. The closeness score formula is as follows: ， in, For the first i The degree of similarity between historical cases and the ideal solution, with a value range of [value range missing]. .

4. The engineering cost control strategy matching method according to claim 2, characterized in that, For all historical cases, the proximity and feature values ​​are sequentially processed using micro-cluster clustering and optimal cluster clustering, including: Calculate the Euclidean distance between the proximity and feature values ​​of each pair of historical cases to obtain the feature distance of multiple cases, and merge historical cases whose feature distance is less than the distance threshold into micro-clusters to obtain multiple micro-clusters; On a micro-cluster basis, calculate the number of cases, feature mean vectors, and Euclidean distance between the two micro-clusters to be merged. Based on the Bayesian information criterion, calculate the merging cost of the two micro-clusters. The expression for the Bayesian information criterion is as follows: ， in, , Let A and B be the number of cases corresponding to the two microclusters A and B to be merged. , Let A and B be the feature mean vectors corresponding to the two microclusters A and B to be merged. The Euclidean distance between the mean vectors; When merger costs When the value is reduced to the minimum, the optimal cluster and its corresponding historical cases are obtained.

5. The engineering cost control strategy matching method according to claim 2, characterized in that, Construct a case feature cloud model, import the standardized feature values ​​of the cases to be decided, and calculate the feature similarity between the feature values ​​of the cases to be decided and the typical feature manifestations based on the case feature cloud model, including: Based on the inverse cloud algorithm, three core parameters are defined for the case feature cloud model. These three core parameters include the expectation used to characterize the trend of feature centers. Ex Entropy, used to characterize the degree of fuzziness En and superentropy used to characterize fuzzy fluctuations He ; Based on the three core parameters of the case feature cloud model, and combined with the membership degree formula, the degree of fuzziness of the feature values ​​of the case to be decided belonging to the typical feature pattern is calculated. , represented as: ， ， in, The random entropy that follows a normal distribution; Import the standardized feature values ​​of the cases to be decided, based on the degree of fuzziness. The feature similarity between the feature values ​​of the case to be decided and the typical feature performance is calculated using the cloud similarity formula, which is: ， in, For feature similarity.

6. The method for matching engineering cost control strategies according to claim 2, characterized in that, A fuzzy Petri net is constructed based on the aforementioned production rules. The feature similarity is input into the fuzzy Petri net, and inference from feature similarity to policy matching degree is performed, including: The core elements of inference for fuzzy Petri nets are constructed based on the feature conditions, triggering logic of feature conditions and policy conclusions, feature importance, triggering threshold, and rule strength in production rules. These core elements include input places, transitions, weights, thresholds, rule strength, intermediate places, and output places. For each transition, a transition trigger judgment is performed on the feature similarity received by the input library based on the transition triggering rule, including: a. Calculate the overall contribution of the feature similarity, expressed as: ， in, For change The overall contribution of inputs, For input library Changes The weights, For input library Credibility, like , λ For change The corresponding trigger threshold then changes Triggered if it is triggered, otherwise not triggered. b. For triggerable transitions, the credibility of the output library is calculated based on the comprehensive contribution of the feature similarity and the rule strength, expressed as: ， in, The updated credibility of the output repository or intermediate repository. To maintain credibility before the update, For change The corresponding rule strength; If there are multi-level inference, the credibility of the intermediate place will be used as the input for the next level of transition, repeating the process from a to b until the credibility propagation that can trigger the transition is completed. The credibility of the output place is then used as the policy matching degree.

7. The engineering cost control strategy matching method according to claim 6, characterized in that, The reasoning process is stopped based on a preset iteration termination condition, and the reasoning results are obtained. From these results, strategies that meet preset strategy selection criteria are selected as engineering cost control strategies, including: When the changes in the credibility of all output libraries obtained from two consecutive inferences satisfy the following conditions At that time, stop reasoning. The preset change threshold; From the strategy matching scores corresponding to the output library, select the strategy with the highest credibility that is greater than the preset strategy triggering standard, and use it as the final project cost control strategy.

8. A matching device for engineering cost control strategies oriented towards multiple feature attributes, characterized in that, include: The knowledge base construction module is used to construct an engineering cost knowledge base, which includes a case feature library and a rule library. The case feature library includes standardized feature values ​​of multiple historical cases, and the rule library includes production rules. The proximity calculation module is used to calculate the proximity to the ideal solution based on the feature values ​​of each historical case in the case feature library, and obtain the proximity corresponding to each historical case. The clustering module is used to perform micro-clustering and optimal clustering on the proximity and feature values ​​corresponding to all historical cases in sequence to obtain the historical cases corresponding to the optimal clustering, and extract typical feature performance from the historical cases. The cloud model building module is used to build a case feature cloud model, import the standardized feature values ​​of the case to be decided, and calculate the feature similarity between the feature values ​​of the case to be decided and the typical feature performance based on the case feature cloud model. The inference module is used to construct a fuzzy Petri net based on the production rules, input the feature similarity into the fuzzy Petri net, and perform inference from feature similarity to policy matching degree. The strategy selection module is used to stop reasoning based on a preset iteration termination condition, obtain the reasoning result, and select the strategy that meets the preset strategy selection condition from the reasoning result as the engineering cost control strategy.

9. A matching device for engineering cost control strategies oriented towards multiple feature attributes, characterized in that, The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the engineering cost control strategy matching method for multiple feature attributes as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the engineering cost control strategy matching method oriented towards multiple feature attributes as described in any one of claims 1 to 7.

Images