Auditing quality evaluation method and evaluation system for auditing process of construction project of power distribution network

Abstract

The invention discloses an auditing quality evaluation method for an auditing process of a power distribution network construction project, and the method comprises the following steps: 1) collecting and compiling a list of all materials and equipment needed in the whole construction process of the power distribution network project, and coding all used materials and equipment; 2) generating construction cost budget data according to project requirements; 3) identifying an actual construction cost file, extracting typical data in the cost file, matching the typical data with the list in the step 1, and coding the typical data; 4) carrying out comparative analysis on the coded typical data construction cost budget data, and identifying the difference between the two data; and 5) summarizing the differences to generate a difference report. The invention further provides an auditing quality evaluation system for the auditing process of the construction project of the power distribution network. The invention establishes and standardizes a cost and cost encoding system, including all related cost categories. In addition, a new coding structure is implemented, and consistency and accuracy of data are ensured.

Description

Technical Field

[0001] The present invention is mainly aimed at the field of engineering cost, and provides a method for evaluating the quality of distribution network project engineering audit data and a related evaluation system to solve the problem of data batch and automatic analysis in distribution network project engineering audit data, and at the same time collect and extract valid data in the file. Background Art

[0002] In the field of distribution network engineering cost, to which the present invention is applied, most projects adopt the quota cost method. When the quota cost is used, the project settlement statement and the corresponding attachments are the common data composition and preparation methods in engineering audit. This method uses the project settlement statement to estimate the total cost of the entire project based on the national or industry standard quota (i.e., the pre-set standard cost of a single project or workload), and provides other attachments for material support.

[0003] The quota cost of distribution network projects is usually based on national or local construction project cost quota standards. These standards include the predetermined prices and consumption of various materials, labor and machinery. By adopting a unified pricing basis, the quota cost aims to achieve standardization and rationalization of project budget compilation and improve the scientificity and accuracy of cost management. At the same time, the corresponding visas, hidden project certification materials and other supporting attachments are used to ensure the authenticity and compliance of the project cost.

[0004] Traditional engineering audit methods mainly rely on manual inspection and verification, which has the following shortcomings:

[0005] Low computational efficiency and high technical requirements: Project audits require checking a large amount of documents and data, including contracts, settlement statements, construction drawings, change orders, etc., which are time-consuming and laborious to process manually. As the audit process is complex and cumbersome, it often takes a long time to complete the entire audit process, affecting the overall progress of the project.

[0006] Cost reduction strategies are prone to subjective bias: different auditors may have different understandings and judgments on the use of the same quota, resulting in highly subjective audit results.

[0007] Lack of consideration of correlation: The work correlation between business departments and audit units is strong, and traditional methods cannot fully reflect the interaction and mutual influence between the two.

[0008] The cost differences throughout the entire process are difficult to compare: the existing audit methods lack scientific and systematic quality measurement standards, making it difficult to objectively measure the quality and effectiveness of the cost differences throughout the entire process.

[0009] These shortcomings significantly limit the efficiency and accuracy of engineering audits, making it difficult to meet the high requirements of modern engineering management for audit work. Summary of the invention

[0010] The purpose of the present invention is to provide a power grid project engineering audit data quality evaluation method and evaluation system to solve the problems of low efficiency, strong subjectivity, lack of correlation consideration and so on in the existing engineering audit process.

[0011] The present invention adopts the following technical means to achieve: an audit quality evaluation method for the audit process of a distribution network construction project, comprising the following steps:

[0012] 1) Collect and compile a list of all materials and equipment required during the construction of the entire distribution network project, and encode all materials and equipment used;

[0013] 2) Generate construction cost budget data according to project requirements;

[0014] 3) Identify the actual construction cost documents, extract typical data from the cost documents, match the typical data with the list in step 1, and then encode the typical data;

[0015] 4) Compare and analyze the coded typical data construction cost budget data to identify the differences between the two;

[0016] 5) Summarize the differences and generate a difference report.

[0017] In step 1), all distribution network construction projects are divided into several decision tree links, each of which is composed of a set of typical power grid design data, which constructs a standardized module cost data with a three-level cross-reference relationship of module-material-quota.

[0018] The typical design data features of the power grid include upper-level data, lower-level data and environmental data. The upper-level data include pure civil engineering project type data, pole erection project type data, overhead line laying project type data, cable laying project type data, electrical equipment project type data and branch box construction type data. The lower-level data is the specific construction project of each upper-level data. The environmental data includes indoor, open-air, mine and underwater. The upper-level data, lower-level data and environmental data of the same group are combined to form several decision tree links, and each decision tree link corresponds to an actual distribution network construction project.

[0019] The lower-level data include the core power distribution equipment and related auxiliary construction materials required for the project. The equipment and auxiliary construction materials are set with initial values, which are the quantities of equipment and materials used under normal circumstances.

[0020] The decision tree link formed in step 1) is optimized. The specific steps are as follows:

[0021] 1-1) Divide the decision tree data features into binary and obtain several groups of binary reflexive features;

[0022] 1-2) Calculate the Gini impurity of each feature and perform weighted average calculation on it;

[0023] 1-3) Select the segmentation feature with the smallest Gini impurity among all features as the best segmentation point;

[0024] 1-4) Use the best split point to divide the data set into two subsets, and repeat steps 1-2) and 1-3) for each subset until the decision tree data node reaches the maximum depth.

[0025] In step 1-2), the weighted average calculation formula is as follows:

[0026] Among them Gini split1 Its representative feature 1 is the Gini impurity of the split point, n is the total number of samples, and n Y , n N The number of samples with and without this feature, Gini Y With Gini N Corresponding to the Gini impurity with and without this feature respectively.

[0027] Gini is the Gini impurity, and its function is In the formula, C represents the classification set, which is all typical designs in the present invention, i represents the i-th classification, and p(i) is the i-th classification frequency of the training sample.

[0028] Step 2) specifically includes the following steps:

[0029] 2-1) According to the actual construction plan, determine the corresponding complete decision tree link, and obtain the initial cost project data from the complete decision tree link;

[0030] 2-2) offset the cost data one by one to generate corresponding offset data sets;

[0031] 2-3) Analyze the initial data and the data around the offset point using a local weighted regression model, and solve for the local slope;

[0032] 2-4) Analyze the impact of design parameters on total cost based on local slopes and integrate relevant information into a report.

[0033] The local weighted regression model is: Y≈β0(x)+β1(x)·X, where Y is the cost result, X is the variable composed of all parameters affecting the cost, X={x1,x2,x3,....,x n}, β0(x) and β1(x) are the local constant and local slope based on point x.

[0034] In step 3), several coding mapping tables are used to identify the actual cost documents to ensure that all data can be correctly and completely converted.

[0035] The differences mentioned include missing attachments to the cost document, abnormal number of adjustments to the cost document data, and inconsistency between the cost document data and the construction cost budget data.

[0036] Step 4) also includes prejudgment of suspicious points, where the suspicious points are quota items with high data change frequency, i.e., frequent items. The mining steps of the frequent items are as follows:

[0037] 4-1) Construct an initial FP tree and mine frequent itemsets from it;

[0038] 4-2) Analyze frequent item sets and extract association rules.

[0039] The initial FP tree construction steps are as follows:

[0040] First, find the combination of modules, materials, and quotas that are different from the budget data, and sort them in descending order of support. Then sort each module, material, and quota according to its frequency of occurrence, and then start from the root node and add entries to the tree in the sorted order. If the prefix of a path is the same as an existing path, reuse this path and increase the corresponding count; otherwise, create a new branch, and finally form the initial FP tree.

[0041] The association rule extraction method is as follows:

[0042] For each frequent item set, the possible associations between the items are analyzed. If the quota changes under the condition that the specified modules and materials appear, the data set is recorded as a rule, and then the support and confidence of each rule are calculated. The support is the proportion of the rule in all cost data sets, and the confidence is the probability of the rule occurring.

[0043] An audit system for the audit process of a distribution network construction project, comprising:

[0044] Data entry module: used to enter actual cost documents;

[0045] Data conversion module: used to identify and encode cost documents;

[0046] A data generation module is used to generate construction cost budget data;

[0047] Data comparison module, used to compare construction cost budget data with cost document data;

[0048] Report generation module, used to generate the final differential report.

[0049] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that the memory stores the computer program, and when the processor executes the computer program, it implements any of the aforementioned methods for evaluating the audit quality of the audit process of a distribution network construction project.

[0050] A computer-readable storage medium having a computer program stored thereon, wherein the storage medium includes a stored program, wherein when the program is run, the device where the storage medium is located is controlled to execute any of the aforementioned methods for evaluating the audit quality of a distribution network construction project audit process.

[0051] The present invention establishes and standardizes the coding system of construction cost, including all relevant cost categories. In addition, a new coding structure is implemented to ensure the consistency and accuracy of the data. Through the corresponding mechanism and version maintenance of the old and new codes, even the data input based on the old codes can be correctly parsed and analyzed; the module of the present invention can conduct in-depth comparative analysis of structured cost data, identify and report price differences and quantity differences, and support multi-dimensional data analysis. Through detailed difference reports, it helps the audit team to quickly understand the problem and provide a basis for adjusting the budget and improving project management; and the present invention uses the FP-growth algorithm to automatically mine frequent item sets, identify quota items that are often replaced under specific conditions, and automatically mark suspicious points that have not been adjusted in the review stage. Provide a systematic method to identify and focus on potential cost problem points, and enhance the automation and intelligence level of auditing. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is the overall functional architecture diagram of the present invention;

[0053] Figure 2 An example flow chart of editing cost rules of the present invention;

[0054] Figure 3 A flow chart of the cost coding of the present invention at different stages;

[0055] Figure 4 This is a flow chart of data recording and mining of cost difference items of the present invention;

[0056] Figure 5 Prepare a flow chart for the settlement statement of the present invention;

[0057] Figure 6 This is a flowchart of doubt point extraction and settlement audit of the present invention. DETAILED DESCRIPTION

[0058] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0059] The present invention relates to an audit quality evaluation method for a distribution network construction project audit process, comprising the following steps:

[0060] 1) Collect and compile a list of all materials and equipment required during the construction of the entire distribution network project, and encode all materials and equipment used;

[0061] 2) Generate construction cost budget data according to project requirements;

[0062] 3) Identify the actual construction cost documents, extract typical data from the cost documents, match the typical data with the list in step 1, and then encode the typical data;

[0063] 4) Compare and analyze the coded typical data construction cost budget data to identify the differences between the two;

[0064] 5) Summarize the differences and generate a difference report.

[0065] Specifically, the cost rule editing process is as follows:

[0066] Step 101) Determine the list of construction materials and standard format for the distribution network project: This step involves collecting and compiling a list of all materials and equipment required throughout the distribution network construction process, and coding all materials and equipment used. This includes the demand for core distribution equipment such as cables, switches, transformers, and auxiliary materials required for construction. Develop a standardized format for recording various documents and data generated during the construction process to ensure that all information is listed in a timely and accurate manner for effective auditing and management. Define a systematic relationship mapping to correspond the materials in the material requisition list to specific engineering quantities (such as line laying length, number of switches, etc.). According to the engineering design and construction requirements, the distribution network project is broken down into several smaller, manageable modules. Each module contains a certain amount of engineering and corresponding material requirements, and a standardized module cost data with a three-level cross-reference relationship of module-material-quota is constructed.

[0067] Step 102) The system automatically organizes typical design data of the power grid: typical design project categories such as pure civil engineering, pole-related, overhead part, cable part, electrical part, branch box, etc.; sub-categories of subordinate design projects, such as the cable part can be subdivided into branch line modification, cable maintenance, new cable construction, etc.; subordinate design environmental factors, such as the laying environment of new cables, indoor, open-air, mine, underwater, etc. At the same time, the environmental selection will also have the same indoor, open-air, mine, underwater, etc. options in pole-related. Finally, each leaf node in the decision tree is defined to correspond to a typical design, which details the parameters, module types and quantities used under the specific conditions. For example, for cable laying in overhead lines, the parameter is 100 meters, and the system automatically derives the module type and quantity based on the typical design.

[0068] Step 103) Train the decision tree using real items:

[0069] Step 103-1) Real project decision data collection: Based on the classification and selection rules of typical power grid designs, a decision tree is designed to assist cost estimators in identifying typical designs. The system reads the decision tree links of batches of real projects, i.e., a complete decision tree link that includes the leaf nodes (i.e., typical designs) reached by the appropriate project categories, subcategories, and environmental factors.

[0070] Step 103-2) Binary division of real project decision data features: The above decision tree link data is transformed, and each non-stratified choice is used as a typical design identification feature. At the same time, only the binary division of features is recorded at this time, that is, pure civil engineering and non-pure civil engineering are a set of features, and indoor and non-indoor are a set of features to form a training data set S. Design a typical design identification decision tree based on the CART classification tree, and use the Gini impurity index as the basis for node division.

[0071] Step 103-3) Calculate the Gini impurity of each feature: traverse each feature. For each feature, since the above steps have been binary divided, there is only a split point, and the data is divided into two parts: one part contains the data of the feature, and the other part does not contain the data of the feature. Calculate the Gini impurity of each part, and calculate the weighted average Gini impurity according to the following formula: Among them Gini split1 Its representative feature 1 is the Gini impurity of the split point, n is the total number of samples, and n Y , n N The number of samples with and without this feature, Gini Y With Gini N Corresponding to the Gini impurity with and without this feature, Gini is the Gini impurity function In the formula, C represents the classification set, which is all typical designs in the present invention, i represents the ith classification, and p(i) represents the probability of the ith classification, which is the ith classification frequency of the training sample in the present invention.

[0072] Step 103-4) Select the best segmentation feature: Select the segmentation feature with the smallest Gini impurity among all features as the best segmentation point.

[0073] Step 103-5) Recursive segmentation: Use the optimal segmentation point determined above to divide the data set into two subsets. Repeat steps 103-3) and 103-4) on each subset until the decision tree node reaches the maximum depth, that is, the typical design is fully covered.

[0074] Step 104) Constructing the cost rule editing interface: This step involves creating an interactive web interface for editing and managing the cost rules of the distribution network project. The interface is designed as a tree structure, allowing users to define and refine the composition of the construction investment step by step. Each cost includes a name and allows users to edit the calculation logic (such as which items are added and multiplied by the corresponding rate). In addition, users can edit and adjust the rates directly on the interface according to the quota calculation rules to adapt to the specific needs of different projects.

[0075] Step 105) Edit the construction investment composition: In this step, the user will specifically edit the various components of the construction investment. This includes engineering costs (such as construction and installation engineering costs, main materials, equipment and tools purchase costs, etc.), other construction costs (such as project construction management fees, land and engineering preparation costs, etc.) and reserve costs (basic reserve costs and price difference reserve costs). Each type of cost has a corresponding code, and the user can define the calculation logic of these costs through the interface, for example, adding specific cost items and multiplying them by the corresponding rate to calculate the total cost.

[0076] Step 106) Mapping the logical relationship of cost rules: In this step, the system will allow users to establish and maintain the logical relationship between cost rules to ensure that the cross-reference relationship of cost data meets industry standards and actual project needs. This includes matching specific engineering quantities (such as line laying length, number of switches, etc.) with related material requirements and quota standards. The system provides tools to help users automatically detect and correct logical errors in cost rule definitions to ensure the accuracy and completeness of cost data.

[0077] The cost coding process is as follows:

[0078] Step 201) Define the cost coding structure: This step involves establishing and standardizing the cost coding system for unified use throughout the project audit system. The design of the coding structure will take into account all relevant cost categories, such as engineering costs, equipment costs, material costs, and labor costs. Each cost category will be assigned a unique code identifier, which helps ensure consistency and accuracy during data entry, processing, and analysis.

[0079] Step 202) Unification and maintenance of coding logic:

[0080] Step 202-1) Implement the new tree structure coding: In this step, the system will update and unify all relevant cost data according to the newly defined tree structure coding logic. This includes remapping existing data entries to the new coding structure and ensuring that new projects adopt this new coding. During this process, the system will provide tools and guidance to help users understand the structure of the new coding and correctly classify data into the appropriate coding level.

[0081] Step 202-2) Correspondence and version maintenance of new and old codes: In order to solve the data compatibility problems that may be caused by the new and old coding systems, this step will implement a new and old coding correspondence mechanism. The system will create a mapping table to record in detail the correspondence between the new and old codes, as well as their respective version numbers. In this way, even data input based on the old code can be correctly parsed and analyzed. The system will support the maintenance of the coding version to ensure that after each coding update, the old data can still be accurately analyzed through the mapping relationship, thereby avoiding data analysis errors or omissions caused by coding updates.

[0082] Step 203) Extract key cost items based on local weighted regression analysis of typical design: After the cost rules are created or modified, the system will use local weighted regression analysis of typical design to extract key cost items to verify the effectiveness and impact on project cost under the new rules and coding. This step determines how different input variables (such as specific rates or material costs) affect the final total project cost by carefully analyzing the data changes around the typical design points. This reveals which variables (such as specific rates or material costs) have the greatest impact on the total project cost, and records the results and codes with the greatest impact. The following are the specific steps:

[0083] Step 203-1) Typical design input by project type: In this step, the system first needs to collect typical design data of various project types from users or historical databases. This includes detailed cost structure, main materials and equipment usage, and other important cost drivers. The typical design of each project type should record its standard specifications and budget allocation in detail, and according to past data, the actual project is generally used as a benchmark for sensitivity analysis.

[0084] Step 203-2) Cost item offset: Based on the typical design, the system will simulate small-scale offsets to simulate possible design changes or cost fluctuations. These offsets will be set according to various situations that may be encountered in actual projects, such as rising material costs, changes in labor costs, etc. When setting the offset parameters, only small-scale offsets are performed, and the offsets are designed based on experience. The offsets that change frequently are set larger, such as a 5% increase in material costs, a 1% increase in labor costs, etc. The corresponding offset data set is generated by the system for further sensitivity analysis.

[0085] Step 203-3) Local weighted regression model construction: Use the local weighted regression (LOESS) model to analyze the data around the typical design and its offset point. Use the local weighted regression model to fit the impact of a small range of cost offsets based on the typical design on the overall cost. Assume that all parameters that affect the cost constitute the variable X = {x1, x2, x3, ...., x n}, the cost result is Y, and the LOESS method uses weighted least squares to fit a local polynomial regression model. Use local linearity to fit the cost fluctuations around the typical design under the overall nonlinear model, that is, Y≈β0(x)+β1(x)·X, where x represents the data point centered on the data set, which is the data point where the typical design is located in this case, and β0(x) and β1(x) are local constants and local slopes based on point x.

[0086] Step 203-4) Local slope solution: Extract local slopes β1(x) from the local weighted regression model. These slopes represent the sensitivity of the cost to various changes near each typical design point. Local slopes are key indicators for evaluating the impact of various design parameters. Solve local derivatives (slopes) for each LOESS model. These slopes directly reflect the sensitivity of the output cost to changes in the input variables. Analyze all calculated local slopes to determine which design parameters have a significant impact on the total cost, and integrate this information into a detailed sensitivity analysis report for decision makers.

[0087] Step 204) Configure cost codes by cost stage: In this step, the system will provide a mechanism that allows users to configure specific cost codes according to different project stages. For example, the preliminary design stage, the construction stage, and the final settlement stage may require different cost categories and details. This stage-by-stage coding configuration facilitates the subsequent preparation of cost documents for different stages. At the same time, as the project progresses and possible changes, the stage-to-cost correspondence rules can be dynamically adjusted to reflect the added costs or modified budgets.

[0088] The cost differentiation comparison process is as follows:

[0089] Step 301) Initialize the cost difference comparison module: The cost difference comparison module of the present invention can conduct in-depth comparison and analysis on structured cost data and identify cost differences in the entire project audit process. The module uses a preset coding system to classify and mark all relevant data for effective data tracking and comparison.

[0090] Step 302) Coding data matching and comparison: In this step, the system will match the structured cost data of each stage according to the defined coding logic. The corresponding comparison across coding versions can also be achieved through the new and old coding mapping table. By matching the coding of the project, the system can accurately compare the data at different time points or audit stages, such as preliminary budget, submitted settlement statement, approved settlement statement, etc. This process automatically identifies price differences and quantity differences and provides clear difference reports.

[0091] Step 303) Multi-dimensional difference analysis: Difference comparison is not limited to a single dimension. The system supports multi-dimensional data analysis, including cities, business departments, financial departments, and audit units. Users can select a specific analysis dimension, and the system will display the corresponding difference data based on the selected dimension. This multi-angle analysis helps auditors understand the efficiency and effectiveness of cost control from different levels.

[0092] Step 304) Generate detailed report of price difference and quantity difference: Once the system identifies a specific price difference or quantity difference, it will automatically generate a detailed difference report. The report includes the specific value of the difference, relevant influencing factors and possible cause analysis. These reports not only help the audit team quickly understand the problem, but also serve as a basis for adjusting budgets and improving project management.

[0093] The present invention also includes prejudgment of suspicious points, where the suspicious points are quota items with high data change frequency, i.e., frequent items. The mining steps of the frequent items are specifically as follows:

[0094] Step 401) Apply FP-growth algorithm to mine frequent item sets: construct a transaction data set, including those modules, materials, and quota combinations that have changed between the approval stage and the submission stage. The data filtered out in this step represents the instances of actual changes in the project, and these change instances will be used to generate the FP tree. Apply the FP-growth algorithm on the prepared transaction data set, focusing only on those modules, materials, and quota combinations that are inconsistent in the structured full-process cost documents. The purpose is to discover item sets that appear frequently and are prone to change in different projects. These frequent item sets reveal common problems or audit risk points that may exist in the audit process. Specifically, it includes the following sub-steps:

[0095] Step 401-1) Construct the initial FP tree: define the transaction in the present invention as the combination of modules, materials, and quotas to be audited and adjusted, and sort them in descending order of support. Sort each entry (module, material, quota) in the cost data set according to its frequency of occurrence. Starting from the root node, add entries to the tree in the sorted order. If the prefix of a path is the same as an existing path, reuse this path and increase the corresponding count; otherwise, create a new branch.

[0096] Step 401-2) Mining frequent item sets: For each item in the tree, create a "conditional pattern base" by tracing the links connecting all identical items. The conditional pattern base refers to the set of all paths ending with a certain item (such as a certain quota). These paths are all prefix paths from the root of the FP tree to the item. For each item in the FP tree, we first find all instances of the item in the tree, and then for each instance, trace its path to the root. These paths plus the count of each path (that is, the number of times the path appears, this information has been recorded when constructing the FP tree) constitute the conditional pattern base of the item. Use the conditional pattern base to construct a conditional FP tree, and recursively mine frequent item sets from it. These frequent item sets represent combinations of modules, materials, and quotas that are often adjusted together during the audit process.

[0097] Step 402) Analyze frequent item sets and generate association rules: Through the frequent item sets generated in the previous steps, we can further analyze these patterns and extract association rules. Specifically, it includes the following sub-steps:

[0098] Step 402-1) Extraction of association rules: For each frequent item set, analyze the possible associations between the items therein to form rules such as "If module X and material Y appear, the quota Z used for review is often adjusted." Calculate the support and confidence of each rule to confirm the effectiveness and reliability of these rules. Support: Indicates the frequency of occurrence of the item set in all transactions. Assume that the occurrence of module X and material Y is event A, and the quota Z used for review is adjusted as event B. For rule A→B (that is, if module X and material Y appear, the quota Z used for review is often adjusted), its support is defined as the ratio of A∪B in all cost data sets. Confidence indicates the probability of the occurrence of the consequent B under the condition that the antecedent A appears, that is, the probability of B occurring when it is known that A occurs. Calculated as Only those rules whose support and confidence are higher than the user-set threshold are retained to complete the association rule extraction.

[0099] Step 402-2) Application of rules: Apply these rules to new audit cases to predict and identify possible audit adjustment points. These rules help auditors focus on those audit issues that may require special attention.

[0100] Step 403) Rule verification and optimization: Apply these rules in the actual audit process and evaluate their performance based on the audit results. Update and adjust the support and confidence thresholds of the rules based on actual experience and feedback to optimize the accuracy and coverage of the rules.

[0101] The present invention can also prepare a settlement statement according to demand, and the specific steps are as follows:

[0102] Step 501) Initialization of cost data structuring: This step sets the foundation for the entire data structuring process, ensuring accurate data collection and effective storage. This includes setting up data collection and storage modules and defining structured templates for settlement statements.

[0103] Step 502) Design of preset template for settlement statement: According to the cost composition defined in step 101 and step 102 and the corresponding relationship between the cost stage and the cost cost code, the module decision tree is used to design the preset template for settlement statement of each stage. This template will include all the necessary cost data fields and equip each field with a unique code to ensure the uniformity and accuracy of the data.

[0104] Step 503) Utilize the module decision tree to form the module: User project features use the decision tree to input corresponding parameters, such as cable length, transformer capacity, etc., according to the leaf node reached (i.e., typical design). The data acquisition and storage module automatically calculates and outputs the required module types and quantities based on the user's input and the selected typical design. For example, the modules that may be required for direct burial cable laying include trenching, pipe laying, direct burial cable laying, etc. The system will provide a specific number of modules based on the cable length. Moreover, under each cost component, the user can also upload related attachments. These attachments may include, but are not limited to, contract texts, engineering change orders, acceptance records, and other related documents. The system provides a simple interface that allows users to directly upload and associate these files to the corresponding cost items, generate a master settlement statement, and ensure that all supporting documents are structured and stored and managed together with the data.

[0105] Step 504) Data integration of attachments: Once the attachments are uploaded, the system automatically extracts key information from the attachments and associates them with the relevant items in the main statement. This step enhances the integrity and tracking capabilities of the data, facilitating subsequent audits and data verification. In addition, the system is able to detect and mark any possible inconsistencies or missing information to ensure the accuracy and completeness of the statement.

[0106] In the present invention, it is also necessary to optimize the decision tree, and the specific steps are as follows:

[0107] Step 503-1) Decision tree selection data statistics: Due to the use of binary feature division and the training of actual decision tree projects, users should not stay too long, make mistakes, or choose a path with a lower probability. According to the actual path selection data of users in the process of using the decision tree, special attention is paid to those paths that eventually lead users to return to the previous level. The possible stop points of users in the process of using the decision tree are counted, and the number of times all paths are selected is recorded to record the successful path selection. The events such as selecting a less used path, staying for more than 1 minute, and returning to the previous level in the path are regarded as abnormal events.

[0108] Step 503-2) Continuous optimization of decision tree: When the frequency of abnormal events exceeds 50%, we believe that it seriously affects the user selection experience and needs to retrain the CART decision tree model. The training process is the same as step 103, and the training data brought in is the original training plus the actual decision tree usage data.

[0109] In the present invention, the steps of searching for cost data in a preset format in step 502 are as follows:

[0110] Step 502-1) Define and implement data matching rules: Set a series of data matching rules to first determine the current cost stage and automatically identify and extract cost data in the settlement statement and related documents. These rules are based on text matching technology and can identify the codes and corresponding data items and specific data in the document.

[0111] Step 502-2) Automatic data extraction and matching: The system will automatically scan the uploaded documents, such as settlement statements, contracts, etc., and search and extract relevant cost data according to the preset data matching rules. The system uses OCR technology and text analysis tools to identify and extract data from documents of various formats.

[0112] Step 502-3) Handling of abnormal and missing data: During the data matching process, if the system cannot find cost data that matches the preset code in the document, a prompt mechanism will be triggered. This includes issuing a warning to the user, requesting more information or manually entering the missing data. At the same time, the system will record all instances of failure to match for subsequent review and analysis.

[0113] After the present invention completes the preparation of cost documents at different stages and the uploading of attachments, the next step is to upload the structured cost documents at different stages to the cloud server. This step not only ensures the secure storage of data, but also facilitates future access and auditing. During the upload process, all cost document data at different stages and related attachments will be encrypted using MD5 to protect the security and privacy of information. Encryption ensures security during data transmission and prevents unauthorized access and data leakage.

[0114] The cost documents and all related attachments of the structured different stages of the present invention are stored on the cloud server, which not only provides a high degree of data security and reliability, but also enables the data to be quickly accessed between different locations and different devices. The flexibility and scalability of cloud storage ensure that the system can handle the data needs of small to large projects. The cloud server provides automatic data backup functions to ensure that data will not be lost under any circumstances. In addition, the fast data recovery function ensures that data can be quickly restored when it is accidentally deleted or damaged, ensuring business continuity. The cloud server also provides detailed access control and audit tracking functions, allowing managers to set who can access which data and track all access and modification operations to the data, further enhancing data security and compliance.

[0115] The present invention uses the data acquisition storage module to perform the settlement audit process steps as follows:

[0116] Step 601) Initialize the settlement audit process: In this step, the audit team begins to conduct a detailed review of the cost documents at different stages. The audit process includes verifying the various costs and data listed in the cost documents at different stages to confirm their accuracy and compliance.

[0117] Step 602) Missing attachment reminder mechanism during the audit process: During the preparation of cost documents at different stages, specific cost components often require attachments as supporting evidence, such as contract modifications, approval documents for additional engineering quantities, etc. The system will set up an automatic detection mechanism to identify those cost components that usually require attachments but lack corresponding attachments in the current cost documents submitted at different stages.

[0118] Step 603) Determination of suspicious quota items during the audit process

[0119] Step 603-1) suspicious quota item marking: Using the FP-growth rules optimized by steps 401)-403), the system analyzes the new project data and automatically identifies quota items that meet frequent patterns. This includes paying special attention to quota items that appear in the rules but have not been adjusted during the review process and marking them as suspicious quota items.

[0120] Step 603-2) Suspicious point report and audit intervention: The system will compile the automatically marked suspicious points into a report and provide it to the audit review team. Auditors will conduct more in-depth manual inspection and analysis based on this information to confirm whether there are any audit or cost issues.

[0121] Step 603-3) Feedback collection and system iteration and update: Collect feedback from the audit review team to evaluate the accuracy and practicality of the automatic identification system for suspicious quota items. Based on new data and feedback collected in actual applications, continuously iterate and update rules and algorithms to improve the overall performance and accuracy of the system.

[0122] Step 603-4) Overview of suspicious quota items: After completing all the above audit steps, the system will provide a data overview and a comprehensive audit result display. This step is the final stage of the audit process. The overview will detail which modules, materials and quota combinations are frequently adjusted during the audit. The purpose of listing suspicious quota items is to find supporting evidence or to hold the audit quality accountable, especially those quota combinations that are frequently adjusted but not properly adjusted in the project.

[0123] Step 604) Modify and supplement the settlement statement submitted for review: Based on the audit findings, the settlement statement submitted for review may need to be modified and supplemented. The auditor will modify the settlement data based on the issues found, and may require the business department to submit necessary attachments, such as newly added contract documents, change orders or other supporting documents to prove the rationality and accuracy of the modification.

[0124] Step 605) Form a structured verification statement: The revised statement and all newly uploaded attachments will be integrated into a structured verification statement. This document will include all original data and its modification records to ensure the integrity and traceability of the verification statement.

[0125] Step 606) Record the modification content: During the settlement audit process, all modifications and updates will be recorded in detail. This includes recording the reason for each modification, the amount involved, the relevant code and any relevant attachments. These records are an important part of the difference item data record and provide key information for subsequent data analysis and mining.

[0126] Step 607) Recording and mining of difference data: Using the recorded modification content, the system will mine the difference data. This process involves analyzing the pattern and frequency of modification and identifying possible trends or anomalies, which is of great significance for optimizing future audit strategies and improving the settlement process.

[0127] Step 608) Upload to the cloud server: The final audited settlement statement and all its related records and attachments will be uploaded to the cloud server. Cloud storage not only ensures the security and accessibility of data, but also facilitates cross-regional project team access and collaboration, while providing powerful data backup and disaster recovery capabilities.

[0128] The present invention also relates to an audit system for the audit process of a distribution network construction project, comprising:

[0129] Data entry module: used to enter actual cost documents;

[0130] Data conversion module: used to identify and encode cost documents;

[0131] A data generation module is used to generate construction cost budget data;

[0132] Data comparison module, used to compare construction cost budget data with cost document data;

[0133] Report generation module, used to generate the final differential report.

[0134] The present invention also relates to a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and is characterized in that the memory stores the computer program, and when the processor executes the computer program, it implements any of the aforementioned methods for evaluating the audit quality of the distribution network construction project audit process.

[0135] The present invention also relates to a computer-readable storage medium having a computer program stored thereon, wherein the storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute any of the aforementioned methods for evaluating the audit quality of the distribution network construction project audit process.

Claims

1. A method for evaluating the audit quality of a distribution network construction project audit process, characterized in that: The steps include: 1) Collect and compile a list of all materials and equipment required during the construction of the entire distribution network project, and encode all materials and equipment used; 2) Generate construction cost budget data according to project requirements; 3) Identify the actual construction cost documents, extract typical data from the cost documents, match the typical data with the list in step 1, and then encode the typical data; 4) Compare and analyze the coded typical data construction cost budget data to identify the differences between the two; 5) Summarize the differences and generate a difference report.

2. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 1 is characterized in that: In step 1), all distribution network construction projects are divided into several decision tree links, each of which is composed of a set of typical power grid design data, which constructs a standardized module cost data with a three-level cross-reference relationship of module-material-quota.

3. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 2 is characterized in that: The typical design data features of the power grid include upper-level data, lower-level data and environmental data. The upper-level data include pure civil engineering project type data, pole erection project type data, overhead line laying project type data, cable laying project type data, electrical equipment project type data and branch box construction type data. The lower-level data is the specific construction project of each upper-level data. The environmental data includes indoor, open-air, mine and underwater. The upper-level data, lower-level data and environmental data of the same group are combined to form several decision tree links, and each decision tree link corresponds to an actual distribution network construction project.

4. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 3 is characterized by: The lower-level data include the core power distribution equipment and related auxiliary construction materials required for the project. The equipment and auxiliary construction materials are set with initial values, which are the quantities of equipment and materials used under normal circumstances.

5. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 3 is characterized by: The decision tree link formed in step 1) is optimized. The specific steps are as follows: 1-1) Divide the decision tree data features into binary and obtain several groups of binary reflexive features; 1-2) Calculate the Gini impurity of each feature and perform weighted average calculation on it; 1-3) Select the segmentation feature with the smallest Gini impurity among all features as the best segmentation point; 1-4) Use the best split point to divide the data set into two subsets, and repeat steps 1-2) and 1-3) for each subset until the decision tree data node reaches the maximum depth.

6. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 5 is characterized by: In step 1-2), the weighted average calculation formula is as follows: Among them Gini split1 Its representative features 1 is the Gini impurity of the split point, n is the total number of samples, n Y , n N The number of samples with and without this feature, Gini Y With Gini N The Gini impurity with and without this feature respectively; Gini is the Gini impurity, and its function is In the formula, C represents the classification set, which is all typical designs in the present invention, i represents the i-th classification, and p(i) is the i-th classification frequency of the training sample.

7. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 1 is characterized in that: Step 2) specifically includes the following steps: 2-1) According to the actual construction plan, determine the corresponding complete decision tree link, and obtain the initial cost project data from the complete decision tree link; 2-2) offset the cost item data one by one to generate corresponding offset data sets; 2-3) Analyze the initial data and the data around the offset point using a local weighted regression model, and solve for the local slope; 2-4) Analyze the impact of design parameters on total cost based on local slopes and integrate relevant information into a report.

8. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 7 is characterized in that: The local weighted regression model is: Y≈β0(x)+β1(x)·X, where Y is the cost result, X is the variable composed of all parameters affecting the cost, X={x1,x2,x3,....,x n }, β0(x) and β1(x) are the local constant and local slope based on point x.

9. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 1 is characterized in that: In step 3), several coding mapping tables are used to identify the actual cost documents to ensure that all data can be correctly and completely converted.

10. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 1 is characterized in that: The differences mentioned include missing attachments to the cost document, abnormal number of adjustments to the cost document data, and inconsistency between the cost document data and the construction cost budget data.

11. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 1 is characterized in that: Step 4) also includes prejudgment of suspicious points, where the suspicious points are quota items with high data change frequency, i.e., frequent items. The mining steps of the frequent items are as follows: 4-1) Construct an initial FP tree and mine frequent itemsets from it; 4-2) Analyze frequent item sets and extract association rules.

12. The audit quality evaluation method for the audit process of a distribution network construction project according to claim 11, characterized in that: The initial FP tree construction steps are as follows: First, find the combination of modules, materials, and quotas that are different from the budget data, and sort them in descending order of support. Then sort each module, material, and quota according to its frequency of occurrence, and then start from the root node and add entries to the tree in the sorted order. If the prefix of a path is the same as an existing path, reuse this path and increase the corresponding count; otherwise, create a new branch, and finally form the initial FP tree; The association rule extraction method is as follows: For each frequent item set, the possible associations between the items are analyzed. If the quota changes under the condition that the specified modules and materials appear, the data set is recorded as a rule, and then the support and confidence of each rule are calculated. The support is the proportion of the rule in all cost data sets, and the confidence is the probability of the rule occurring.

13. An audit quality evaluation system for the audit process of a distribution network construction project, characterized in that: include: Data entry module: used to enter actual cost documents; Data conversion module: used to identify and encode cost documents; A data generation module is used to generate construction cost budget data; Data comparison module, used to compare construction cost budget data with cost document data; Report generation module, used to generate the final differential report.

14. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program, and is characterized in that the memory stores the computer program, and when the processor executes the computer program, it implements the audit quality evaluation method for the audit process of a distribution network construction project described in any one of claims 1 to 12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: The storage medium includes a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute an audit quality evaluation method for a distribution network construction project audit process according to any one of claims 1 to 12.

Images