Project cost intelligent auditing method and system based on artificial intelligence

Abstract

The invention provides a project cost intelligent auditing method and system based on artificial intelligence, and the method comprises the steps: carrying out the structural analysis and project item extraction of an original document related to the project cost, and forming a standardized project item structural unit set; based on the project item structure unit set, constructing a track chain of each project item in three key stages, namely a budgeting stage, a contract stage and a settlement stage, recording attribute state sequences of the track chain in different stages, and calculating a difference vector to represent the variation amplitude of a key field; performing abnormal behavior identification based on the trajectory chain, identifying a potential abnormal trajectory through a multi-dimensional abnormal scoring function combining numerical value change intensity and clause unit matching missing risk, and generating a structured interpretation prompt unit; and a unified explanation prompt structure is constructed according to the structured explanation prompt unit, and the explanation prompt structure summarizes the serial numbers of the abnormal items, the change fields and the clause missing rate to serve as a finally output structured audit result.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to an intelligent auditing method and system for engineering costs based on artificial intelligence. Background Technology

[0002] In engineering construction projects, the accuracy and rationality of project costs directly affect the realization of investment benefits and the standardization of project management. As a crucial link in the entire process of cost control, project cost auditing covers multiple stages, from project initiation, preliminary budget, budget, bidding, contract signing, to settlement and final accounts. Its core tasks include bill of quantities preparation and verification, accuracy assessment of quota application, rationality evaluation of main material prices and labor costs, and analysis of the completeness and compliance of contract terms. However, in current practical engineering consulting and auditing work, cost auditing still heavily relies on manual experience and rule-based checks, mainly depending on engineers comparing paper or electronic documents line by line and making logical judgments. This is inefficient, highly subjective, and makes it difficult to trace errors. Especially in large or complex projects, with long project cycles, numerous document types, and multiple version changes, information fragmentation and data disconnect at different stages often occur, leading to inconsistencies between drawings and bill of quantities, discrepancies between contract and quotation terms, and drastic price fluctuations at different stages of the same project. Ultimately, this results in audit blind spots and frequent settlement disputes. While some cost estimation software currently possesses basic data verification functions, its technical principles still rely on static field comparisons and preset rule bases. This makes it unable to handle dynamic semantic relationships across stages and documents, and it struggles to effectively link abnormal numerical changes with underlying clause support, unit consistency, and responsibility attribution. Furthermore, it cannot generate standardized, deliverable audit documents. Especially when faced with scenarios involving complex text descriptions, inconsistent expression methods, and highly heterogeneous data sources, existing systems cannot effectively achieve a closed loop from document parsing to result output, resulting in significant deficiencies in the intelligent, interpretable, and standardized management of engineering cost audits. Summary of the Invention

[0003] The purpose of this invention is to design an intelligent engineering cost auditing method and system based on artificial intelligence, which can cover the entire cost auditing chain, has cross-stage behavior modeling capabilities, automatic anomaly identification capabilities, clause interpretation and suggestion generation capabilities, and can ultimately output structured result documents.

[0004] To achieve the above objectives, a first aspect of the present invention provides an intelligent engineering cost auditing method based on artificial intelligence, the method comprising:

[0005] The four types of original documents related to project cost—bill of quantities, contract text, budget list, and settlement list—are structured and parsed to extract project items, forming a standardized set of project item structural units. Each project item includes information such as number, name, specifications, unit, quantity, unit price, source, and referenced clauses.

[0006] Based on the set of structural units of the project items, a trajectory chain of each project item in the three key stages of budget stage, contract stage and settlement stage is constructed, the attribute state sequence of each item in different stages is recorded, and the difference vector is calculated to characterize the change range of key fields.

[0007] Based on the trajectory chain, abnormal behavior is identified. By combining the intensity of numerical change with the risk of missing clause unit matching in a multi-dimensional abnormal scoring function, potential abnormal trajectories are identified and structured explanation prompt units are generated.

[0008] Based on the structured explanation prompt unit, a unified explanation prompt structure is constructed. The explanation prompt structure summarizes the abnormal item number, changed fields and clause missing rate as the final structured audit result.

[0009] Furthermore, the structured parsing steps include: identifying the header and extracting row data from tabular documents; extracting key information from natural language text, including clause numbers, material categories, and pricing methods; and matching and associating project items with contract clauses through similarity calculations.

[0010] Furthermore, the steps for matching project items with contract terms through similarity calculation are as follows: a weighted scoring function is used, which comprehensively calculates the textual similarity between the project item name and the clause description, as well as the numerical similarity between the project item specifications and the clause description.

[0011] Furthermore, the trajectory chain construction steps include: performing cross-stage matching of project items through a trajectory aggregation scoring function, wherein the scoring function comprehensively considers name specification similarity, clause semantic similarity, and quantity differences after unit conversion; extracting the status nodes of each project item in the budget stage, contract stage, and settlement stage, including quantity, unit price, unit, clause reference, and source location; and calculating the difference vector, including the logarithmic scaling values ​​of the unit price change rate and the quantity change rate.

[0012] Furthermore, in the trajectory chain construction step, the trajectory aggregation scoring function used for cross-stage matching includes a unit conversion residual penalty term. This penalty term is calculated based on the difference in quantity ratio under a unified unit, in order to prevent entries with similar names but huge differences in quantity from being mistakenly aggregated.

[0013] Furthermore, in the step of calculating the difference vector to characterize the magnitude of change in key fields, the difference vector consists of two components: the first component is the unit price change rate between the settlement stage and the budget stage; the second component is the value obtained after logarithmically scaling the magnitude of quantity change.

[0014] Furthermore, the multidimensional anomaly scoring function includes: a squared term for unit price change; a squared term for quantity change; and a penalty term for interpretability. The squared term for unit price change is used to indicate whether the unit at any stage in the trajectory chain is inconsistent with the unit at the budget stage. If they are inconsistent, the value is 1; otherwise, it is 0.

[0015] Furthermore, the explanatory prompt structure includes: project item number; a vector of abnormal behavior fields, including the absolute value of unit price change rate, the absolute value of quantity change rate, and a unit change flag; and a clause citation missing rate, reflecting the completeness of clause support at each stage.

[0016] Furthermore, the abnormal behavior identification step also includes: comparing the calculation result of the multidimensional abnormality scoring function with a preset threshold; if the scoring result exceeds the threshold, the corresponding trajectory chain is determined to be a potential abnormal trajectory.

[0017] A second aspect of the present invention provides an intelligent engineering cost auditing system based on artificial intelligence, the system comprising:

[0018] The document parsing module is used to perform structured parsing and extract project items from four types of original documents related to project cost: bill of quantities, contract text, budget list, and settlement list. This forms a standardized set of project item structural units, with each project item containing information such as number, name, specifications, unit, quantity, unit price, source, and referenced clauses.

[0019] The trajectory modeling module is used to construct the trajectory chain of each project item in the three key stages of budget stage, contract stage and settlement stage based on the set of structural units of the project item, record the attribute state sequence of each stage, and calculate the difference vector to characterize the change magnitude of key fields.

[0020] An anomaly identification module is used to identify abnormal behavior based on the trajectory chain. By combining a multi-dimensional anomaly scoring function that matches the intensity of numerical changes with the risk of missing clause units, it identifies potential abnormal trajectories and generates structured explanation prompt units.

[0021] The explanation construction module is used to construct a unified explanation prompt structure based on the structured explanation prompt unit. The explanation prompt structure summarizes the abnormal item number, changed fields and clause missing rate as the final structured audit result.

[0022] The beneficial technical effects of the present invention are at least as follows:

[0023] To address the aforementioned issues, this invention provides an AI-based intelligent auditing method and system for engineering costs. By performing structured parsing of original documents, a unified engineering item unit is established, and a cross-stage trajectory chain is constructed based on this unit, enabling the evolutionary modeling of bill of quantities, contract, and settlement data over time. By introducing a multi-dimensional anomaly identification mechanism, the system can automatically detect potential problems in various scenarios such as numerical mutations, missing clauses, and unit changes, and generate structured prompts. Furthermore, this invention designs an explanation prompt construction mechanism, compressing information such as anomaly causes, missing clause rates, and unit consistency into a standardized input structure to trigger corresponding audit rules. Finally, this invention proposes a result output encapsulation mechanism, automatically mapping the explanation prompt structure into a deliverable audit document, forming a complete list containing risk descriptions and suggested actions. This invention is the first to achieve a closed-loop technical route from document parsing, trajectory modeling, anomaly identification, explanation prompts to result delivery, providing an intelligent engineering cost auditing solution that is structured, interpretable, and practical, significantly improving the efficiency, accuracy, and standardization of auditing work. Attached Figure Description

[0024] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0025] Figure 1 This is a flowchart of the intelligent engineering cost auditing method based on artificial intelligence according to the present invention.

[0026] Figure 2 This is a framework diagram of the AI-based intelligent engineering cost auditing system of the present invention. Detailed Implementation

[0027] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0028] In one or more embodiments, such as Figure 1 As shown, an intelligent engineering cost auditing method based on artificial intelligence is disclosed, the method comprising the following:

[0029] S1: Perform structured parsing and project item extraction on four types of original documents related to project cost: bill of quantities, contract text, budget list, and settlement list, to form a standardized set of project item structural units. Each project item includes information such as number, name, specifications, unit, quantity, unit price, source, and referenced clauses.

[0030] Specifically, the goal of this step is to uniformly parse and structurally process the various types of original documents involved in the project cost audit, forming a standardized set of structural units for project items, which will serve as input for subsequent trajectory modeling and anomaly identification. The key to this step lies in extracting fields, semantically normalizing, and associating clauses from document data of different sources and expressions, so that the same component or list item can be consistently represented at the structural level.

[0031] The input is a collection of documents. These correspond to the bill of quantities, contract text, budget or settlement list, and drawing appendices, respectively. The bill of quantities and budget list are mostly tabular data, and the headers and cells can be directly parsed; the contract text is usually a natural language description; and the drawing appendices are converted into text structure using optical character recognition (OCR). The output of this step is a set of project items. Each project item includes information such as number, name, specifications, unit, quantity, unit price, source, and referenced clauses.

[0032] During processing, the system first uses table parsing methods to identify keywords such as "project name," "unit," "quantity," and "unit price" in the table header, converting each row of data in the tabular document into initial fields for project items. For example, a bill of quantities item "Rebar HRB400 φ16, unit: tons, quantity: 32.54, unit price: 4860" will be directly parsed into name, unit, quantity, and unit price. Next, for descriptions in the contract text, key information is identified using sequence analysis methods, such as "Rebar unit price is based on the cost information of May 2023." To achieve this identification, the following scoring function is used:

[0033] ;

[0034] in, Indicates the first The probability that a word is the target field; Indicates from the first word to the... The embedding sequence of words; This represents a context representation function based on a bidirectional recurrent neural network, used to capture the contextual relationships between words; and For parameter matrices; This represents an element-wise nonlinear transformation; This is the activation function. The probability calculated by this function can effectively highlight fields such as clause numbers, material categories, and pricing methods in the contract text.

[0035] Next, the structured fields need to be matched with the natural language fields in the contract terms. For this purpose, a similarity calculation function is introduced:

[0036] ;

[0037] in, For engineering projects With contract terms Match score; Indicates the name of the project item; Indicates the specifications and parameters of the project item; This represents a name similarity function based on text edit distance; This represents a specification similarity function based on numerical normalization; and This is a weighting factor used to balance the weight of name and specification in the matching process. For example, when the name in the bill of quantities is "φ16 rebar," and the contract terms describe it as "the contractor shall bear the price adjustment for φ16 HRB400 rebar," after normalization, the specification field will be converted to the value "16." The system assigns the clause number as a high score. Reference fields .

[0038] The final output set of project items In, each Includes: Number (Generated from document number and line number), Name ,Specification ,unit ,quantity ,unit price ,source Referenced Terms For example, the result for a rebar item might be: Number "BOQ-15", Name "Rebar HRB400 φ16", Unit "tons", Quantity "32.54", Unit Price "4860", Source "Bill of Quantities, Line 15", Referenced Clause "Contract Clause 3.2.1". This output ensures consistent fields, traceable sources, and consistent clause associations, providing high-quality input for subsequent trajectory modeling.

[0039] S2: Based on the set of structural units of the project items, construct the trajectory chain of each project item in the three key stages of budget stage, contract stage and settlement stage, record its attribute state sequence in different stages, and calculate the difference vector to characterize the change range of key fields.

[0040] Specifically, this step involves the set of structural units for the engineering items output in step one. Based on this, for each project item, we construct its numerical and source evolution path at each key stage (such as budgeting, bidding, contracting, and settlement) throughout the entire project process, forming a set of trajectory chains. Each trajectory chain It describes the project items The sequence of attribute states exhibited at each stage is used for subsequent anomaly identification and review suggestion generation.

[0041] The core technical objective of this step is to construct a unified cross-stage expression by deeply comparing and standardizing information such as the representation, name variations, specification expressions, unit conversions, and clause references of the same component or clause item in different documents, enabling the system to have the capability to handle "structural items" from the perspective of cross-stage expression. "to the evolutionary chain" The ability to map data is crucial. Engineering projects often have issues such as inconsistent naming, slight adjustments to specifications, and changes in quantity without recording the reasons for the modifications. Without effective modeling, costs can easily spiral out of control during the settlement phase.

[0042] The system will first structural items in By number Preliminary classification, and identification of documents from each stage that are identical or similar. and The entries in the field are matched. To this end, a trajectory aggregation scoring function is introduced. The function is designed to score whether two project items should be aggregated into the same trajectory chain.

[0043] ;

[0044] in: For structural items and Trajectory aggregation scoring between; The similarity score for the combination of name and specification fields is achieved by combining character edit distance and numerical normalization. For the clause number and The score for whether it points to similar logical semantic content is based on keyword matching in the clause title and comparison of contextual semantic fragment vectors; The difference in quantity ratio after unit conversion is specifically defined as:

[0045] ;

[0046] in and These refer to quantities under a unified unit; coefficients. , Weighting coefficients for similarity scores (empirically set as follows) ), For the coefficient of the abnormal penalty term (e.g.) This feature effectively prevents entries that are highly similar in fields but have vastly different numbers from being mistakenly aggregated.

[0047] Compared to traditional field matching methods, this trajectory aggregation scoring function adds a unit conversion residual penalty term. As a "quantity rationality filtering mechanism" unique to engineering scenarios, it enables the effective differentiation of different measurement unit items with similar names in the list (such as "rebar φ16" and "rebar φ16 (allocation)"), which has strong engineering practicality and creativity.

[0048] After the aggregation of entries is completed, for each project item trajectory chain... The system will extract its state nodes at each stage. ,in Indicates the stage number (e.g., budget is...). The contract is The settlement is Each state node This includes attributes such as quantity, unit price, terms and conditions, and location of the source document.

[0049] To further characterize the trajectory evolution features, a trajectory difference vector is introduced. This is used to describe the magnitude of changes in key fields of an engineering item across different stages, and is defined as follows:

[0050] ;

[0051] in: Representing project items The unit price during the budget and settlement phase comes from... and of Field; These are quantity fields for the budget and settlement stages, respectively. To avoid division by zero errors, the smallest positive number (set to) The first term reflects the price increase, and the second term is the logarithmic scaling of the quantity change, which has better engineering robustness and is more sensitive to small changes.

[0052] The vector These features will serve as important input features for subsequent identification models and can also be directly used to develop risk control scores.

[0053] Ultimately, the system will each Corresponding trajectory chain Stored as a structure containing:

[0054] Project Item Number (Continue with the output from step one);

[0055] Each stage Corresponding state nodes This includes fields such as quantity, unit price, unit, and referenced terms;

[0056] Difference vector , as an indicator of trajectory evolution;

[0057] Aggregated ratings Matching details with each field can be used for source tracing visualization.

[0058] Specifically, the quantity of the steel reinforcement component numbered "BOQ-15" in the budget stage is... Ton, unit price Yuan, the quantity during the settlement stage is Ton, unit price If the element is a given element, then its trajectory difference vector is:

[0059] ;

[0060] The result indicates that the price increased by 17.1% and the quantity increased by approximately 20%, exceeding the internal warning threshold set by the company. Therefore, this project will subsequently enter the anomaly identification module.

[0061] This step involves constructing a trajectory chain. and evolution vector It realizes the transformation of the original static structural items into cross-stage semantic evolution objects, and is the core supporting structure for subsequent cost anomaly identification, responsibility stage positioning and contract interpretation generation.

[0062] S3: Based on the trajectory chain, perform abnormal behavior identification, and identify potential abnormal trajectories by combining the numerical change intensity with the clause unit matching missing risk through a multi-dimensional abnormal scoring function, and generate structured explanation prompt units;

[0063] Specifically, this step uses the trajectory chain set constructed in step two. As input, the system identifies anomalous behaviors in each project item's evolution across multiple project phases, focusing particularly on whether the changes are reasonable and whether they are justifiable. Anomaly identification goes beyond simply judging numerical differences; it involves a comprehensive assessment considering multiple dimensions, including project context, clause references, unit changes, and missing phase behaviors, generating structured prompts for subsequent semantic interpretation. This step serves as both the judgment engine and a pre-interpretation mechanism.

[0064] This step introduces a multidimensional anomaly scoring function for each trajectory chain. Perform fine-grained anomaly modeling.

[0065] Considering the unique problem scenarios in engineering cost auditing, such as "the quantity remains unchanged but the price jumps sharply", "the unit changes but the pricing method is not updated", and "the contract is missing key clauses", this step designs an anomaly scoring function with penalty constraints. Modeling is performed by combining the intensity of numerical change with the risk of missing clause-unit matching:

[0066] ;

[0067] in: The unit prices for the budget and settlement stages are respectively derived from... and ; The number of corresponding stages; The minimum value set to prevent the denominator from being zero is set to 0. ;coefficient Sensitivity weights are commonly used in engineering practice. The "interpretation impossibility penalty" is used to measure the degree of missing or conflicting information in terms and units of the trajectory.

[0068] Among them, the penalty item The definition of is:

[0069]

[0070] in, For trajectory chain The total number of stages (usually 3); Representation phase The units are inconsistent with those in the initial stage; Representation phase Missing clause number; indicator function The value is 1 if the condition is met, and 0 otherwise.

[0071] Overall, The larger the value, the more "unexplained field changes" or "missing clause references" exist in the trajectory, which is a major cause of actual audit disputes.

[0072] This structure breaks away from the traditional approach of only looking at numerical changes and proposes an abnormal cognitive model of "insufficient explanation is risk". It reflects that audit engineers pay more attention to the standards of "whether it is reasonable" and "whether it is compliant" in practice, rather than just "whether it has changed".

[0073] Scoring function After standardization in the system, it is compared with the threshold. In comparison, if ,but Identified as a potentially anomalous trajectory. Threshold. This setting can be based on the company's historical data and experience; the recommended initial value is [value missing]. .

[0074] When a trajectory is identified as abnormal, the system will generate a structured explanation prompt unit for that entry. It contains the following fields:

[0075] Project Item Number ;

[0076] Abnormal fields and their descriptions of phased changes (e.g., "Quantity changed from 30 tons to 36 tons, clause missing");

[0077] The relative changes in unit price and quantity;

[0078] all A snapshot of the fields, including name, specifications, unit, and clause number;

[0079] Source location, such as "BOQ line 15" or "Contract clause 3.2.1";

[0080] Missing or inconsistent clause references;

[0081] A set of stage numbers where a match failed.

[0082] Specifically, if a component is numbered "BOQ-20", its budgeted unit price is 4800 yuan, and its settlement unit price is 5680 yuan, with the unit changing from "tons" to "kilograms", and the contract clause lacks a reference to the component number, then its... The value is:

[0083] ;

[0084] As set If so, the item is identified as abnormal and generated. As input for the subsequent explanation module.

[0085] The output of this step is a collection of abnormal trajectories. With structured hint set Each of them This can serve as structural input to the semantic generation module, forming a "hint context." This step proposes an explanatory power constraint. It effectively connects the needs of cost trajectory modeling and semantic interpretation, and realizes a closed-loop connection from "trajectory chain → anomaly identification → interpretation construction" in terms of structural logic, ensuring that the system can not only identify data anomalies in audit operations, but also provide audit explanations with sufficient context and logical reasoning.

[0086] S4: Construct a unified explanation prompt structure based on the structured explanation prompt unit. The explanation prompt structure summarizes the abnormal item number, changed fields and clause missing rate as the final structured audit result.

[0087] Specifically, this step involves the set of abnormal trajectories output in step three. With structured hint set Based on this, construct a unified set of explanation and prompt structures. The core task of this step is to aggregate the structural information into a set of interpretation-driven fields, which can then be used by the next step to directly generate the review report or suggestion text. Therefore, its goal is not to generate natural language, but to design a set of intermediate layer structures that provide a unified semantic entry point and judgment reference for subsequent output logic.

[0088] Each From the previous step, the fields include: Project Item Number ; Difference vector ,in This represents the rate of change in unit price. The rate of change in quantity is derived from the trajectory chain; the set of missing terms. This indicates whether there are clause numbers at each stage; unit set This indicates the unit of measurement for each stage; anomaly scoring The calculation has already been performed in the previous step.

[0089] To construct the explanation and hint structure, the system defines:

[0090] ;

[0091] in: It is the structural item number; It is a vector of combined abnormal behavior fields, which is an extension of the difference vector; This is the clause citation missing rate, used to measure the completeness of the explanation and support.

[0092] Explanation field vector The construction method is as follows:

[0093] ;

[0094] in: This represents the rate of change in unit price between the settlement stage and the budget stage; , representing the corresponding rate of change in quantity; This is used to prevent the denominator from being zero; This indicates whether the units used in the settlement phase differ from those used in the budget phase, and is used to indicate the risk of unit consistency.

[0095] Clause citation missing rate The calculation is as follows:

[0096] ;

[0097] in This is the total number of stages, typically 3. This indicator reflects the degree of lack of supporting evidence for the clauses and is an important signal for assessing the reasonableness of the interpretation.

[0098] All fields come from the structure hint unit in the previous step. No new data sources or redundant variables are introduced. This structure can be directly fed into the template matching system of the generation module or the language generation module to achieve precise selection of "interpreting semantic fragments" or "suggesting template numbers." This step is not responsible for generating natural language text; it only constructs the structured input that drives the process.

[0099] Specifically, suppose the component with the number "BOQ-17" has the following anomaly:

[0100] Yuan, Yuan, ;

[0101] ton, ton, ;

[0102] The unit is "ton" in both stages. ;

[0103] The clauses were missing during the contract stage. ;

[0104] but for:

[0105] ;

[0106] This structure will ultimately be used to trigger the rule "There is a unit price increase, but the contract terms are missing. It is recommended to supplement the basis for the main material adjustment".

[0107] Based on structured interpretation hint set For each anomaly, construct a standardized output structure unit. This generates a document structure that can be exported as audit reports, audit checklists, and recommendation records. Anomaly identification, scoring, or semantic interpretation is no longer performed here; the process only transforms the explanatory prompts into a standardized audit deliverable structure.

[0108] Each The structure includes the following:

[0109] : Structural item number;

[0110] : These represent the rate of change in unit price, the rate of change in quantity, and whether the unit has changed (1 or 0), respectively.

[0111] Missing citation rate, from statistics of the previous period;

[0112] System constructs output structure unit The format is defined as follows:

[0113] ;

[0114] in: For risk description field, based on and The values ​​are used to generate templates; For the suggested action field, the system uses a unit change flag. With the rate of missing clauses Determine whether supplementary clauses or explanations are needed; Risk description The generation uses a standard template matching mechanism, with the following logic:

[0115] like The following message is generated: "The unit price of this project item increased by more than 15% during the settlement stage."

[0116] like Added: "Quantity change exceeds 10%";

[0117] like Add: "The unit has changed";

[0118] like Add: "The phase clause is missing."

[0119] Suggested content The generation rules are as follows:

[0120] like The output reads: "It is recommended to supplement the basis for unit adjustments during the settlement phase and the corresponding clause explanations."

[0121] like and The output reads: "It is recommended to provide supplementary materials explaining the price adjustment, or to confirm whether there was an error in the information."

[0122] For other cases, the output should read: "It is recommended that auditors manually review and confirm the handling method."

[0123] This structured output field It is automatically constructed through a template mapping mechanism, requiring no manual input of language descriptions, and possesses controllability and reproducibility. All results are uniformly packaged into a unified output set. It supports exporting to standard report formats (such as JSON, Excel, PDF, etc.).

[0124] Specifically, assuming the input for:

[0125] = "BOQ-22";

[0126] ;

[0127] ;

[0128] The matching template is as follows:

[0129] → “Unit price increased by more than 15%”;

[0130] → "Quantity change exceeds 10%";

[0131] → “Unit change”;

[0132] → “Terms missing”.

[0133] Final output:

[0134] ;

[0135] all Output set As a core structural part of the final audit document, the system can provide an export interface, allowing users to generate a final audit recommendation list, opinion appendix, or system review result summary with one click.

[0136] In one or more embodiments, such as Figure 2 As shown, an intelligent engineering cost auditing system based on artificial intelligence is disclosed, the system comprising:

[0137] The document parsing module is used to perform structured parsing and extract project items from four types of original documents related to project cost: bill of quantities, contract text, budget list, and settlement list. This forms a standardized set of project item structural units, with each project item containing information such as number, name, specifications, unit, quantity, unit price, source, and referenced clauses.

[0138] The trajectory modeling module is used to construct the trajectory chain of each project item in the three key stages of budget stage, contract stage and settlement stage based on the set of structural units of the project item, record the attribute state sequence of each stage, and calculate the difference vector to characterize the change magnitude of key fields.

[0139] An anomaly identification module is used to identify abnormal behavior based on the trajectory chain. By combining a multi-dimensional anomaly scoring function that matches the intensity of numerical changes with the risk of missing clause units, it identifies potential abnormal trajectories and generates structured explanation prompt units.

[0140] The explanation construction module is used to construct a unified explanation prompt structure based on the structured explanation prompt unit. The explanation prompt structure summarizes the abnormal item number, changed fields and clause missing rate as the final structured audit result.

[0141] It is worth noting that the specific workflow of the AI-based intelligent engineering cost audit system provided in this embodiment is the same as that of the AI-based intelligent engineering cost audit method described in the above embodiments, and will not be repeated here.

[0142] This invention also provides an artificial intelligence-based intelligent engineering cost auditing device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above-described embodiment of the artificial intelligence-based intelligent engineering cost auditing method, for example... Figure 1 The steps S1 to S4 described above; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0143] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the AI-based intelligent engineering cost auditing device.

[0144] The AI-based intelligent cost auditing device can be a desktop computer, laptop, handheld computer, or cloud server, among other computing devices. This AI-based intelligent cost auditing device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the AI-based intelligent cost auditing device may also include input / output devices, network access devices, buses, etc.

[0145] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASACs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the AI-based intelligent cost auditing equipment, connecting all parts of the equipment via various interfaces and lines.

[0146] The memory can be used to store the computer programs and / or modules. The processor realizes various functions of the AI-based intelligent engineering cost auditing device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0147] The modules integrated into the AI-based intelligent engineering cost auditing equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0148] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0149] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based intelligent construction cost auditing method, characterized in that, The method comprises: The method comprises: The method comprises: Based on the engineering item structure unit set, the trajectory chain of each engineering item in the budget stage, the contract stage and the settlement stage is constructed, the attribute state sequence of each engineering item in different stages is recorded, and the difference vector is calculated to represent the change amplitude of the key field. Based on the trajectory chain, abnormal behavior identification is performed, a multi-dimensional abnormal scoring function is combined with the numerical change intensity and the missing risk of the clause unit to identify potential abnormal trajectories, and a structured explanation prompt unit is generated. 2.The method of claim 1, wherein, According to the structured explanation prompt unit, a unified explanation prompt structure is constructed, which summarizes the number of abnormal items, the change field and the clause missing rate as the final output of the structured audit results. 3.The method of claim 1, wherein, The structured analysis step comprises: table header recognition and row data extraction for table type documents; key information extraction for natural language text, including clause number, material category and pricing method; and matching and association of engineering items and contract clauses through similarity calculation. 4.The method of claim 1, wherein, The step of matching engineering items and contract clauses through similarity calculation is: a weighted scoring function is used to comprehensively calculate the text similarity of engineering item name and clause description, and the numerical similarity of engineering item specification parameters and clause description. 5.The method of claim 1, wherein, The trajectory chain construction step comprises: cross-stage matching of engineering items through a trajectory aggregation scoring function, which comprehensively considers the name specification similarity, clause semantic similarity and quantity difference after unit conversion; extracting the state nodes of each engineering item in the budget stage, the contract stage and the settlement stage, including quantity, unit price, unit, clause reference and source location; and calculating the difference vector, including the logarithmic scaling value of the unit price change rate and the quantity change rate. 6.The method of claim 1, wherein, In the trajectory chain construction step, the trajectory aggregation scoring function used for cross-stage matching contains a unit conversion residual penalty term, which is calculated based on the quantity proportion difference under the unified unit, and is used to prevent items with similar names but large quantity differences from being mis-aggregated. 7.The method of claim 1, wherein, In the step of calculating the difference vector to represent the change amplitude of the key field, the difference vector is composed of two components: the first component is the unit price change rate in the settlement stage and the budget stage; and the second component is the value obtained by logarithmic scaling of the quantity change amplitude. 8.The method of claim 1, wherein, The multi-dimensional abnormal scoring function comprises: a unit price change square term; a quantity change square term; and an explanation impossibility penalty term; the unit price change square term is used to indicate whether the unit in any stage of the trajectory chain is inconsistent with the unit in the budget stage, and if inconsistent, the value is 1, otherwise 0. The explanation prompt structure comprises: engineering item number; abnormal behavior field combination vector, including unit price change rate absolute value, quantity change rate absolute value and unit change flag; and clause reference missing rate, reflecting the integrity of the clause support in each stage. The explanation prompt structure comprises: engineering item number; abnormal behavior field combination vector, including unit price change rate absolute value, quantity change rate absolute value and unit change flag; and clause reference missing rate, reflecting the integrity of the clause support in each stage. 9.The method of claim 1, wherein, The abnormal behavior identification step further includes: comparing the calculation result of the multi-dimensional anomaly scoring function with a preset threshold, and if the scoring result exceeds the threshold, determining that the corresponding trajectory chain is a potential abnormal trajectory.

10. The artificial intelligence-based intelligent engineering cost auditing system, characterized in that, The system comprises: A document analysis module is configured to analyze and extract four types of original documents related to project cost, including bill of quantities, contract text, budget list, and settlement list, to form a set of standardized engineering item structure units, each engineering item including number, name, specification, unit, quantity, unit price, source, and reference clause information. A trajectory modeling module is configured to construct a trajectory chain of each engineering item in three key stages, i.e., the budget stage, the contract stage, and the settlement stage, based on the set of engineering item structure units, record the attribute state sequence of each engineering item in different stages, and calculate a difference vector to represent the change amplitude of key fields. An abnormality identification module is configured to identify potential abnormal trajectories based on the trajectory chain by combining a multi-dimensional anomaly scoring function of numerical change intensity and clause unit matching missing risk, and generate a structured explanation prompt unit. An explanation construction module is configured to construct a unified explanation prompt structure according to the structured explanation prompt unit, the explanation prompt structure summarizing the number of abnormal items, the changed fields, and the clause missing rate as the final output of the structured review results.

Images