Commission settlement and reconciliation method and device for social e-commerce, computer equipment and storage medium

Abstract

The invention provides a commission settlement and reconciliation scheme for social e-commerce. According to the social e-commerce commission settlement and reconciliation scheme, a user bill table and a platform bill table of a target user are acquired respectively, a mapping relation between fields of the user bill table and the platform bill table is established by using a target large model, after the fields of the user bill table are mapped, a unique identifier is generated according to a selected identifier field, and record matching and similarity and similarity comparison are realized. The technical effect is remarkable, the account checking problem caused by table format and field setting differences between the user and the platform bill is solved, non-standardized user data is converted into a standardized form capable of being directly compared with platform data, and the data processing efficiency is greatly improved. The matching records are accurately positioned through the unique identifier, and similarities and differences are compared, so that the complexity and errors of manual checking are avoided, and the account checking accuracy is remarkably improved. All links are closely matched to form a complete reconciliation process, and the efficiency and accuracy of social e-commerce commission settlement can be ensured in a scene with various user data formats.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a commission settlement and reconciliation method, computer equipment, and storage medium for social e-commerce. Background Art

[0002] With the rapid development of e-commerce, financial payments, and the sharing economy, commission settlement, a core component of multi-party collaboration, has become increasingly important, with its efficiency directly impacting the conduct of a company's business. Currently, commission settlement systems rely primarily on traditional financial processing methods, including manual data entry and spreadsheet calculations and verification. While some advanced systems utilize computer software to automate some processes, billing data parsing is cumbersome, with varying formats across various businesses. Each access requires the development of a corresponding data cleansing module, increasing development workload and making them prone to errors. Summary of the Invention

[0003] The purpose of this application is to solve at least one of the above-mentioned technical deficiencies, and in particular to provide a solution for efficient and accurate commission settlement and reconciliation.

[0004] In a first aspect, this application provides a commission settlement and reconciliation method for social e-commerce, including:

[0005] Obtain the user bill table and platform bill table corresponding to the target user respectively;

[0006] Input the user billing table and the set field set into the target big model to instruct the target big model to determine the mapping relationship between the fields in the user billing table and the fields in the set field set according to the set field set; the set field set includes all fields that appear in the platform billing table;

[0007] Map the fields of the user bill table according to the mapping relationship;

[0008] Generate a unique identifier for each record in the platform billing table and the mapped user billing table based on the value corresponding to the selected identification field;

[0009] Each record in the platform billing table and the mapped user billing table is compared based on the unique identifier.

[0010] In one embodiment, obtaining a user bill table corresponding to a target user includes:

[0011] Subscribe to the preset user bill topic in the distributed message queue to receive the user bill data stream pushed by each service node;

[0012] Split each user's billing data stream by transaction time window and generate a user bill shard table stored at the set granularity;

[0013] According to the user ID of the target user, the data of the corresponding partition is extracted from the user bill shard table to form the user bill table corresponding to the target user.

[0014] In one embodiment, the process of updating the field set includes:

[0015] Parse the data definition language logs of the database corresponding to the platform billing table and extract metadata descriptions of newly added fields;

[0016] Determine the field name of the new field based on the metadata description;

[0017] Add a new field to the set field set based on the field name of the new field.

[0018] In one embodiment, a unique identifier is generated for each record in the platform bill table and the mapped user bill table based on the value corresponding to the selected identification field, including:

[0019] Generate a unique identifier based on the values ​​of the order number field and transaction time field of each record in the platform bill table and the mapped user bill table.

[0020] In one embodiment, inputting the user billing table and the set field set into the target macro model to instruct the target macro model to determine the mapping relationship between the fields in the user billing table and the fields in the set field set according to the set field set includes:

[0021] The prompt word is determined based on the mapping relationship generated by the user bill table, the set field set and the preset thinking chain; the preset thinking chain is used to instruct the target large model to compare any first field in the user bill table with the semantic similarity of the first field and each second field in the set field set, establish a field mapping between the first field and the second field with the highest semantic similarity, and output the first number of second fields with the highest semantic similarity and their semantic similarity.

[0022] The mapping relationship determination prompt words are input into the target large model to obtain the mapping relationship.

[0023] In one embodiment, before mapping the fields of the user bill table according to the mapping relationship, the method further includes:

[0024] Determine a group of field mappings in which the difference in semantic similarity between the second field with the highest semantic similarity and the remaining second fields is less than a first threshold as a relationship to be reviewed;

[0025] Send the relationship to be reviewed to manual review;

[0026] Update the mapping relationship based on the results of manual review.

[0027] In one embodiment, after updating the mapping relationship according to the result of manual review, the method further includes:

[0028] The manually reviewed mapping relationships and their corresponding user bill tables and set field sets are stored in the training sample library;

[0029] When the training samples stored in the training sample library exceed the second quantity, the target large model is updated using the training sample library, and the training sample library is cleared.

[0030] In one embodiment, after comparing each record in the platform bill table and the mapped user bill table according to the unique identifier, the method further includes:

[0031] Two records with the same unique identifier but with differences are identified as difference records;

[0032] Generate a reconciliation report based on the difference records.

[0033] In one embodiment, after generating the reconciliation report based on the difference records, the method further includes:

[0034] Desensitize data in reconciliation reports;

[0035] The reconciliation report after data desensitization is input into the target large model to instruct the target large model to summarize the reconciliation report after data desensitization.

[0036] In a second aspect, the present application also provides a commission settlement and reconciliation device for social e-commerce, including:

[0037] The data acquisition module is used to obtain the user bill table and platform bill table corresponding to the target user respectively;

[0038] A mapping relationship determination module is used to input the user billing table and the set field set into the target macro model to instruct the target macro model to determine the mapping relationship between the fields in the user billing table and the fields in the set field set based on the set field set; the set field set includes all fields that appear in the platform billing table;

[0039] The mapping module is used to map the fields of the user bill table according to the mapping relationship;

[0040] The identification generation module is used to generate a unique identification for each record in the platform bill table and the mapped user bill table according to the value corresponding to the selected identification field;

[0041] The matching and reconciliation module is used to compare each record in the platform bill table and the mapped user bill table based on the unique identifier.

[0042] In a third aspect, the present application provides a computer device comprising one or more processors and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the steps of the commission settlement and reconciliation method for social e-commerce in any of the above embodiments are executed.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the commission settlement and reconciliation method for social e-commerce in any of the above embodiments.

[0044] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0045] This social e-commerce commission settlement and reconciliation solution obtains the target user's user billing table and platform billing table respectively, uses the target big model to establish a mapping relationship between the two fields, and after mapping the user billing table fields, generates a unique identifier based on the selected identification field to achieve record matching and comparison of similarities and differences. Its technical effect is significant. It solves the reconciliation problem caused by differences in the format and field settings of user and platform billing tables, converts non-standardized user data into a standardized form that can be directly compared with platform data, and greatly improves data processing efficiency. Through the unique identifier, the matching records are accurately located and the differences are compared, which avoids the tediousness and errors of manual verification and significantly improves the accuracy of reconciliation. All links work closely together to form a complete reconciliation process, which can ensure the efficiency and accuracy of social e-commerce commission settlement even in scenarios with diverse user data formats. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0047] Figure 1 A flowchart of a commission settlement and reconciliation method for social e-commerce is provided for one embodiment of the present application;

[0048] Figure 2 A flowchart of obtaining a user bill table corresponding to a target user in one embodiment of the present application is shown;

[0049] Figure 3 This is a flowchart of updating a setting field set in one embodiment of the present application;

[0050] Figure 4A schematic diagram of a process for determining a mapping relationship in one embodiment of the present application;

[0051] Figure 5 A schematic diagram of a process for manually reviewing the mapping relationship output by the target large model in one embodiment of the present application;

[0052] Figure 6 A diagram of the internal structure of a computer device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application and do not belong to all the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0054] This application provides a commission settlement and reconciliation method for social e-commerce. Figure 1 , including steps S102 to S110.

[0055] S102, respectively obtaining a user bill table and a platform bill table corresponding to the target user.

[0056] Target users are understood to be specific entities generating commission settlement needs, encompassing individuals such as individual shop owners and influencers who earn commissions by promoting products through social networks. User billing tables, serving as a record of target users' transaction activity, include transaction flow information such as product sales records and commission income details, serving as a crucial basis for commission settlement and financial accounting. Platform billing tables, generated by the platform, integrate transaction data from all users, platform fee deductions, and platform revenue, serving as a core data source for the platform's financial accounting, business analysis, and reconciliation and settlement with users. The user billing table and platform billing table record transaction information from different perspectives, complementing each other. The user billing table focuses on individual business activities, while the platform billing table presents an overall transaction overview. Only by simultaneously obtaining both billing tables can a comprehensive data foundation be provided for subsequent reconciliation. The absence of either form results in incomplete reconciliation information, making accurate commission settlement impossible. The integration of these two data forms establishes the foundational framework for comparing and analyzing user and platform transaction data, serving as the starting point for the entire reconciliation process.

[0057] S104: Input the user billing table and the set field set into the target macro model to instruct the target macro model to determine the mapping relationship between the fields in the user billing table and the fields in the set field set according to the set field set. The set field set includes all fields that appear in the platform billing table.

[0058] As can be understood, a set field set is a predefined collection of all fields in the platform billing table. It serves as a reference standard for standardizing the data formats and field meanings of both user and platform billing tables, providing a foundation for data mapping and reconciliation. The target big model is an AI model built using deep learning technology. It possesses powerful data analysis, pattern recognition, and semantic understanding capabilities. It analyzes the relationship between the user billing table and the set field set and determines the corresponding mapping relationship between the two fields. Due to the diverse business models and data recording methods of users on social e-commerce platforms, the field settings and formats of user billing tables vary, while the platform billing table format and field standards are unified. For effective reconciliation, a field mapping relationship between the two is necessary. The target big model trains on a large amount of user billing table and set field set data, discovering potential semantic associations and data patterns between fields. Leveraging its powerful capabilities, it matches user billing table fields with the set field set, completing the conversion from non-standardized user data to standardized platform data. This lays the foundation for subsequent reconciliation and resolves reconciliation challenges caused by data format differences. The target big model allows for pre-trained open source big models and fine-tuning. First, preprocess the user bill table and set field set data into text sequences, then input them into the model. By setting training parameters and optimization algorithms, the model learns how to determine the mapping relationship.

[0059] S106: Map the fields of the user bill table according to the mapping relationship.

[0060] It can be understood that field mapping is the process of converting and adjusting the user billing table fields according to the platform billing table format and semantics based on the correspondence between the determined user billing table fields and the set field set (platform billing table field set), aiming to make the data of the two consistent in structure and meaning, so as to facilitate subsequent comparison and reconciliation. After the mapping relationship is determined in step S104, this step applies it to the user billing table data. Through operations such as replacement, renaming, and data type conversion, the user billing table data that originally did not conform to the platform billing table format is made consistent with the platform billing table in terms of field name, data type, data meaning, etc. Field mapping is a key link in connecting user personalized data with platform standardized data, eliminating obstacles of data format differences, making the two billing table data comparable and compatible, and providing a basis for subsequent record matching and commission settlement verification.

[0061] S108 , generating a unique identifier for each record in the platform bill table and the mapped user bill table according to the value corresponding to the selected identification field.

[0062] It can be understood that the identification field is selected from the platform billing table and the user billing table. It is a field that is unique or can uniquely identify key information of the record. It can be a single field such as transaction ID, order number, or a combination of multiple fields. The unique identifier is generated by a specific algorithm based on the selected identification field value. It is used to uniquely identify each record in the billing table. It plays an indexing and distinguishing role in reconciliation to ensure that the record is accurately identified and located in subsequent matching. There are a large number of transaction records in the platform billing table and the mapped user billing table. In order to achieve accurate matching, a unique identifier needs to be generated. By selecting the appropriate identification field, the uniqueness of its value or the combined uniqueness is used to generate a unique identifier. The unique identifier can establish a clear correspondence between the records in the two billing tables, converting complex multi-record comparisons into precise matches based on unique identifiers, thereby improving reconciliation efficiency and accuracy.

[0063] S110 , comparing each record in the platform bill table and the mapped user bill table according to the unique identifier.

[0064] As can be understood, the unique identifier generated in step S108 is used as an index to locate matching records in the two billing tables. When a successful match is found, the system automatically performs a field-by-field comparison of all fields in the record, including the commission amount, transaction status, and settlement time. This comparison goes beyond simple equality checks and can also incorporate complex logical checks based on business rules, such as determining whether the service fee deducted by the platform meets the agreed-upon ratio or whether the actual commission received by the user is consistent with the calculated result. In this way, this step organically combines record matching with data difference comparison, achieving both data alignment and identifying reconciliation discrepancies between user-side and platform-side records.

[0065] This social e-commerce commission settlement and reconciliation solution obtains the target user's user billing table and platform billing table respectively, uses the target big model to establish a mapping relationship between the two fields, and after mapping the user billing table fields, generates a unique identifier based on the selected identification field to achieve record matching and comparison of similarities and differences. Its technical effect is significant. It solves the reconciliation problem caused by differences in the format and field settings of user and platform billing tables, converts non-standardized user data into a standardized form that can be directly compared with platform data, and greatly improves data processing efficiency. Through the unique identifier, the matching records are accurately located and the differences are compared, which avoids the tediousness and errors of manual verification and significantly improves the accuracy of reconciliation. All links work closely together to form a complete reconciliation process, which can ensure the efficiency and accuracy of social e-commerce commission settlement even in scenarios with diverse user data formats.

[0066] In one embodiment, see Figure 2 , obtaining the user bill table corresponding to the target user includes steps S202 to S206.

[0067] S202: Subscribe to a preset user bill topic in a distributed message queue to receive user bill data streams pushed by each service node.

[0068] Distributed message queues, such as Kafka and RabbitMQ, are middleware systems that utilize a distributed architecture and a publish-subscribe model. These systems enable asynchronous communication and decoupling between applications in a distributed environment, ensuring reliable message transmission and processing, and buffering and peak shaving in high-concurrency scenarios. The user billing topic is a predefined message classification identifier within a distributed message queue that aggregates user billing-related data messages. User billing data generated by each service node is published under this topic, serving as a key classification basis for centralized collection of user billing data. The user billing data stream is generated in real time by each service node during the processing of user transactions, commission calculations, and other business processes. It exists as a continuous data sequence and covers transaction details, commission calculation details, and other information.

[0069] In social e-commerce systems, business modules are dispersed across multiple service nodes, with frequent transactions and real-time and bursty data generation. By subscribing to the user billing topic in a distributed message queue, the data collection module can asynchronously receive user billing data streams pushed by each service node, achieving unified data aggregation. This step is the starting point for acquiring user billing data, providing the raw data source for subsequent data processing. By decoupling the service nodes from the data collection module through message queues, high system availability and scalability are ensured. Together with the subsequent data segmentation and extraction steps, this completes the data acquisition process, ensuring comprehensive and timely data capture.

[0070] S204 , splitting each user bill data stream according to the transaction time window, and generating a user bill shard table stored at a set granularity.

[0071] As you can understand, a transaction time window is a pre-set time interval used to segment the continuous user billing data stream into time segments for data processing and storage. It can be set to minutes, hours, or days based on business needs. Granularity is a parameter that defines the level of detail and data block size for user billing data storage. This granularity determines the amount of data and time span contained in each data shard in the user billing shard table. The user billing shard table is a structured table that organizes and stores data according to the set granularity after segmenting the user billing data stream by transaction time window. Each shard table corresponds to the data within a time window, facilitating data management, querying, and subsequent processing.

[0072] Because user billing data streams are continuously generated and the data volume is massive, to improve data processing efficiency and storage management convenience, the data stream is segmented by transaction time window and then stored as sharded tables at a set granularity. This approach converts the continuous data stream into discrete data blocks. During data queries and statistical analysis, relevant sharded tables can be quickly located based on the time range, avoiding the need to scan the entire data set. Segmentation and sharded storage make data processing more targeted and efficient, providing an organized, structured data foundation for subsequent data extraction based on user IDs. The size of the transaction time window can be adaptively adjusted based on user transaction activity. When transaction volume exceeds a peak threshold, the time window is automatically narrowed and the set granularity is refined to ensure timely and accurate data processing. When transaction volume falls below a trough threshold, the time window is expanded and the granularity is coarsened to conserve storage resources.

[0073] S206 , extracting data of a corresponding partition from the user bill shard table according to the user identifier of the target user, and forming a user bill table corresponding to the target user.

[0074] As can be understood, the user ID is an identifier used to uniquely identify the target user within the system. It can be a user ID, mobile phone number, email address, etc., and is the key basis for locating the target user's data among numerous users. The corresponding partition is determined within the user billing shard table based on the time distribution of the target user's transaction data. This partition is determined by filtering using the user ID and time range. After generating the user billing shard table, to obtain the target user's billing data, the relevant data must be extracted from the shard based on the user ID. The user ID is searched across the shard tables, and the corresponding partition containing the target user's data is located based on the transaction time range. The data from these partitions is extracted and integrated to form a user billing table that fully reflects the target user's transactions and commissions. This step is critical for accurately obtaining the target user's data from massive amounts of user billing data. Based on the data obtained and processed in the previous steps, the target user's data is ultimately obtained for reconciliation, providing an accurate data foundation for subsequent data mapping and matching with the platform billing table. This is the final step in the entire user billing data acquisition process.

[0075] In one embodiment, see Figure 3 , the updating process of setting the field set includes steps S302 to S306.

[0076] S302: Parse the data definition language log of the database corresponding to the platform bill table and extract metadata descriptions of the newly added fields.

[0077] As you can understand, a data definition language (DDL) log is a database system's log file that records DDL (Data Definition Language) operations. It contains detailed records of operations such as creating, modifying, and deleting table structures, and is a key data source for tracking database schema changes. Metadata descriptions describe characteristics such as the data structure and attributes within a database, including field names, data types, constraints, and annotations. As social e-commerce systems evolve, the structure of platform billing tables dynamically changes with business development, with the addition of new fields being a common form of change. Parsing DDL logs allows for real-time capture of these structural changes. When executing DDL operations, the database generates a log containing operation details. Parsing this log can extract metadata descriptions of the newly added fields, providing essential information for subsequent processing. This step is the starting point for updating the set field set, providing a basis for discovering and identifying newly added fields. Together with the subsequent determination of field names and updating of the set field set, it forms a complete dynamic field set update mechanism, ensuring that the set field set remains synchronized with the platform billing table structure.

[0078] S304: Determine the field name of the newly added field according to the metadata description.

[0079] It can be understood that newly added fields are data items newly added to the table during the process of changing the platform bill table structure, representing new dimensions or new attribute information that the business system needs to record. The field name is a unique identifier assigned to the newly added field, which is used to reference and operate the field in the database. It is an important part of the data structure and must follow the database naming conventions and be business-readable. After obtaining the metadata description of the newly added field, key information needs to be extracted from it to determine the field name. The metadata description may contain a variety of information, such as field type, constraints, etc., and the field name is the core element. By analyzing the specific identification or location information in the metadata description, the field name can be accurately identified and extracted. Determining the field name is the process of converting technical metadata into business-understandable identification. It provides an accurate identification basis for the subsequent addition of new fields to the set field set. It is a key link connecting metadata parsing and field set updates to ensure that the newly added fields can be correctly identified and referenced.

[0080] S306: Add a new field to the set field set according to the field name of the newly added field.

[0081] It can be understood that adding new fields in this step is an operation to add the newly added field names determined in step S304 to the set field set, so that the set field set can reflect the latest structure of the platform billing table and maintain consistency with the actual business data. After determining the field name of the newly added field, it needs to be added to the set field set. By updating the set field set, it is ensured that it contains all the fields of the platform billing table, providing a complete reference basis for the subsequent field mapping between the user billing table and the platform billing table. The newly added fields reflect changes in business needs, and timely incorporating them into the set field set can enable the entire commission settlement and reconciliation system to adapt to business development and accurately handle new types of data. This step is the ultimate goal of updating the set field set. Together with the previous steps, it ensures the dynamics and accuracy of the set field set, laying the foundation for the subsequent reconciliation process.

[0082] In one embodiment, a unique identifier is generated for each record in the platform billing table and the mapped user billing table based on the value corresponding to the selected identification field, including: generating a unique identifier based on the value of the order number field and the transaction time field of each record in the platform billing table and the mapped user billing table.

[0083] As you can understand, the order number field is a key data item used to uniquely identify each transaction order in a social e-commerce system. It remains unchanged throughout the entire transaction lifecycle and serves as an index for information throughout the entire process, including order creation, payment, delivery, and after-sales service. The transaction time field records the exact moment a transaction occurred, accurate to a timestamp (e.g., year-month-day-hour), reflecting the sequence and temporal characteristics of transactions. The unique identifier is a globally unique string or code generated based on the values ​​of the order number and transaction time fields. It is used to accurately identify each transaction record in the platform billing table and the mapped user billing table.

[0084] In social e-commerce scenarios, the platform and user billing tables record large amounts of transaction data from different sources. Relying solely on a single field cannot ensure accurate correspondence between records. While order numbers can identify orders, the same order number may exist at different times. While transaction times can indicate sequential ordering, multiple transactions may occur at the same time. Combining the order number field with the transaction time field ensures the uniqueness of individual transactions, while the time field distinguishes between transactions with the same order number at different times. These two fields complement each other and provide a reliable data foundation for generating unique identifiers.

[0085] In one embodiment, the user bill table and the set field set are input into the target big model to instruct the target big model to determine the mapping relationship between the fields in the user bill table and the fields in the set field set according to the set field set. Figure 4 , including steps S402 to S404.

[0086] S402: Determine a prompt word based on a mapping relationship generated by the user bill table, the set field set, and a preset thought chain. The preset thought chain instructs the target macromodel to compare the semantic similarity of any first field in the user bill table with each second field in the set field set, establish a field mapping between the first field and the second field with the highest semantic similarity, and output a first number of second fields with the highest semantic similarity and their semantic similarities.

[0087] As can be understood, the preset thought chain is a predefined logical reasoning path that guides the target large model in analyzing the semantic relationships between the user billing table and the set field set. It includes substeps such as field semantic analysis, similarity calculation, and mapping decisions. The mapping relationship determination prompt is a text instruction generated based on the user billing table structure, the set field set content, and the preset thought chain. It guides the target large model in performing the field mapping task, including specific analysis requirements, constraints, and output format requirements. In social e-commerce scenarios, user billing tables have diverse structures, requiring mapping relationships with standard platform fields. The preset thought chain provides a systematic analysis framework, guiding the large model to first extract the semantic features of the first field in the user billing table, then compare them with the semantic features of each second field in the set field set. The degree of association is quantified by calculating a similarity metric, and the top several second fields with the highest similarity are selected as candidate mappings after sorting by similarity. The generated prompt translates this logic into natural language instructions that the large model can understand, ensuring that the model analyzes and outputs the results according to the expected path. This step provides a clear reasoning framework and input guidance for the subsequent large model to perform the mapping task, serving as a critical bridge between data preparation and model analysis.

[0088] S404: Inputting the mapping relationship determination prompt words into the target large model to obtain the mapping relationship.

[0089] As can be understood, after the prompt word is generated in step S402, it is input into the target macro model. Based on the language knowledge and patterns learned during the pre-training phase, the macro model parses the prompt word and performs field semantic analysis and similarity calculation according to the pre-set thought chain. Relying on the macro model's intelligent analysis capabilities, the transformation from unstructured prompts to structured mappings is achieved, providing a basis for subsequent field mapping in the user bill table.

[0090] In one embodiment, before mapping the fields of the user bill table according to the mapping relationship, refer to Figure 5 , also including steps S502 to S506.

[0091] S502 : Determine a group of field mappings in which the difference in semantic similarity between the second field with the highest semantic similarity and the remaining second fields is less than a first threshold as a relationship to be reviewed.

[0092] It can be understood that the semantic similarity difference refers to the difference between the score of the second field with the highest semantic similarity and the scores of the remaining second fields for a first field in the user billing table in the mapping relationship output by the target large-scale model. It is used to measure the discrimination between the highest and second-highest similarity. The first threshold is a pre-set critical value used to determine whether the discrimination between the second field with the highest semantic similarity and the remaining second fields is sufficiently significant. It can be dynamically adjusted based on business needs and historical mapping data. Relationships awaiting review are field mappings where the semantic similarity difference between the second field with the highest semantic similarity and the remaining second fields is less than the first threshold. These mappings are highly ambiguous and require further manual review and confirmation. In real-world business scenarios, due to the diversity of language expressions and the complexity of business terminology, the target large-scale model may have uncertainty in certain field mappings. When the score difference between the second field with the highest semantic similarity and the second-highest similarity is small, it indicates that the model's ability to distinguish these fields is weak, and the mapping results are less reliable. By setting the first threshold, we can filter out relationships with insufficient discrimination and await review, effectively identifying potential mapping errors and providing clear targets for subsequent manual review. This step complements the large model mapping process, using a threshold judgment mechanism to enhance the reliability of the mapping results, providing accurate data to be processed for subsequent manual review, and ensuring the accuracy of the final mapping relationship.

[0093] S504: Send the relationship to be reviewed to manual review.

[0094] Manual review is the process by which professional business personnel or data experts manually verify and correct the pending relationships output by the large model, resolving complex mapping issues that are difficult for the model to handle. This step delivers the pending relationships to the reviewer in a specific format and through a specific channel. This ensures that the reviewer has sufficient information to make a judgment and provides a convenient review feedback mechanism. After determining the pending relationships in step S502, these relationships must be passed to the manual review process. By formatting the pending relationship data to include user billing table fields, candidate set fields, and their semantic similarity values, and providing business context and sample data examples for the fields, reviewers can fully understand the meaning and mapping of the fields. A workflow system or review platform is used to assign the pending relationships to the appropriate reviewer, allowing reviewers to confirm, correct, or add comments through the interface. Manual review, as a supplement to the model's automatic mapping, can effectively address the model's limitations in semantic understanding and business logic judgment, improve the accuracy and reliability of mapping relationships, and ensure the quality of the data ultimately used for reconciliation.

[0095] S506: Update the mapping relationship according to the result of manual review.

[0096] It is understandable that after the manual review is completed, the system needs to update the mapping relationship based on the review results. For the mapping relationship confirmed by the review, it will be marked as confirmed and included in the final mapping result; for the mapping relationship adjusted by the reviewer, the original mapping relationship will be replaced with a new mapping field pair; for special cases discovered during the review, such as fields with ambiguity or requiring special processing, additional mapping rules or remarks will be added. The updated mapping relationship will serve as the basis for subsequent user bill table field mapping, directly affecting the accuracy of the reconciliation results. This step is the final link in the entire mapping review process. By incorporating the professional judgment of manual review into the mapping relationship, it makes up for the shortcomings of the model's automatic mapping and improves the overall performance and reliability of the system.

[0097] In one embodiment, after updating the mapping relationship based on the manual review results, the method further includes storing the manually reviewed mapping relationship and its corresponding user bill table and set of settings in a training sample library. When the training samples stored in the training sample library exceed a second number, the target large model is updated using the training sample library, and the training sample library is cleared.

[0098] It can be understood that the training sample library is a storage system used to store data related to the training of the target large model. In this scenario, it mainly stores manually reviewed mapping relationships, as well as the corresponding user billing tables and set field sets. This data contains a large number of field mapping cases in actual business scenarios and is an important data source for large model learning and optimization. The second quantity is a pre-set threshold used to measure whether the number of samples in the training sample library meets the conditions for triggering a model update. This threshold can be dynamically adjusted based on factors such as model training requirements, computing resources, and the frequency of business data changes.

[0099] In the social e-commerce commission settlement and reconciliation process, while the target large model can initially determine field mapping relationships based on pre-set thought chains and prompts, manual review can correct errors caused by semantic ambiguity and misinterpretation of business rules. Storing these manually reviewed mapping relationships and their corresponding data in the training sample library is the process of accumulating high-quality training data for the model. When the number of samples in the training sample library exceeds a second threshold, indicating a sufficient number of representative new cases, the target large model is updated. This allows the model to learn the business knowledge and judgment logic reflected in the manual review process and optimize its mapping capabilities. After the update, the training sample library is cleared to avoid repeated training and ensure that the model is continuously exposed to new business changes and data features in subsequent training. This process works closely with the previous field mapping and manual review steps, forming a cycle of business processing and model updates, achieving continuous improvement in model performance. This ensures that field mapping tasks can be accurately completed in complex and changing social e-commerce business scenarios and provides reliable support for subsequent reconciliation work.

[0100] In one embodiment, after comparing each record in the platform billing table and the mapped user billing table based on the unique identifier, the process further includes: identifying two records with the same unique identifier but with discrepancies as discrepant records. A reconciliation report is generated based on the discrepant records. Discrepant records are understood to be records in the platform billing table and the mapped user billing table that have the same unique identifier but whose key data fields do not match. These records reflect inconsistencies in transaction data between the user and the platform, such as discrepancies in values ​​for fields like commission amounts and service fee deductions. A reconciliation report is a structured document generated by integrating discrepant record information. It contains details about the discrepancies and a classification of the discrepancies by type, providing a basis for financial personnel to verify and resolve discrepancies. In social e-commerce commission settlement reconciliation, unique identifiers are used to link records in the two billing tables. If the key data field values ​​of records with the same unique identifier are inconsistent, these records are identified as discrepancies. This is because identical unique identifiers indicate that the records correspond to the same transaction, while data discrepancies indicate that there are inconsistencies between the user's and the platform's records of the transaction, requiring further verification. Generating a reconciliation report systematically organizes the discrepancies, providing clear guidance for subsequent financial processing. This step is closely linked to the unique identifier generation and record matching steps. Unique identifier generation is the basis for difference identification, record matching is the process of difference discovery, and difference record determination and report generation are the output links of reconciliation results. Together, they constitute a complete reconciliation closed loop to ensure that differences are traceable and processable.

[0101] In one of the embodiments, after generating a reconciliation report based on the difference records, it also includes: desensitizing the data of the reconciliation report. The reconciliation report after data desensitization is input into the target big model to instruct the target big model to summarize the reconciliation report after data desensitization. It can be understood that data desensitization is the process of deforming sensitive information in the data through a specific algorithm. In the reconciliation report scenario, it involves replacing, masking or generalizing sensitive fields such as user ID, transaction amount, bank account, etc., to ensure that the data maintains business analysis value without leaking privacy. The summary here refers to the target big model generating a refined text containing core information such as difference type distribution, main problem sources, and processing suggestions based on the content of the reconciliation report to provide data support for decision-making.

[0102] In the social e-commerce reconciliation process, reconciliation reports contain a wealth of sensitive user and transaction information. Directly distributing this information poses a risk of privacy breaches. Data desensitization replaces sensitive fields with dummy values ​​or masks, such as replacing the user ID "123456" with "UID-789." This protects privacy while preserving the data's statistical characteristics and business logic. The desensitized report is then fed into the target large-scale model. Based on the language knowledge and business models learned during the pre-training phase, the model automatically identifies key differences, trends, and potential issues within the report. Through abstraction, classification, and logical reasoning, it generates a highly summarized and decision-making summary. This step is closely integrated with the identification of discrepancies and report generation, forming a complete chain of "discrepancy discovery - report generation - privacy protection - intelligent analysis," ensuring both data security and improved information utilization efficiency.

[0103] On the second aspect, the present application also provides a commission settlement and reconciliation device for social e-commerce, including: a data acquisition module, which is used to respectively obtain the user bill table and the platform bill table corresponding to the target user. A mapping relationship determination module, which is used to input the user bill table and the set field set into the target big model to instruct the target big model to determine the mapping relationship between the fields in the user bill table and the fields in the set field set according to the set field set. The set field set includes all fields appearing in the platform bill table. A mapping module, which is used to map the fields of the user bill table according to the mapping relationship. An identifier generation module, which is used to generate a unique identifier for each record in the platform bill table and the mapped user bill table according to the value corresponding to the selected identifier field. A matching and reconciliation module, which is used to compare each record in the platform bill table and the mapped user bill table according to the unique identifier.

[0104] For the specific limitations of the commission settlement and reconciliation device of social e-commerce, please refer to the limitations of the commission settlement and reconciliation method of social e-commerce above, which will not be repeated here. The various modules in the above-mentioned commission settlement and reconciliation device of social e-commerce can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0105] The present application provides a computer device comprising one or more processors and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the one or more processors, the steps of the commission settlement and reconciliation method for social e-commerce in any of the above embodiments are executed.

[0106] Schematically, as Figure 6 As shown, Figure 6 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. Figure 6 Computer device 600 includes a processing component 602, which further includes one or more processors, and memory resources represented by memory 601 for storing instructions executable by processing component 602, such as applications. The applications stored in memory 601 may include one or more modules, each corresponding to a set of instructions. Furthermore, processing component 602 is configured to execute instructions to perform the steps of the social e-commerce commission settlement and reconciliation method described in any of the above-described embodiments.

[0107] The computer device 600 may further include a power supply component 603 configured to perform power management of the computer device 600 , a wired or wireless network interface 604 configured to connect the computer device 600 to a network, and an input / output (I / O) interface 605 .

[0108] The present application provides a computer-readable storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the commission settlement and reconciliation method for social e-commerce in any of the above embodiments.

[0109] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0110] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referenced to each other.

[0111] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one 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 present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A commission settlement and reconciliation method for social e-commerce, characterized in that: include: Obtain the user bill table and platform bill table corresponding to the target user respectively; Inputting the user billing table and the set field set into the target macro model to instruct the target macro model to determine the mapping relationship between the fields in the user billing table and the fields in the set field set according to the set field set; the set field set includes all fields appearing in the platform billing table; Mapping the fields of the user bill table according to the mapping relationship; Generate a unique identifier for each record in the platform bill table and the mapped user bill table according to the value corresponding to the selected identification field; Each record in the platform bill table and the mapped user bill table is compared according to the unique identifier.

2. The commission settlement and reconciliation method according to claim 1, characterized in that: Obtaining the user bill table corresponding to the target user includes: Subscribe to the preset user bill topic in the distributed message queue to receive the user bill data stream pushed by each service node; Segmenting each user bill data stream according to the transaction time window to generate a user bill shard table stored at a set granularity; According to the user identifier of the target user, data of the corresponding partition is extracted from the user bill shard table to form the user bill table corresponding to the target user.

3. The commission settlement and reconciliation method according to claim 1, characterized in that: The updating process of the setting field set includes: Parsing the data definition language log of the database corresponding to the platform bill table and extracting metadata descriptions of the newly added fields; Determine the field name of the newly added field according to the metadata description; According to the field name of the newly added field, a new field is added to the set field set.

4. The commission settlement and reconciliation method according to claim 1, characterized in that: Generating a unique identifier for each record in the platform bill table and the mapped user bill table according to the value corresponding to the selected identification field includes: The unique identifier is generated according to the values ​​of the order number field and the transaction time field of each record in the platform bill table and the mapped user bill table.

5. The commission settlement and reconciliation method according to claim 1, characterized in that: Inputting the user bill table and the set field set into the target macro model to instruct the target macro model to determine, according to the set field set, a mapping relationship between fields in the user bill table and fields in the set field set, includes: A prompt word is determined based on a mapping relationship generated according to the user bill table, the set field set, and a preset thought chain; the preset thought chain is used to instruct the target large model to compare any first field in the user bill table with the semantic similarity of each second field in the set field set, establish a field mapping between the first field and the second field with the highest semantic similarity, and output a first number of the second fields with the highest semantic similarity and their semantic similarities; The mapping relationship determination prompt words are input into the target large model to obtain the mapping relationship.

6. The commission settlement and reconciliation method according to claim 5, characterized in that: Before mapping the fields of the user bill table according to the mapping relationship, the method further includes: Determining, in the mapping relationship, a group of the field mappings, in which the difference in the semantic similarity between the second field with the highest semantic similarity and the remaining second fields is less than a first threshold, as a relationship to be reviewed; Send the relationship to be reviewed to manual review; The mapping relationship is updated according to the result of the manual review.

7. The commission settlement and reconciliation method according to claim 6, characterized in that: After the mapping relationship is updated according to the result of the manual review, the method further includes: Storing the manually reviewed mapping relationship and its corresponding user bill table and setting field set in a training sample library; When the training samples stored in the training sample library exceed a second number, the target large model is updated using the training sample library, and the training sample library is cleared.

8. The commission settlement and reconciliation method according to claim 1, characterized in that: After comparing each record in the platform bill table and the mapped user bill table according to the unique identifier, the method further includes: Determine two records that have the same unique identifier and are different as difference records; Generate a reconciliation report based on the difference records.

9. The commission settlement and reconciliation method according to claim 8, characterized in that: After generating the reconciliation report according to the difference record, the method further includes: Desensitizing the data in the reconciliation report; The reconciliation report after the data is desensitized is input into the target big model to instruct the target big model to summarize the reconciliation report after the data is desensitized.

10. A commission settlement and reconciliation device for social e-commerce, characterized in that: include: The data acquisition module is used to obtain the user bill table and platform bill table corresponding to the target user respectively; a mapping relationship determination module, configured to input the user billing table and the set field set into a target macro model, and instruct the target macro model to determine, based on the set field set, a mapping relationship between fields in the user billing table and fields in the set field set; the set field set includes all fields appearing in the platform billing table; A mapping module, configured to map the fields of the user bill table according to the mapping relationship; an identification generation module, configured to generate a unique identification for each record in the platform bill table and the mapped user bill table, respectively, based on the value corresponding to the selected identification field; A matching and reconciliation module is used to compare each record in the platform bill table and the mapped user bill table according to the unique identifier.

11. A computer device, characterized in that: It includes one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the commission settlement and reconciliation method for social e-commerce described in any one of claims 1 to 9 are executed.

12. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the commission settlement and reconciliation method for social e-commerce as described in any one of claims 1 to 9.

Images