Intelligent report filling system and method based on machine account association

Abstract

The invention provides an intelligent report filling system and method based on machine account association, and belongs to the technical field of enterprise data association, and the system comprises the steps that a machine account management module obtains machine account data from each source machine account system, and stores the machine account data in a machine account database; the intelligent filling engine module queries and associates the corresponding machine account data from the machine account database to generate report pre-filling data; the multi-stage data verification module sequentially executes format verification, business logic verification and fluctuation anomaly detection on the report pre-filling data, and outputs the verified report data; the data closed-loop management module receives the final report data, identifies increment change between the final report data and the source machine account data through a difference detection algorithm, synchronizes the change data back to a corresponding source machine account system, and generates an operation log; and visually configuring interface definition and configuring a field mapping relation, a verification rule and a backflow rule. According to the method, the machine account filling report is automatically associated, the difference is written back and traces are reserved, and errors are reduced.

Description

Technical Field

[0001] This application belongs to the field of enterprise data association technology, specifically involving an intelligent report filling system and method based on ledger association. Background Technology

[0002] In enterprise data management, report filling is a core, periodic operation in business scenarios such as finance, supply chain, and human resources. Traditional report filling methods mainly rely on manual data entry, which is prone to errors and omissions, resulting in low efficiency. Some existing technologies use primary keys (such as ID or code) to extract simple fields (such as name or date) from the database, but they cannot handle complex business logic or unstructured data mapping. Existing technologies also use fixed template matching methods, such as predefined report templates provided by some ERP systems, with fixed field association rules that cannot flexibly adapt to business changes.

[0003] Existing technologies suffer from the following drawbacks. First, data silos exist. Ledger data (such as financial records and inventory records) is separated from report data, lacking dynamic correlation. This necessitates manual cross-system queries and verification, increasing workload and error risk. For example, while related patents implement automatic report filling, they only support fixed template matching and cannot achieve cross-system data correlation. Second, mapping flexibility is insufficient. Existing technologies rely on fixed field name matching, which cannot adapt to semantic differences between heterogeneous systems, such as synonyms with different names, preventing automatic field correlation. Finally, automatic correction is lacking. When automatic filling errors occur, the system typically only provides a simple error message, lacking intelligent recommendations or user behavior learning functions, and cannot optimize candidate filling strategies.

[0004] Therefore, there is an urgent need for a method that can intelligently link ledger and report data, support dynamic mapping, be self-learning, and achieve closed-loop data management. Summary of the Invention

[0005] In a first aspect, embodiments of this application provide an intelligent report filling system based on ledger association, including: The ledger management module is used to obtain structured or unstructured ledger data from various source ledger systems through data interfaces and store it in the ledger database. The intelligent data entry engine module is used to query and associate the corresponding ledger data from the ledger database based on the field mapping relationship pre-configured by the user, and generate pre-filled data for reports. The multi-level data verification module is used to sequentially perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report according to the user's pre-configured verification rules, and output the verified report data with anomaly markers and correction suggestions to the user. The data closed-loop management module is used to receive the final report data confirmed by the user, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the user's pre-configured backflow rules, while generating an operation log containing the changed content and version hash value. A visual configuration interface provides users with an interactive interface for defining and configuring field mapping relationships, validation rules, and reflow rules.

[0006] Furthermore, the intelligent data entry engine module includes a dynamic field mapping unit and a context-aware controller; The dynamic field mapping unit includes: The semantic parser receives the original ledger field names configured by the user or from the source ledger system, converts them into ledger semantic vectors through a semantic vectorization model, and calculates the semantic similarity between the ledger semantic vectors and the semantic vectors of the target report fields. The thesaurus is used to store pairs of fields that have the same semantics but different names across different business systems, providing auxiliary mapping for the semantic parser; The fuzzy matching unit is used to receive keywords or regular expressions input by the user when the semantic similarity is below a threshold, and to perform a fuzzy query in the field list of the source ledger system to establish a mapping. The visual drag-and-drop component is used to display a list of source ledger fields and target report fields to users, and to receive field relationships established by users through drag-and-drop operations, and generate mapping rules. The context-aware controller includes: The behavior collection component is used to listen to and record users' correction operations on pre-filled data in real time in the data entry workbench, and generate user behavior logs. The strategy optimization component is used to receive and analyze the user behavior logs, and dynamically adjust and optimize the population strategy based on the analysis results. The real-time recommendation component is used to retrieve relevant data from the ledger database based on the context of the field currently being operated on, and to provide users with real-time recommendations for fill values; The semantic parser further includes: The field segmentation component is used to receive the original field names from the source ledger system or the target report, perform word segmentation, and output a word sequence composed of basic word units. The semantic vectorization component is used to receive word segmentation sequences, convert each word unit and contextual relationship into a numerical vector of a preset dimension through a pre-trained semantic model, and finally aggregate them into a field semantic vector that represents the overall semantics of the corresponding field. The similarity calculation component receives the semantic vectors of the source ledger field and the target report field, calculates the cosine similarity between the two vectors using a vector space model, and outputs a quantified similarity score.

[0007] Furthermore, the multi-level data verification module includes: The basic validation component is used to receive pre-filled report data, perform validation according to user-configured format validation rules, and output preliminary validation results. The business validation component receives data that has passed the initial validation and calls logical validation rules from the business rule base to perform deep validation. The intelligent early warning component is used to receive data that has passed deep verification, calculate the data volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed preset thresholds. The business verification component further includes: The associated verification unit, when a field value is marked as abnormal, calls entity data of the same type and scale from the ledger database, calculates the reasonable value range of the corresponding field value based on the called entity data, and verifies the reasonableness of the current field value. A logic rule base is used to store and maintain business logic validation rules; The anomaly marking unit is used to uniformly add visual markings and textual correction suggestions to all abnormal data found during the verification process.

[0008] Furthermore, the data closed-loop management module includes: The difference detection component is used to compare the report field values ​​in the final report data with the source data of the corresponding report fields in the ledger database during the feedback process, and to identify the fields that have changed and the old and new values. The backflow rule base is used to store data synchronization strategies, including full update, incremental update, or update rules triggered based on preset conditions. The version control component is used to generate an operation log containing change elements such as operation time, operator, changed field, old value, and new value when data changes, and to generate a unique version identifier for each change set; The version control component further includes: The operation log recording unit is used to capture and persistently store data operation flow data; A version identifier generator is used to generate a globally unique version identifier for each data change based on the content of the operation log using a hash algorithm. The change tracking interface provides users with a query function, which can display a complete history of data changes based on version identifiers or time ranges.

[0009] Furthermore, the visual configuration interface includes: The field mapping configuration interface is used to display the field list of the source ledger system and the field list of the target report, and provides configuration entry points for semantic parsing, synonyms, fuzzy matching and drag-and-drop association, and finally generates a visual field mapping relationship diagram. The rules settings interface is used to configure format validation rules, business validation logic rules, and data return rules; The workbench interface is used to display pre-filled reports to users, highlight fields that require manual entry or verification, and output exception markers, correction suggestions, and real-time fill value recommendations.

[0010] Secondly, this application also provides a method for intelligent report filling based on ledger association, including the following steps: S1. Obtain structured and unstructured ledger data from various source ledger systems through data interfaces, and store the ledger data in the ledger database after cleaning and standardization. S2. Based on the field mapping relationship predefined by the user through the visual configuration interface, query and associate the corresponding ledger data from the ledger database, and automatically generate pre-filled report data; S3. Based on the verification rules predefined by the user through the visual configuration interface, perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report in sequence, and output the verified report data with anomaly markers and correction suggestions to the user; S4. Receive the final report data after user confirmation and correction, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the pre-configured backflow rules. At the same time, generate an operation log containing the changed content and version hash value.

[0011] Furthermore, the specific steps of step S2 are as follows: S21. Receive and process user-defined field mapping relationships through a visual configuration interface; S22. Based on the established mapping relationship, query the corresponding ledger data from the ledger database, automatically fill the corresponding fields of the target report, and generate pre-filled data for the report; S23. Monitor and record users' correction operations on the pre-filled data in the data entry workbench in real time, and generate user behavior logs; S24. Receive and analyze user behavior logs, and optimize the filling strategy using a pre-trained machine learning model; wherein, the machine learning model is trained using the gradient descent algorithm to minimize the error function:

[0012] Where, is the sample size. Correct the value for the user. These are the model's predicted values; S25. Based on the context of the field currently being operated on, retrieve relevant data from the ledger database and provide the user with real-time populate value recommendations.

[0013] Furthermore, the user-defined field mapping relationship in step S21 is established through at least one of the following methods: S211. Receive the original field names from the source ledger system or the target report, convert the original field names into field semantic vectors using a semantic vectorization model, and calculate the semantic similarity between the semantic vectors of the source ledger fields and the semantic vectors of the target report fields; wherein, the cosine similarity calculation formula is:

[0014] in, A semantic vector representing the source ledger field. This represents the semantic vector of the target report field, and a mapping is automatically established when the similarity exceeds a preset threshold. S212. Query the pre-built thesaurus containing semantically identical but differently named field pairs across different business systems to assist in establishing mappings; S213. When the semantic similarity is below the threshold, receive keywords or regular expressions input by the user and perform a fuzzy search in the field list of the source ledger system to establish a mapping; S214. Display a list of source ledger fields and target report fields to the user, and receive field associations directly established by the user through drag-and-drop operations, and generate mapping rules.

[0015] Furthermore, the specific steps of step S3 are as follows: S31. Receive pre-filled report data, perform validation according to user-configured format validation rules, and output preliminary validation results; S32. Receive the data that has passed the initial verification and call the logical verification rules in the business rule base to perform deep verification; S33. Receive data that has passed deep verification, calculate the volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed a preset threshold; wherein, the volatility calculation formula is:

[0016] When volatility exceeds the volatility threshold, it is marked as abnormal.

[0017] Furthermore, step S32 also includes the following association verification steps: When a field value is marked as abnormal, other entity data of the same type and scale are retrieved from the ledger database to calculate the reasonable value range of the field to verify the reasonableness of the current value; wherein, reasonableness verification is performed by calculating the deviation rate:

[0018] The average value of the same type is calculated using the K-means clustering algorithm. When the deviation rate exceeds the deviation rate threshold (e.g., 20%), the value is confirmed to be abnormal.

[0019] Furthermore, the specific steps of step S4 are as follows: S41. Compare the field values ​​in the final report data with the corresponding source data in the ledger database to identify the fields that have changed and the old and new values; the difference detection is based on the hash comparison of the field values, and the hash value calculation uses the SHA-256 algorithm: Hash value = SHA - 256 (field value) When the hash value of the source ledger is inconsistent with the hash value of the report, it is determined to be a changed field; S42. Perform full update, incremental update or condition-triggered synchronization on the changed data according to the pre-configured backflow rules; S43. When data changes, generate an operation log containing the operation time, operator, changed field, old value, and new value, and generate a unique version identifier for each change set; the version identifier is generated using the MD5 algorithm. Version hash = MD5(operation type + field name + old value + new value + timestamp).

[0020] As can be seen from the above technical solutions, this application has the following advantages: The intelligent report filling system and method based on ledger association provided in this application periodically extracts ledger data from various business systems through a standard data interface, cleans and transforms it, and then centrally stores it in the ledger database. Based on pre-configured field mapping relationships, it automatically generates pre-filled report data, reducing manual data entry. Through three layers of validation—format, business logic, and historical fluctuations—it alerts to anomalies and provides correction suggestions. After user confirmation, the system compares the final report with the source ledger, writes the changes back to the source system according to the feedback rules, and generates an immutable operation log using hashing. This reduces the workload of manual copying and verification, maintains consistency between ledgers and reports, provides traceable change records, and meets audit requirements. Attached Figure Description

[0021] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the intelligent report filling system based on ledger association of the present invention.

[0023] Figure 2 This is a flowchart illustrating the intelligent report filling method based on ledger association of the present invention. Detailed Implementation

[0024] The following detailed description of the intelligent report filling system based on ledger association will provide a more comprehensive overview of various embodiments of this disclosure. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0025] This embodiment provides an intelligent report filling system based on ledger association. It automatically associates multi-source ledgers according to preset mapping to generate pre-filled reports. The reports undergo three levels of verification: format, business logic, and historical fluctuations, and anomalies are detected. After user confirmation, the differences are written back to the source system, and a unique version log is generated using hashing. This reduces manual copying, verification, and writing back, lowers the risk of data inconsistency, and meets audit traceability requirements.

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figure 1 The diagram shown is a schematic of a smart report filling system based on ledger association in a specific embodiment. The system includes: The ledger management module is used to obtain structured or unstructured ledger data from various source ledger systems through data interfaces and store it in the ledger database. It should be noted that the ledger management module provides a unified data interface to reduce redundant development; it parses structured and unstructured ledger data and stores it in the ledger database to form a unique data source, avoiding inconsistencies between multiple sources. The intelligent data entry engine module is used to query and associate the corresponding ledger data from the ledger database based on the field mapping relationship pre-configured by the user, and generate pre-filled data for reports. It should be noted that the intelligent data entry engine module automatically queries the ledger and fills the report based on the field mapping relationship, saving manual copying and pasting, and supports batch processing. Reports with thousands of rows can be pre-filled in seconds. The multi-level data verification module is used to sequentially perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report according to the user's pre-configured verification rules, and output the verified report data with anomaly markers and correction suggestions to the user. It should be noted that format validation checks data type, length, and required fields to prevent basic errors; business logic validation calls preset rules to verify numerical range and consistency, reducing business errors; and fluctuation anomaly detection compares with historical data from the same period to automatically mark indicators that deviate from the threshold and prompt users to review them. The data closed-loop management module is used to receive the final report data confirmed by the user, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the user's pre-configured backflow rules, while generating an operation log containing the changed content and version hash value. It should be noted that the difference detection uses field-level hash comparison, which can identify any character change; the reflow rules support full, incremental and condition-triggered updates to adapt to different system performance requirements; the operation log records the time, operator, old and new field values ​​and version hash, and supports quick query by time or version number to meet audit and internal control requirements; A visual configuration interface provides users with an interactive interface for defining and configuring field mapping relationships, validation rules, and reflow rules; It should be noted that maintaining mapping, validation, and reflow parameters through a web page requires no coding; import and export functions are provided to facilitate configuration migration between test and production environments.

[0028] This embodiment aggregates multi-source ledgers through an interface, automatically generates reports with pre-filled data according to preset mappings, and writes back the differences after format, business, and fluctuation verification to generate an unalterable log, reducing the workload of manual data entry and verification, and keeping the ledgers and reports consistent.

[0029] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, another intelligent report filling system based on ledger association is provided, which includes: The ledger management module is used to obtain structured or unstructured ledger data from various source ledger systems through data interfaces and store it in the ledger database. For example, a manufacturing company has three source ledger systems: SAP financial system (structured data), WMS inventory system (structured data), and Excel manual ledger (unstructured data). The ledger management module connects to SAP and WMS through a JDBC interface, reads the manual ledger through the Excel parsing interface, and cleans data such as "accounts receivable", "inventory quantity", and "raw material loss record" into a unified "field name-value-unit-timestamp" format and stores it in the ledger database. The handwritten notes field in Excel is standardized and stored after being recognized by OCR. The intelligent data entry engine module is used to query and associate the corresponding ledger data from the ledger database based on the field mapping relationship pre-configured by the user, and generate pre-filled data for reports. The intelligent data entry engine module includes a dynamic field mapping unit and a context-aware controller; The dynamic field mapping unit includes: The semantic parser receives the original ledger field names configured by the user or from the source ledger system, converts them into ledger semantic vectors through a semantic vectorization model, and calculates the semantic similarity between the ledger semantic vectors and the semantic vectors of the target report fields. The thesaurus is used to store pairs of fields that have the same semantics but different names across different business systems, providing auxiliary mapping for the semantic parser; The fuzzy matching unit is used to receive keywords or regular expressions input by the user when the semantic similarity is below a threshold, and to perform a fuzzy query in the field list of the source ledger system to establish a mapping. The visual drag-and-drop component is used to display a list of source ledger fields and target report fields to users, and to receive field relationships established by users through drag-and-drop operations, and generate mapping rules. The context-aware controller includes: The behavior collection component is used to listen to and record users' correction operations on pre-filled data in real time in the data entry workbench, and generate user behavior logs. The strategy optimization component is used to receive and analyze the user behavior logs, and dynamically adjust and optimize the population strategy based on the analysis results. The real-time recommendation component is used to retrieve relevant data from the ledger database based on the context of the field currently being operated on, and to provide users with real-time recommendations for fill values; The semantic parser further includes: The field segmentation component is used to receive the original field names from the source ledger system or the target report, perform word segmentation, and output a word sequence composed of basic word units. The semantic vectorization component is used to receive word segmentation sequences, convert each word unit and contextual relationship into a numerical vector of a preset dimension through a pre-trained semantic model, and finally aggregate them into a field semantic vector that represents the overall semantics of the corresponding field. The similarity calculation component receives the semantic vectors of the source ledger field and the target report field, calculates the cosine similarity between the two vectors using a vector space model, and outputs a quantified similarity score. For example, users need to generate a "Monthly Production and Operation Report" and configure mapping relationships through a visual interface: the "Main Business Revenue" in SAP is automatically associated with the "Operating Revenue" in the report through semantic parsing (cosine similarity 0.92 > threshold 0.8), the "Inventory Turnover Days" in WMS is associated with the "Inventory Turnover Cycle" in the report through thesaurus matching, and the "Equipment Maintenance Costs" in the Excel manual ledger are mapped with the "Operation and Maintenance Expenses" in the report through drag-and-drop operations; when filling in the report, the system extracts the main business revenue of 12 million yuan, inventory turnover days of 28 days, and equipment maintenance costs of 850,000 yuan for March from the ledger database and automatically fills them into the corresponding fields of the report; at the same time, the behavior collection component records the user's operation of correcting the "Operation and Maintenance Expenses" to 880,000 yuan, and the strategy optimization component will prioritize recommending ledger data close to this value when filling in the report later; The multi-level data verification module is used to sequentially perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report according to the user's pre-configured verification rules, and output the verified report data with anomaly markers and correction suggestions to the user. The multi-level data verification module includes: The basic validation component is used to receive pre-filled report data, perform validation according to user-configured format validation rules (such as data type, length, and required fields), and output preliminary validation results. The business validation component receives data that has passed the initial validation and calls logical validation rules (such as numerical range and cross-table reconciliation relationships) in the business rule base to perform in-depth validation. The intelligent early warning component is used to receive data that has passed deep verification, calculate the data volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed preset thresholds. The business verification component further includes: The associated verification unit, when a field value is marked as abnormal, calls entity data of the same type and scale from the ledger database, calculates the reasonable value range of the corresponding field value based on the called entity data, and verifies the reasonableness of the current field value. A logic rule base is used to store and maintain business logic validation rules; The anomaly marking unit is used to uniformly add visual markings and textual correction suggestions to all abnormal data found during the verification process; For example, after the report is pre-filled, the basic validation component detects that the "Taxes Payable" field is formatted as text (it should be a number) and directly marks it as abnormal; the business validation component calls the "VAT = Output Tax - Input Tax" rule in the rule base and finds that the pre-filled VAT of 560,000 yuan does not match the calculated output tax of 2.1 million yuan and input tax of 1.4 million yuan, and adds the text suggestion "It is recommended to check the accuracy of the output / input tax entry"; the intelligent early warning component queries historical data and finds that the fluctuation rate of "Raw Material Procurement Cost" of 8 million yuan in March compared with 5 million yuan in the same period last year is 60% (exceeding the ±30% threshold), automatically marks it and prompts "The procurement cost fluctuation is abnormal, and it is necessary to verify whether there is a batch procurement situation"; the correlation validation unit calls the procurement data of 3 companies of the same size in the same industry, calculates the average value of the same type through K-means clustering, which is 5.5 million yuan, and the current value deviation rate is 45.45% > 20%, confirming the abnormality; The data closed-loop management module is used to receive the final report data confirmed by the user, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the user's pre-configured backflow rules, while generating an operation log containing the changed content and version hash value. The data closed-loop management module includes: The difference detection component is used to compare the report field values ​​in the final report data with the source data of the corresponding report fields in the ledger database during the feedback process, and to identify the fields that have changed and the old and new values. The backflow rule base is used to store data synchronization strategies, including full update, incremental update, or update rules triggered based on preset conditions. The version control component is used to generate an operation log containing change elements such as operation time, operator, changed field, old value, and new value when data changes, and to generate a unique version identifier for each change set; The version control component further includes: The operation log recording unit is used to capture and persistently store data operation flow data; A version identifier generator is used to generate a globally unique version identifier for each data change based on the content of the operation log using a hash algorithm. The change tracking interface provides users with a query function, which can display the complete history of data changes based on version identifiers or time ranges. For example, after the user confirms the report data, the difference detection component calculates the field hash value using the SHA-256 algorithm. It finds that the hash values ​​of "Operation and Maintenance Costs" and "Raw Material Procurement Costs" are inconsistent with the source ledger data, identifying them as changed fields. According to the pre-configured incremental update rules, only the new values ​​(880,000 yuan and 8,000,000 yuan) of these two fields are synchronized back to the WMS system and the manual Excel ledger. Simultaneously, the version control component generates an operation log: Operation time 2024-04-05 10:30, operator Zhang San, changed fields "Operation and Maintenance Costs" (old value 850,000 yuan / new value 880,000 yuan) and "Raw Material Procurement Costs" (old value 7,500,000 yuan / new value 8,000,000 yuan), and generated a version hash "e3f2d1c0b9a876543210fedcba987654" using the MD5 algorithm. The user can query the complete change record for this version through the change traceability interface. A visual configuration interface provides users with an interactive interface for defining and configuring field mapping relationships, validation rules, and reflow rules; The visual configuration interface includes: The field mapping configuration interface is used to display the field list of the source ledger system and the field list of the target report, and provides configuration entry points for semantic parsing, synonyms, fuzzy matching and drag-and-drop association, and finally generates a visual field mapping relationship diagram. The rules settings interface is used to configure format validation rules, business validation logic rules, and data return rules; The workbench interface is used to display pre-filled reports to users, highlight fields that need to be manually entered or verified, and output anomaly markers, correction suggestions, and real-time fill value recommendations. The field mapping configuration interface displays the field lists of SAP, WMS, and Excel on the left and the target report field list on the right. Users can click the "Semantic Analysis" button to see field matching suggestions, and click "Synonyms Library" to add related fields such as "Amount Excluding Tax - Amount Before Tax". Manual mapping can be completed by dragging "Equipment Maintenance Cost" to "Operation and Maintenance Costs", and the interface automatically generates a tree-shaped mapping relationship diagram. The rule setting interface provides a drop-down box to select the format validation type (such as numeric / text / date), supports custom business rules (such as "Gross Profit Margin ≥ 10%)", and the return flow rule can select "Update only changed fields" (incremental update). The workbench interface highlights three abnormal fields: "Taxes Payable", "Value Added Tax", and "Raw Material Procurement Cost", marked with a red border. Correction suggestions and real-time recommendations such as "Reference value of procurement cost for similar enterprises: 5.5 million yuan" are displayed below.

[0030] like Figure 2As shown, the following are embodiments of the intelligent report filling method based on ledger association provided in this disclosure. This method and the intelligent report filling system based on ledger association in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the intelligent report filling method based on ledger association, please refer to the embodiments of the intelligent report filling system based on ledger association described above.

[0031] The method includes the following steps: S1. Obtain structured and unstructured ledger data from various source ledger systems through data interfaces, and store the ledger data in the ledger database after cleaning and standardization. It should be noted that data is captured periodically through a standard interface, reducing manual export; and consistent field naming, date format, and units prevent subsequent mapping errors. S2. Based on the field mapping relationship predefined by the user through the visual configuration interface, query and associate the corresponding ledger data from the ledger database, and automatically generate pre-filled report data; It should be noted that the data is extracted from the ledger database in one go according to the preset mapping, reducing the need for line-by-line entry; S3. Based on the verification rules predefined by the user through the visual configuration interface, perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report in sequence, and output the verified report data with anomaly markers and correction suggestions to the user; It should be noted that the system provides real-time prompts for format validation, allowing users to correct errors on the spot; it sorts the results by error level through background business validation, prioritizing the display of serious anomalies; and it provides multi-dimensional comparisons of year-on-year, month-on-month, and moving average data through fluctuation detection, helping to pinpoint the causes of anomalies. S4. Receive the final report data after user confirmation and correction, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the pre-configured backflow rules, while generating an operation log containing the changed content and version hash value. It should be noted that by packaging and sending back only the changed fields, the load on the source system is reduced; version hashing ensures the uniqueness of each commit and supports rollback to any historical version; and operation logs are centrally stored and can be retrieved in multiple dimensions by user, time period, and field, meeting compliance auditing requirements.

[0032] This embodiment aggregates multi-source ledgers through an interface, automatically generates reports with pre-filled data according to preset mappings, and writes back the differences after format, business, and fluctuation verification to generate an unalterable log, reducing the workload of manual data entry and verification, and keeping the ledgers and reports consistent.

[0033] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, another intelligent report filling method based on ledger association is provided, which includes the following steps: S1. Obtain structured and unstructured ledger data from various source ledger systems through data interfaces, and store the ledger data in the ledger database after cleaning and standardization. For example, a retail company periodically retrieves data from the ERP system (structured), the online store backend (structured), and manually recorded PDF ledgers from stores (unstructured) via an HTTP interface at 2:00 AM daily. The system standardizes the "sales quantity" (unit: "pieces") in the ERP system, the "order amount" (unit: "yuan") in the online store, and the "store traffic" (extracted after OCR recognition) in the PDF ledgers: the date format is uniformly "YYYY-MM-DD", the amount is retained to two decimal places, and the traffic is converted to an integer. Finally, the data is stored in the ledger database to ensure that the data formats such as "sales quantity 500 pieces on 2024-03-01" and "order amount 35,000.00 yuan on 2024-03-01" are consistent. S2. Based on the field mapping relationships predefined by the user through the visual configuration interface, query and associate the corresponding ledger data from the ledger database to automatically generate pre-populated report data; the specific steps of step S2 are as follows: S21. Receive and process user-defined field mapping relationships through a visual configuration interface; The user-defined field mapping relationship in step S21 is established through at least one of the following methods: S211. Receive the original field names from the source ledger system or the target report, convert the original field names into field semantic vectors using a semantic vectorization model, and calculate the semantic similarity between the semantic vectors of the source ledger fields and the semantic vectors of the target report fields; wherein, the cosine similarity calculation formula is:

[0034] in, A semantic vector representing the source ledger field. This represents the semantic vector of the target report field, and a mapping is automatically established when the similarity exceeds a preset threshold. S212. Query the pre-built thesaurus containing semantically identical but differently named field pairs across different business systems to assist in establishing mappings; S213. When the semantic similarity is below the threshold, receive keywords or regular expressions input by the user and perform a fuzzy search in the field list of the source ledger system to establish a mapping; S214. Display a list of source ledger fields and target report fields to the user, and receive field associations directly established by the user through drag-and-drop operations, and generate mapping rules; For example, the user configures the field mapping of the "Quarterly Sales Analysis Report": the ERP "Product Sales Quantity" and the report "Sales Volume" are converted into semantic vectors through a semantic vectorization model, and the cosine similarity is calculated to be 0.95, automatically establishing the mapping; the thesaurus is queried to associate the online mall "Customer Actual Payment" with the report "Actual Sales Revenue"; since the semantic similarity between the PDF ledger "Store Reception Personnel" and the report "Store Traffic" is 0.75 < the threshold of 0.8, the user enters the keyword "Traffic Flow" to perform a fuzzy search, and the mapping is established by matching; the ERP "Inventory Occupied Funds" and the report "Inventory Costs" are directly associated through drag-and-drop operations; S22. Based on the established mapping relationship, query the corresponding ledger data from the ledger database, automatically fill the corresponding fields of the target report, and generate pre-filled data for the report; S23. Monitor and record users' correction operations on the pre-filled data in the data entry workbench in real time, and generate user behavior logs; S24. Receive and analyze user behavior logs, and optimize the filling strategy using a pre-trained machine learning model; wherein, the machine learning model is trained using the gradient descent algorithm to minimize the error function:

[0035] Where, is the sample size. Correct the value for the user. These are the model's predicted values; S25. Based on the context of the field currently being operated on, retrieve relevant data from the ledger database and provide the user with real-time population value recommendations; For example, based on the mapping relationship, the system extracts the following data from the ledger database: 12,000 units sold, actual sales revenue of 890,000 yuan, 32,000 customer visits, and inventory cost of 450,000 yuan for Q1 2024, and automatically populates the report. When a user corrects the "actual sales revenue" from 890,000 yuan to 910,000 yuan (to supplement orders not yet entered offline), the behavior collection component records this operation. The system optimizes the model using a gradient descent algorithm to minimize the error between the "user-corrected value of 910,000 yuan" and the "model-predicted value of 890,000 yuan," and adjusts the prediction weights when filling similar reports subsequently. When a user operates on the "inventory cost" field, the system retrieves data from the ledger database based on the context and recommends in real time the "reference value for inventory cost of the same category in Q1 2024: 420,000-480,000 yuan." S3. Based on the verification rules predefined by the user through the visual configuration interface, perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report in sequence, and output the verified report data with anomaly markers and correction suggestions to the user; The specific steps of step S3 are as follows: S31. Receive pre-filled report data, perform validation according to user-configured format validation rules (such as data type, length, and required fields), and output preliminary validation results; For example, after the report is pre-populated, the basic validation component detects that the "In-store Customer Traffic" field is mistakenly filled with "32,000" (the default format is an integer), and immediately outputs the preliminary validation result "Data format error, an integer must be entered"; S32. Receive the data that has passed the initial verification, and call the logical verification rules (such as numerical range and cross-table reconciliation relationship) in the business rule base to perform deep verification; Step S32 also includes the following association verification steps: When a field value is marked as abnormal, other entity data of the same type and scale are retrieved from the ledger database to calculate the reasonable value range of the field to verify the reasonableness of the current value; wherein, reasonableness verification is performed by calculating the deviation rate:

[0036] The average value of the same type is calculated using the K-means clustering algorithm. When the deviation rate exceeds the deviation rate threshold (e.g., 20%), the value is confirmed to be abnormal. For example, after initial verification, the business verification component calls the rule "Actual sales revenue = number of goods sold × average order value" in the rule base to calculate 12,000 items × 75 yuan / item = 900,000 yuan, which deviates from the pre-filled 910,000 yuan by 10,000 yuan, triggering deep verification; the correlation verification unit calls the Q1 customer traffic data of 3 stores of the same size in the same area (28,000, 30,000, and 31,000), calculates the average of the same type of 29,700 people through K-means clustering, and the current value of 32,000 people has a deviation rate of 7.74% < 20%, which is considered reasonable; however, the deviation rate of "actual sales revenue" from the calculated value is 1.11%, prompting "It is recommended to check the average order value or that order data has not been entered"; S33. Receive data that has passed deep verification, calculate the volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed a preset threshold; wherein, the volatility calculation formula is:

[0037] When volatility exceeds the volatility threshold (±30%), it is marked as abnormal; For example, the intelligent early warning component queries the Q1 2023 data, showing that the number of goods sold was 9,500. It calculates the volatility of Q1 2024 as (12,000-9,500) / 9,500×100%≈26.3%<30%, which is considered normal. However, the inventory cost in Q1 2023 was 320,000 yuan, and the volatility of Q1 2024 as (450,000-320,000) / 320,000×100%≈40.6%>30%, which is automatically marked as abnormal. S4. Receive the final report data confirmed and corrected by the user, identify incremental changes between the final report data and the source ledger data using a difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the pre-configured backflow rules. Simultaneously, generate an operation log containing the changed content and version hash value. The specific steps of step S4 are as follows: S41. Compare the field values ​​in the final report data with the corresponding source data in the ledger database to identify the fields that have changed and the old and new values; the difference detection is based on the hash comparison of the field values, and the hash value calculation uses the SHA-256 algorithm: Hash value = SHA - 256 (field value) When the hash value of the source ledger is inconsistent with the hash value of the report, it is determined to be a changed field; For example, after the user confirms the final report data, the difference detection component calculates the SHA-256 hash value for each field and finds that the hash values ​​of "actual sales" (910,000 yuan) and "inventory cost" (450,000 yuan) are inconsistent with the hash values ​​of the source ledger data (890,000 yuan and 430,000 yuan), and determines them to be changed fields; S42. Perform full update, incremental update or condition-triggered synchronization on the changed data according to the pre-configured backflow rules; For example, based on the pre-configured return flow rules, an incremental update is performed on "actual sales" (only the changed values ​​are synchronized) and synchronized to the online mall backend; a condition-triggered update is performed on "inventory cost" (because the fluctuation exceeds 30%, it is synchronized to the ERP system and triggers the inventory review process). S43. When data changes, generate an operation log containing the operation time, operator, changed field, old value, and new value, and generate a unique version identifier for each change set; the version identifier is generated using the MD5 algorithm. Version hash = MD5(operation type + field name + old value + new value + timestamp); For example, the system generates an operation log: Operation time 2024-04-10 15:20, operator xx, changed fields "Actual sales amount" (old value 890,000 yuan / new value 910,000 yuan) and "Inventory cost" (old value 430,000 yuan / new value 450,000 yuan), and calculated the version hash "a1b2c3d4e5f67890abcdef1234567890" using the MD5 algorithm. This version identifier is bound to the operation log and stored.

[0038] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0039] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A report intelligent entry system based on ledger association, characterized in that, include: The ledger management module is used to obtain structured or unstructured ledger data from various source ledger systems through data interfaces and store it in the ledger database. The intelligent data entry engine module is used to query and associate the corresponding ledger data from the ledger database based on the field mapping relationship pre-configured by the user, and generate pre-filled data for reports. The multi-level data verification module is used to sequentially perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report according to the user's pre-configured verification rules, and output the verified report data with anomaly markers and correction suggestions to the user. The data closed-loop management module is used to receive the final report data confirmed by the user, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the user's pre-configured backflow rules, while generating an operation log containing the changed content and version hash value. A visual configuration interface provides users with an interactive interface for defining and configuring field mapping relationships, validation rules, and reflow rules.

2. The intelligent report filling system based on ledger association according to claim 1, characterized in that, The intelligent data entry engine module includes a dynamic field mapping unit and a context-aware controller; The dynamic field mapping unit includes: The semantic parser receives the original ledger field names configured by the user or from the source ledger system, converts them into ledger semantic vectors through a semantic vectorization model, and calculates the semantic similarity between the ledger semantic vectors and the semantic vectors of the target report fields. The thesaurus is used to store pairs of fields that have the same semantics but different names across different business systems, providing auxiliary mapping for the semantic parser; The fuzzy matching unit is used to receive keywords or regular expressions input by the user when the semantic similarity is below a threshold, and to perform a fuzzy query in the field list of the source ledger system to establish a mapping. The visual drag-and-drop component is used to display a list of source ledger fields and target report fields to users, and to receive field relationships established by users through drag-and-drop operations, and generate mapping rules. The context-aware controller includes: The behavior collection component is used to listen to and record users' correction operations on pre-filled data in real time in the data entry workbench, and generate user behavior logs. The strategy optimization component is used to receive and analyze the user behavior logs, and dynamically adjust and optimize the population strategy based on the analysis results. The real-time recommendation component is used to retrieve relevant data from the ledger database based on the context of the field currently being operated on, and to provide users with real-time recommendations for fill values; The semantic parser further includes: The field segmentation component is used to receive the original field names from the source ledger system or the target report, perform word segmentation, and output a word sequence composed of basic word units. The semantic vectorization component is used to receive word segmentation sequences, convert each word unit and contextual relationship into a numerical vector of a preset dimension through a pre-trained semantic model, and finally aggregate them into a field semantic vector that represents the overall semantics of the corresponding field. The similarity calculation component receives the semantic vectors of the source ledger field and the target report field, calculates the cosine similarity between the two vectors using a vector space model, and outputs a quantified similarity score.

3. The intelligent report filling system based on ledger association according to claim 1, characterized in that, The multi-level data verification module includes: The basic validation component is used to receive pre-filled report data, perform validation according to user-configured format validation rules, and output preliminary validation results. The business validation component receives data that has passed the initial validation and calls logical validation rules from the business rule base to perform deep validation. The intelligent early warning component is used to receive data that has passed deep verification, calculate the data volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed preset thresholds. The business verification component further includes: The associated verification unit, when a field value is marked as abnormal, calls entity data of the same type and scale from the ledger database, calculates the reasonable value range of the corresponding field value based on the called entity data, and verifies the reasonableness of the current field value. A logic rule base is used to store and maintain business logic validation rules; The anomaly marking unit is used to uniformly add visual markings and textual correction suggestions to all abnormal data found during the verification process.

4. The intelligent report filling system based on ledger association according to claim 3, characterized in that, The data closed-loop management module includes: The difference detection component is used to compare the report field values ​​in the final report data with the source data of the corresponding report fields in the ledger database during the feedback process, and to identify the fields that have changed and the old and new values. The backflow rule base is used to store data synchronization strategies, including full update, incremental update, or update rules triggered based on preset conditions. The version control component is used to generate an operation log containing change elements such as operation time, operator, changed field, old value, and new value when data changes, and to generate a unique version identifier for each change set; The version control component further includes: The operation log recording unit is used to capture and persistently store data operation flow data; A version identifier generator is used to generate a globally unique version identifier for each data change based on the content of the operation log using a hash algorithm. The change tracking interface provides users with a query function, which can display a complete history of data changes based on version identifiers or time ranges.

5. The intelligent report filling system based on ledger association according to claim 4, characterized in that, The visual configuration interface includes: The field mapping configuration interface is used to display the field list of the source ledger system and the field list of the target report, and provides configuration entry points for semantic parsing, synonyms, fuzzy matching and drag-and-drop association, and finally generates a visual field mapping relationship diagram. The rules settings interface is used to configure format validation rules, business validation logic rules, and data return rules; The workbench interface is used to display pre-filled reports to users, highlight fields that require manual entry or verification, and output exception markers, correction suggestions, and real-time fill value recommendations.

6. A method for intelligent report filling based on ledger association, characterized in that, Includes the following steps: S1. Obtain structured and unstructured ledger data from various source ledger systems through data interfaces, and store the ledger data in the ledger database after cleaning and standardization. S2. Based on the field mapping relationship predefined by the user through the visual configuration interface, query and associate the corresponding ledger data from the ledger database, and automatically generate pre-filled report data; S3. Based on the verification rules predefined by the user through the visual configuration interface, perform format verification, business logic verification, and fluctuation anomaly detection on the pre-filled data of the report in sequence, and output the verified report data with anomaly markers and correction suggestions to the user; S4. Receive the final report data after user confirmation and correction, identify the incremental changes between the final report data and the source ledger data through the difference detection algorithm, and synchronize the changed data back to the corresponding source ledger system according to the pre-configured backflow rules. At the same time, generate an operation log containing the changed content and version hash value.

7. The intelligent report filling method based on ledger association according to claim 6, characterized in that, The specific steps of step S2 are as follows: S21. Receive and process user-defined field mapping relationships through a visual configuration interface; S22. Based on the established mapping relationship, query the corresponding ledger data from the ledger database, automatically fill the corresponding fields of the target report, and generate pre-filled data for the report; S23. Monitor and record users' correction operations on the pre-filled data in the data entry workbench in real time, and generate user behavior logs; S24. Receive and analyze user behavior logs, and optimize the filling strategy using a pre-trained machine learning model; wherein, the machine learning model is trained using the gradient descent algorithm to minimize the error function: Where, is the sample size. Correct the value for the user. These are the model's predicted values; S25. Based on the context of the field currently being operated on, retrieve relevant data from the ledger database and provide the user with real-time populate value recommendations.

8. The intelligent report filling method based on ledger association according to claim 7, characterized in that, The user-defined field mapping relationship in step S21 is established through at least one of the following methods: S211. Receive the original field names from the source ledger system or the target report, convert the original field names into field semantic vectors using a semantic vectorization model, and calculate the semantic similarity between the semantic vectors of the source ledger fields and the semantic vectors of the target report fields; wherein, the cosine similarity calculation formula is: in, A semantic vector representing the source ledger field. This represents the semantic vector of the target report field, and a mapping is automatically established when the similarity exceeds a preset threshold. S212. Query the pre-built thesaurus containing semantically identical but differently named field pairs across different business systems to assist in establishing mappings; S213. When the semantic similarity is below the threshold, receive keywords or regular expressions input by the user and perform a fuzzy search in the field list of the source ledger system to establish a mapping; S214. Display a list of source ledger fields and target report fields to the user, and receive field associations directly established by the user through drag-and-drop operations, and generate mapping rules.

9. The intelligent report filling method based on ledger association according to claim 6, characterized in that, The specific steps of step S3 are as follows: S31. Receive pre-filled report data, perform validation according to user-configured format validation rules, and output preliminary validation results; S32. Receive the data that has passed the initial verification and call the logical verification rules in the business rule base to perform deep verification; S33. Receive data that has passed deep verification, calculate the volatility of key indicators by querying historical data, and automatically mark abnormal fluctuations that exceed a preset threshold; wherein, the volatility calculation formula is: When volatility exceeds the volatility threshold (±30%), it is marked as abnormal.

10. The intelligent report filling method based on ledger association according to claim 6, characterized in that, The specific steps of step S4 are as follows: S41. Compare the field values ​​in the final report data with the corresponding source data in the ledger database to identify the fields that have changed and the old and new values; the difference detection is based on the hash comparison of the field values, and the hash value calculation uses the SHA-256 algorithm: Hash value = SHA - 256 (field value) When the hash value of the source ledger is inconsistent with the hash value of the report, it is determined to be a changed field; S42. Perform full update, incremental update or condition-triggered synchronization on the changed data according to the pre-configured backflow rules; S43. When data changes, generate an operation log containing the operation time, operator, changed field, old value, and new value, and generate a unique version identifier for each change set; the version identifier is generated using the MD5 algorithm. Version hash = MD5(operation type + field name + old value + new value + timestamp).

Images