An intelligent auditing method based on settlement auditing operation standards

Abstract

The present application belongs to the field of engineering management and computer intelligent auditing technology, and discloses an intelligent auditing method based on settlement auditing operation standard. The method acquires engineering settlement data, extracts auditing behavior characteristics, and constructs an auditing stage characteristic vector. Based on the characteristic vector, the stage confidence probability of the engineering settlement in different auditing stages is calculated to form the auditing stage probability distribution. The settlement auditing operation standard is modeled as a calculable constraint function, the constraint response value is calculated, and the stage modulation of the constraint strength is performed according to the stage confidence probability. The auditing conclusion is generated in parallel under the assumption of different auditing stages, and the stage consistency counterargument index is obtained through consistency analysis. When the counterargument index meets the preset inconsistency condition, the stage modulation amplitude is suppressed, and the stabilized auditing conclusion is generated based on the suppressed constraint strength as the final result output, thereby improving the stability and reliability of the engineering settlement auditing result under the uncertainty of the auditing stage.

Description

Technical Field

[0001] This invention relates to the fields of engineering management and computer intelligent auditing technology, specifically to an intelligent auditing method based on settlement auditing operation standards. Background Technology

[0002] Project settlement auditing is a crucial technical step in project management. It typically involves verifying and analyzing settlement data to determine the reasonableness and compliance of the settlement results. With the increasing scale of engineering projects and the growing complexity of settlement data structures, computer-based settlement auditing methods are increasingly being applied to the project settlement auditing process to improve efficiency and consistency.

[0003] In actual engineering settlement audit processes, settlement audits typically include multiple audit stages such as preliminary review, secondary review, and final review. Although the settlement audit operation standards used in each audit stage are usually consistent in terms of textual content, in practice, there are differences in the tolerance for anomalies, the strength of constraints on deviations, and the certainty requirements for audit conclusions at different audit stages.

[0004] In existing technologies, computer systems often execute settlement audit operation standards as static rules, or simply switch rule versions by using stage identifiers. This makes it difficult to reflect the implicit impact of the audit stage on the constraint strength of the operation standards, and can easily lead to judgment results that are consistent in form but inconsistent in audit semantics for the same settlement data under different audit stages.

[0005] Furthermore, in some project settlement data, the audit stage information is not recorded with explicit fields, leading to uncertainty in the computer's identification of the audit stage. In this situation, directly applying audit rules based on a single stage assumption can easily result in misjudgments due to stage identification bias, thereby affecting the stability and reliability of the settlement audit results.

[0006] Therefore, it is necessary to provide an intelligent settlement and auditing method that can differentiate the strength of operational standard constraints at different auditing stages at the computer execution level and reduce misjudgments caused by uncertainty in the auditing stages. Summary of the Invention

[0007] In view of the above-mentioned shortcomings mentioned in the background art, the purpose of this invention is to provide an intelligent auditing method and a computer storage medium based on settlement auditing operation standards.

[0008] A first aspect of the present invention provides an intelligent auditing method based on settlement auditing operation standards, the method comprising the following steps: Step S1: Obtain the project settlement data to be reviewed, and based on the review rule triggering, settlement data adjustment and review process execution trajectory recorded during the settlement review process, extract the review behavior features to characterize the intensity of review execution, and construct the review stage feature vector; Step S2: Calculate the stage confidence probability of the project settlement being in at least two different review stages based on the feature vector of the review stage, and form a review stage probability distribution; Step S3: Model the settlement audit operation standard as a computable constraint function, calculate the constraint response value for the project settlement data, and perform stage modulation on the constraint strength of the constraint response value based on the audit stage probability distribution; Step S4: Perform audit calculations in parallel under different audit stage assumptions to generate corresponding audit conclusions, and perform consistency analysis on the audit conclusions to obtain stage consistency counter-evidence indicators. Step S5: When the stage consistency counter-evidence index meets the preset inconsistency condition, the stage modulation amplitude of the constraint strength is suppressed, and a stable audit conclusion is generated based on the suppressed constraint strength as the final audit result output.

[0009] As an example, the audit behavior characteristics include at least one of the following: the number of audit rules triggered by unit settlement data, the proportion of adjustment amount of settlement data during the audit process, the proportion of settlement items marked as abnormal, and the number of audit nodes actually executed in the audit process.

[0010] As an example, the feature vector of the review stage is obtained by numerical processing of the review behavior features and combining them according to preset weights; wherein, the preset weights are used to reflect the degree of influence of different review behavior features on the review stage determination.

[0011] As an example, the stage confidence probability is obtained by multi-stage classification calculation of the audit stage feature vector, and the stage confidence probability is used to characterize the probability that the project settlement is in different audit stages.

[0012] As an example, the constraint response value of the constraint function is a continuous numerical value, used to characterize the degree of deviation of the engineering settlement data from the settlement audit operation standard, and is not directly used as the result of determining whether there is a violation.

[0013] As an example, the stage modulation is achieved by introducing a stage modulation coefficient on the constraint response value based on the stage confidence probability, so that the same settlement audit operation standard has different constraint strengths at different audit stages.

[0014] As an example, the phase consistency counter-evidence index is calculated based on the differences in audit conclusions under different audit phase assumptions, and is used to characterize the stability of audit conclusions as audit phases change.

[0015] As an example, when the phase consistency counter-evidence index meets the preset inconsistency conditions, the dominant role of a single audit phase hypothesis in the audit conclusion can be suppressed by reducing the change amplitude of the phase modulation coefficient.

[0016] As an example, the stabilization audit conclusion may include the audit conclusion obtained by re-performing the audit calculation after suppression processing, or it may include risk-based audit information that reflects the sensitivity of the audit conclusion to the audit stage.

[0017] A second aspect of the present invention also provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the preceding claims.

[0018] Compared with the prior art, the present invention has the following significant advantages: This invention quantifies and models the audit behavior generated during the engineering settlement audit process, modulates the constraint strength of the settlement audit operation standards in stages by combining the confidence probability of the audit stage, and generates audit conclusions in parallel under the assumption of multiple audit stages. By using consistency rebuttal and modulation suppression mechanisms, it reduces the impact of uncertainty in a single audit stage on the audit results. Thus, even when it is difficult to accurately determine or there are overlaps in the audit stages, it can still output stable, reliable, and stage-sensitive intelligent audit results, thereby improving the robustness, interpretability, and engineering applicability of engineering settlement audits. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall process of an intelligent auditing method based on settlement auditing operation standards disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the multi-stage classification model disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the layout of computer storage media in a computer according to an embodiment of the present invention. Detailed Implementation

[0020] The following detailed description, in conjunction with specific embodiments, illustrates an intelligent auditing method based on settlement auditing operation standards proposed in this application. It should be understood that the following embodiments are merely illustrative of the technical solutions of this invention and do not constitute a limitation on the scope of protection of this invention. Various equivalent substitutions or modifications made to these embodiments by those skilled in the art without departing from the technical concept of this invention should fall within the scope of protection of this invention.

[0021] The method described in this embodiment is executed by a computer, which can be a server, audit terminal, or cloud computing node in an engineering settlement audit system. The computer includes a processor, a memory, and a computer program stored in the memory and executable by the processor. When the computer program runs, it causes the computer to perform the steps of the method described in this invention.

[0022] Please see Figure 1 This invention provides an intelligent auditing method based on settlement auditing operation standards, the method comprising the following steps: Step S1: Obtain the project settlement data to be reviewed, and based on the review rule triggering, settlement data adjustment and review process execution trajectory recorded during the settlement review process, extract the review behavior features to characterize the intensity of review execution, and construct the review stage feature vector; In this step, the computer first acquires the project settlement data to be reviewed. This project settlement data can originate from multiple system modules, such as a settlement application system, contract management system, cost management system, or historical review database. In actual project management scenarios, project settlement data often has the following characteristics: large data volume, containing numerous list items; complex data structure, including both numerical fields and hierarchical relationship fields; and multiple version iterations during the review process. Therefore, in this embodiment, the computer preferably performs structured parsing on the acquired project settlement data, uniformly mapping data from different sources and in different formats to an internal settlement data model for subsequent unified processing.

[0023] It should be noted that the engineering settlement data in this invention does not refer only to the final settlement amount, but rather to the entire set of data that runs through the entire review process and can be accessed and modified multiple times by the rule engine. For example, the same list item may only undergo format and logic checks in the initial review stage, further rationality analysis in the review stage, and strict compliance checks in the final review stage.

[0024] While acquiring project settlement data, the computer further acquires audit process data generated during the settlement review process. This audit process data is automatically generated and recorded by the audit system during operation, without relying on manual input, and is objective and traceable. The audit process data includes at least: audit rule trigger records, settlement data adjustment records, and audit process execution trajectory records. Audit rule trigger records are typically generated by the rule engine module; for example, whenever a piece of settlement data meets the rule trigger condition, the system records information such as the rule identifier, trigger time, and associated data entries. Settlement data adjustment records describe the data modifications made during the audit process, reflecting the degree of intervention by the audit system in the settlement data. Audit process execution trajectory records describe the node paths traversed during the actual execution of the audit process, which may differ from the theoretical process.

[0025] Based on the above audit process data, the computer extracts audit behavior features to characterize the intensity of audit execution.

[0026] As an example, the audit behavior characteristics include at least one of the following: the number of audit rules triggered by unit settlement data, the proportion of adjustment amount of settlement data during the audit process, the proportion of settlement items marked as abnormal, and the number of audit nodes actually executed in the audit process.

[0027] In this embodiment, the aforementioned audit behavior characteristics are all automatically statistically obtained by the computer during the audit process, and are used to quantitatively describe the objective intensity of audit execution. For example, the number of audit rules triggered by unit settlement data reflects the density of audit rule calls; the proportion of settlement data adjustment amounts during the audit process reflects the extent of intervention in settlement data; the proportion of settlement items marked as abnormal reflects the strength of abnormal identification; and the number of audit nodes actually executed in the audit process reflects the complexity of the audit process.

[0028] In practical implementation, the above-mentioned characteristics of review behavior can be represented as follows:

[0029] in, Indicates the number of times the review rule is triggered. Indicates the number of settlement items. This represents the absolute value of the cumulative adjustment amount during the settlement review process. This indicates the original declared amount. This indicates the number of settlement entries marked as abnormal. This indicates the actual number of audit nodes executed.

[0030] After obtaining the characteristics of each audit behavior, the computer performs numerical processing on each audit behavior characteristic and combines them to construct the feature vector of the audit stage.

[0031] As an example, the feature vector of the review stage is obtained by numerical processing of the review behavior features and combining them according to preset weights; wherein, the preset weights are used to reflect the degree of influence of different review behavior features on the review stage determination.

[0032] In this embodiment, the feature vector of the review stage can be represented as: .in, Preset weights are used to reflect the relative importance of different review behavior characteristics in the review stage judgment. Through this vectorized expression, the computer can comprehensively analyze different types of review behaviors using a unified data structure.

[0033] Step S2: Calculate the stage confidence probability of the project settlement being in at least two different review stages based on the feature vector of the review stage, and form a review stage probability distribution; In this step, the computer performs probability modeling on the probability that the project settlement is in different review stages based on the review stage feature vector constructed in step S1.

[0034] As an example, the stage confidence probability is obtained by multi-stage classification calculation of the audit stage feature vector, and the stage confidence probability is used to characterize the probability that the project settlement is in different audit stages.

[0035] This process includes, but is not limited to, preliminary review, secondary review, and final review stages. It should be noted that in actual project settlement review, settlement data may not strictly correspond to a specific review stage, especially when data flows across stages or stage information is not clearly recorded. Therefore, this embodiment does not employ a single-stage determination method, but rather uses stage confidence probabilities to quantify the probability that the project settlement is in different review stages.

[0036] Specifically, the computer will use the feature vectors from the review stage. The data is input into a multi-stage classification model, which can be trained based on historical review sample data. The output of the multi-stage classification model is the stage confidence probability corresponding to multiple review stages. .in, Let represent the confidence probability that the project settlement is in the k-th review stage, and satisfy: .

[0037] It is understandable that the stage confidence probability is used to characterize the likelihood that the project settlement is in different review stages, rather than to determine a unique review stage.

[0038] For example: see Figure 2 A multi-stage classification model used to calculate the confidence probability during the review stage includes at least the following: The feature encoding module is used to process the input feature vectors from the review stage. A nonlinear feature transformation is performed to extract a high-level semantic representation. The feature encoding module can be composed of a multi-layer feedforward neural network, for example, consisting of at least two fully connected layers and a nonlinear activation function. Its calculation process can be expressed as follows: .in: For the feature vector of the review stage; Features of the intermediate representation after encoding; This is a non-linear activation function. This module maps the original review behavior features to a representation space more suitable for stage-based judgment.

[0039] The stage prototype matching module is used to match the intermediate representation features with the stage prototypes corresponding to each review stage, and to calculate the correlation between the input features and different review stages. For each review stage, a stage prototype vector is pre-set or learned through training. This is used to characterize the typical distribution features of review behavior characteristics at this review stage. The stage prototype matching module calculates the intermediate representation features. The similarity between the input sample and the prototype vectors of each stage indicates the degree of correlation between the input sample and different review stages. For example: , in, This indicates the degree of relevance between the current project settlement and the k-th audit stage.

[0040] The stage gating module is used to gate and filter different review stages based on the relevance level, suppressing review stages with low relevance to the current review status. The stage gating module first normalizes the matching strength to obtain the gating weight: ,in, This refers to the gate temperature parameter.

[0041] Subsequently, the stage gating module sparsifies the gating weights, for example, retaining only a few review stages with high matching degrees and renormalizing the remaining weights, thereby suppressing review stages with low relevance to the current review status. The stage gating module avoids unnecessary interference introduced by all review stages participating in subsequent calculations simultaneously.

[0042] The evidence evaluation module assesses the strength of supporting evidence for the current review behavior characteristics at each review stage. This module can be constructed using linear mapping networks tailored to each review stage, and its output can be represented as: .in, This indicates the strength of evidence supporting the current project settlement being in the kth review stage.

[0043] The refutation assessment module evaluates the strength of refutation evidence against current review behavior characteristics at each review stage. Similar to the evidence assessment module, the refutation assessment module can also be constructed using a linear mapping network, and its output is: .in, This indicates the strength of the disproving evidence that the current project settlement is in the k-th review stage. By introducing a disproving evidence evaluation module, the model can avoid judging the stage solely based on "supporting evidence," thereby improving the robustness of stage determination.

[0044] The stage fusion and probability calculation module is used to fuse the evidence evaluation results. Evaluation results of counter-evidence It also outputs the stage confidence probability of the project settlement at each review stage.

[0045] The stage scores for each review phase can be expressed as follows: .in, This represents the stage gating weights after sparsification.

[0046] Obtain the stage scores corresponding to each review stage. Subsequently, the multi-stage classification model does not directly output the review stage corresponding to the maximum score, but further calculates the stage confidence probability corresponding to each review stage.

[0047] In this embodiment, the stage confidence probability can be obtained by normalizing the stage score, for example:

[0048] in, This is a temperature parameter used to adjust the smoothness of the probability distribution.

[0049] Through the above calculations, we can then obtain the stage confidence probability distribution for project settlement at different review stages. .

[0050] For example, in a project settlement audit instance, the stage score output by the multi-stage classification model is: , , After temperature adjustment and normalization, the stage confidence probability can be obtained as follows: , , The above results indicate that the project settlement is most likely in the second review stage (e.g., the verification stage), while still retaining the possibility of it being in the first or third review stage. In subsequent steps, the confidence probability distribution of this stage will be used to modulate the constraint strength of the settlement review operation standards and support parallel review calculations under the multi-stage review assumption.

[0051] By using the above probability distribution, the computer retains multiple review stage assumptions simultaneously, providing a basis for subsequent review calculations under different review stage assumptions, thereby avoiding review semantic deviations caused by stage misjudgments.

[0052] Step S3: Model the settlement audit operation standard as a computable constraint function, calculate the constraint response value for the project settlement data, and perform stage modulation on the constraint strength of the constraint response value based on the audit stage probability distribution; In this step, the computer models the settlement review operation standards as computable constraint functions. These standards may include engineering quantity calculation specifications, pricing rules, contractual restrictions, and policy-based fee requirements. In traditional review methods, these standards typically exist in textual or institutional form, making them difficult to directly participate in calculations. Therefore, this embodiment transforms the settlement review operation standards into constraint functions, enabling them to generate continuous numerical constraint responses to engineering settlement data.

[0053] Specifically, for the i-th settlement audit operation standard, the following constraint function is constructed: , in, This indicates the input of project settlement data; This indicates the applicable conditions for the work standard, used to limit the scope of the settlement audit work standard in the audit calculation under the specific engineering settlement data status or audit context. The applicable conditions for the work standard are not abstract logical descriptions, but rather a predefined set of determinate conditions used to determine whether to activate the corresponding settlement audit work standard for the current engineering settlement data based on the engineering settlement data or the audit process status.

[0054] In some embodiments, the applicable conditions of the work standards The conditions may include at least one of the following: project type conditions, settlement item type conditions, contractual conditions, settlement stage or audit stage conditions, and numerical conditions related to the range of project settlement data values. The computer will only perform constraint response calculations for the corresponding settlement audit operation standard when the project settlement data meets the applicable conditions of the operation standard; otherwise, the settlement audit operation standard will be skipped to avoid irrelevant or inapplicable rules interfering with the audit results.

[0055] This indicates that the applicable conditions of the aforementioned work standards are met. Under the premise of [specific conditions], the constraint response function calculated based on the project settlement data is used to output a continuous value reflecting the degree of deviation of the project settlement data from the i-th settlement review operation standard. It is understood that the constraint response function itself is used to characterize the judgment logic of the operation standard, and does not directly reflect the difference in the strictness of the execution of the operation standard at different review stages.

[0056] This represents a constraint strength modulation interface, used to convert the constraint response function... The output results are related to the modulation parameters associated with the audit stage (a set of computable parameters used to characterize the differences in audit rigor, tolerance, or constraint enforcement strength at different audit stages; these parameters are used to apply stage-specific effects to the constraint response results while keeping the settlement audit operation standards unchanged, such as those mentioned above). This allows for the association of different audit stages, thereby imposing different levels of constraint on the same settlement audit operation standard.

[0057] Specifically, the constraint strength modulation interface A predefined computable modulation structure is used to receive stage modulation parameters related to the review stage, and to adjust the intensity of the output of the constraint response function based on the stage modulation parameters. The intensity adjustment can be expressed as scaling, weighting, or function mapping of the constraint response value, so that the same constraint response function outputs constraint response results of different magnitudes at different review stages.

[0058] It should be noted that the applicable conditions of the work standards are... Interface with constraint strength modulation The decoupling setting allows for separate control over the scope and intensity of settlement audit operation standards without altering their logical structure. This enables flexible application of settlement audit operation standards under multi-stage audit assumptions and provides a foundation for subsequent stage modulation, consistency verification, and stabilization audit calculations.

[0059] As an example, the constraint response value of the constraint function is a continuous numerical value, used to characterize the degree of deviation of the engineering settlement data from the settlement audit operation standard, and is not directly used as the result of determining whether there is a violation.

[0060] In the specific calculation process, the constraint response function Output continuous constraint response values Among them, the constraint response value It is a continuous numerical value used to characterize the degree of deviation of engineering settlement data from the settlement audit operation standards, and is not directly used as the result of determining whether there is a violation.

[0061] As an example, the stage modulation is achieved by introducing a stage modulation coefficient on the constraint response value based on the stage confidence probability, so that the same settlement audit operation standard has different constraint strengths at different audit stages.

[0062] In this embodiment, after obtaining the constraint response value Then, the computer uses the probability distribution of the review stage obtained in step S2 as a basis. The constraint strength of the constraint response value is modulated in stages. Specifically, the computer introduces different stage modulation coefficients for different review stages: And form the staged modulated constraint response: .

[0063] Through the above methods, the same settlement audit operation standard exhibits different levels of constraint under different audit stage assumptions, thereby reflecting the differences in the tightness of audit at different audit stages without changing the operation standard itself.

[0064] Step S4: Perform audit calculations in parallel under different audit stage assumptions to generate corresponding audit conclusions, and perform consistency analysis on the audit conclusions to obtain stage consistency counter-evidence indicators. Based on the set of stage-modulated constraint response values ​​obtained in step S3 for different audit stages, the computer constructs parallel audit calculation paths under multiple audit stage assumptions.

[0065] Specifically, for each review stage, the assumptions are as follows: The computer is based on the set of constraint response values ​​corresponding to the assumptions of this stage. A complete audit calculation process is executed to generate the audit conclusions under the assumptions of this audit stage.

[0066] In this embodiment, the audit conclusion can be generated in the following form: Among them, the function This refers to the audit calculation function used to comprehensively map multiple constraint response values ​​to an audit conclusion. The audit calculation function can be implemented using methods such as weighted summation, piecewise mapping, or grade determination; this invention does not impose any limitations on this method.

[0067] For example, a computer can first calculate a comprehensive review score: ,in, Indicates the first The weight of settlement audit operation standards in the comprehensive audit.

[0068] Subsequently, an audit conclusion is generated based on the range of the comprehensive audit score: , In this system, 0 indicates approval, 1 indicates attention is needed, and 2 indicates adjustment is needed.

[0069] Using the above method, the computer can obtain a set of parallel review conclusions under different review stage assumptions. .

[0070] After obtaining the above audit conclusions, the computer does not directly select any one of the conclusions as the final result. Instead, it further conducts consistency rebuttal analysis on the audit conclusions under different audit stage assumptions to assess the sensitivity of the audit conclusions to changes in the audit stage assumptions.

[0071] As an example, the phase consistency counter-evidence index is calculated based on the differences in audit conclusions under different audit phase assumptions, and is used to characterize the stability of audit conclusions as audit phases change.

[0072] In this embodiment, the phase consistency disproving analysis adopts a multi-dimensional disproving approach, which includes at least the following disproving indicators: (1) Audit conclusion jump counter-evidence indicator: used to describe the maximum change in the audit conclusion level under different audit stage assumptions: .

[0073] when A larger value indicates that the audit conclusion has undergone significant changes in level under different stage assumptions, and the stability of the audit conclusion is low.

[0074] (2) Indicator for counter-evidence of constraint response distribution differences: used to describe the degree of difference in the overall distribution structure of constraint response values ​​under different audit stage assumptions:

[0075]

[0076] This indicator can identify situations where the internal constraint response structure changes significantly even if the final audit conclusion is the same.

[0077] (3) Stage Probability-Conclusion Deviation Counter-Indicator: Used to measure the degree of consistency between the audit conclusion and the probability distribution of the audit stage. First, calculate the reference audit conclusion in a probabilistic sense:

[0078] Then calculate the degree of deviation:

[0079] When the review conclusion is mainly dominated by the low-confidence probability stage assumption. It will increase significantly.

[0080] The computer weights and fuses the above multiple counter-evidence indicators to obtain the consistency counter-evidence indicator for the synthesis phase:

[0081] in, The preset weights are used to balance the impact of different counter-evidence indicators on the stability judgment.

[0082] Step S5: When the stage consistency counter-evidence index meets the preset inconsistency condition, the stage modulation amplitude of the constraint strength is suppressed, and a stable audit conclusion is generated based on the suppressed constraint strength as the final audit result output.

[0083] In this step, the computer uses the consistency rebuttal index of the comprehensive stage obtained in step S4. They will determine whether the audit conclusions need to be stabilized.

[0084] when When the inconsistency threshold is exceeded, it indicates that the audit conclusion fluctuates significantly with the audit stage. Directly adopting the audit conclusion under the assumption of a certain audit stage may introduce uncertainty risk.

[0085] As an example, when the phase consistency counter-evidence index meets the preset inconsistency conditions, the dominant role of a single audit phase hypothesis in the audit conclusion can be suppressed by reducing the change amplitude of the phase modulation coefficient.

[0086] In this embodiment, the computer first calculates the stage modulation suppression factor based on the consistency disproving index of the synthesis stage: .in, This is a suppression sensitivity parameter used to control the rate at which the degree of suppression changes with the degree of inconsistency.

[0087] Subsequently, the computer performs suppression and fusion processing on the original stage modulation coefficients. For the first... The modulation coefficient of the settlement audit operation standard after suppression can be expressed as: .in, Represents the modulation coefficients in the original stage. This represents a robust statistical value, such as the mean or median, for the modulation coefficients across multiple review stages.

[0088] Through the above suppression measures, the differences in constraint strength under different audit stage assumptions are compressed, thereby reducing the dominant role of a single extreme stage assumption in the audit conclusion.

[0089] After completing the phase modulation suppression, the computer re-executes the audit calculation based on the suppressed constraint strength to generate a stable audit conclusion: .in, This indicates the final audit conclusion after stabilization.

[0090] As an example, the stabilization audit conclusion may include the audit conclusion obtained by re-performing the audit calculation after suppression processing, or it may include risk-based audit information that reflects the sensitivity of the audit conclusion to the audit stage.

[0091] While completing the phase modulation suppression processing and generating stable audit conclusions, the computer can further output risk-type audit information based on the phase consistency counter-evidence index, which is used to quantitatively characterize the sensitivity of the audit conclusions to changes in the audit phase.

[0092] Specifically, in step S4, the computer has calculated the consistency counter-evidence index for the comprehensive stage based on the differences in audit conclusions under different audit stage assumptions. This comprehensive consistency counter-evidence index reflects the stability of the audit conclusion as the audit stage changes, from multiple dimensions such as the magnitude of the jump in audit conclusion, the difference in the distribution of constraint responses, and the degree of deviation between the stage probability and the conclusion.

[0093] when When the value is small, it indicates that the overall audit conclusion changes little under different audit stage assumptions, and the audit conclusion is not sensitive to the audit stage assumptions; when A larger value indicates that the audit conclusion fluctuates significantly with changes in the audit stage, and that the audit conclusion is highly sensitive to the assumptions made at each audit stage.

[0094] Based on this, in order to transform the aforementioned phase consistency counter-evidence index into a risk quantification result that is easier to use for subsequent processing and decision-making, the computer can... Normalization is performed to generate sensitivity risk values ​​for the review stage. For example, in one implementation, risk-based review information can be calculated as follows:

[0095] in, This represents the sensitivity risk value at the review stage, used to reflect the relative sensitivity of the review conclusion to changes at the review stage; Indicates a consistency counter-evidence indicator in the synthesis phase; This represents the preset risk normalization threshold, used to standardize the scale of counter-evidence indicators for different engineering projects and settlement scales.

[0096] Through the above normalization process, the phase consistency counter-evidence index, which originally had an irregular value range, can be mapped to the [0,1] interval. This facilitates horizontal comparisons between different engineering projects and also serves as a unified input for subsequent risk decision-making or manual review modules. A value close to 0 indicates that the review conclusion is not sensitive to changes in the review stage and has high stability; when... A value close to 1 indicates that the review conclusion is highly sensitive to changes in the review stage and has low stability.

[0097] It should be noted that the risk-type audit information is not used to replace the audit conclusion itself, nor is it used directly to determine whether the settlement data is compliant. Instead, it serves as supplementary information to the stabilization audit conclusion, reflecting the reliability of the audit conclusion under different audit stage assumptions. For example, when the stabilization audit conclusion is "passed," but the corresponding risk-type audit information... A higher value indicates that the conclusion is more sensitive to changes during the review stage, suggesting that it should be given attention during manual review or key spot checks; while a stable review conclusion of "needs adjustment" and risky review information... A lower value indicates that the audit conclusion has a high degree of consistency under different audit stage assumptions and can be used as a relatively reliable audit basis.

[0098] Through the above methods, this embodiment can not only output stable audit conclusions, but also simultaneously output risk-type audit information that reflects the sensitivity of the audit conclusion stage. This improves the interpretability and decision-making reference value of the audit results when the project settlement audit stage is uncertain or the audit process is complex.

[0099] Please see Figure 3 This invention also provides a computer storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any of the preceding claims. It is understood that the computer storage medium is disposed within the aforementioned computer, which at least includes a processor, a communication interface, and a bus for enabling communication between the aforementioned components; specific details will not be elaborated further.

[0100] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. An intelligent auditing method based on settlement auditing operation standards, characterized in that, The method includes the following steps: Step S1: Obtain the project settlement data to be reviewed, and based on the review rule triggering, settlement data adjustment and review process execution trajectory recorded during the settlement review process, extract the review behavior features to characterize the intensity of review execution, and construct the review stage feature vector; Step S2: Calculate the stage confidence probability of the project settlement being in at least two different review stages based on the feature vector of the review stage, and form a review stage probability distribution; Step S3: Model the settlement audit operation standard as a computable constraint function, calculate the constraint response value for the project settlement data, and perform stage modulation on the constraint strength of the constraint response value based on the audit stage probability distribution; Step S4: Perform audit calculations in parallel under different audit stage assumptions to generate corresponding audit conclusions, and perform consistency analysis on the audit conclusions to obtain stage consistency counter-evidence indicators. Step S5: When the stage consistency counter-evidence index meets the preset inconsistency condition, the stage modulation amplitude of the constraint strength is suppressed, and a stable audit conclusion is generated based on the suppressed constraint strength as the final audit result output.

2. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The audit behavior characteristics include at least one of the following: the number of audit rules triggered by unit settlement data, the proportion of adjustment amount of settlement data during the audit process, the proportion of settlement items marked as abnormal, and the number of audit nodes actually executed in the audit process.

3. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The feature vector of the review stage is obtained by numerical processing of the review behavior features and combining them according to preset weights; wherein, the preset weights are used to reflect the degree of influence of different review behavior features on the review stage determination.

4. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The stage confidence probability is obtained by multi-stage classification calculation of the review stage feature vector, and the stage confidence probability is used to characterize the probability that the project settlement is in different review stages.

5. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The constraint response value of the constraint function is a continuous numerical value, used to characterize the degree of deviation of the project settlement data from the settlement audit operation standard, and is not directly used as the result of determining whether there is a violation.

6. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The stage modulation is achieved by introducing a stage modulation coefficient on the constraint response value based on the stage confidence probability, so that the same settlement audit operation standard has different constraint strengths at different audit stages.

7. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The phase consistency counter-evidence index is calculated based on the differences in audit conclusions under different audit phase assumptions, and is used to characterize the stability of audit conclusions as audit phases change.

8. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: When the phase consistency counter-evidence index meets the preset inconsistency conditions, the dominant role of a single audit phase hypothesis in the audit conclusion is suppressed by reducing the change amplitude of the phase modulation coefficient.

9. The intelligent auditing method based on settlement auditing operation standards according to claim 1, characterized in that: The stabilization audit conclusion includes the audit conclusion obtained by re-performing the audit calculation after suppression processing, or it includes risk-type audit information that reflects the sensitivity of the audit conclusion to the audit stage.

10. A computer storage medium storing a computer program thereon, characterized in that: When the computer program is executed by a processor, it implements the method as described in any one of claims 1-9.

Images