Data processing method and device

Abstract

The invention discloses a data processing method and device, and relates to the technical field of warehouse logistics. A specific embodiment of the method comprises the following steps: receiving consultation information for overdue money return of a logistics user, performing text analysis on the consultation information, and identifying an entity from a text analysis result by using a preset entity library; retrieving a graph database according to the identified entity to obtain a sub-knowledge graph associated with the logistics user; determining relation nodes having a causal relationship with the logistics user from the sub-knowledge graph, and obtaining node attribute data of each relation node; and inputting the relation node and the node attribute data into a preset model, and taking the money return overdue reason output by the preset model as reply information and displaying the reply information. According to the embodiment of the invention, the method achieves the efficient and automatic processing of the overdue money return problem of the logistics user, reduces the labor cost, and improves the accuracy and speed of solving the problem.

Description

Technical Field

[0001] The present invention relates to the technical field of warehousing logistics, and in particular to a data processing method and device. Background Art

[0002] In today's complex and ever-changing environment, logistics users and logistics service providers work closely together to ensure the efficient transportation and delivery of goods. However, the problem of overdue payment from logistics users often occurs, which brings considerable challenges to the capital circulation and operation of logistics service providers. With the increasingly fierce market competition, the dependence of logistics users on logistics services continues to increase, and logistics service providers play a vital role in the entire supply chain. Logistics service providers need to recover funds in a timely manner to maintain their own financial balance and normal business operations, but overdue payment from logistics users has become a prominent problem, which may lead to a series of adverse consequences such as tight funds for logistics service providers, obstructed project progress, and disrupted financial plans.

[0003] In this context, logistics service providers urgently need to deeply explore the root causes of overdue payment from logistics users so that targeted measures can be taken to solve them and ensure the stable development and interests of enterprises. At the same time, it will also help logistics users and logistics service providers to establish a healthier, more stable and sustainable cooperative relationship and promote the healthy development of the entire industry. The technology currently used is in the form of manual tracking, which has the problems of low efficiency, lack of dynamism and timeliness. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a data processing method and device, which can at least solve the problems of low efficiency, lack of dynamics and timeliness in the manual tracking method in the prior art.

[0005] To achieve the above-mentioned purpose, according to one aspect of an embodiment of the present invention, a data processing method is provided, including: receiving consultation information regarding overdue payment of a logistics user, performing text analysis on the consultation information, and using a preset entity library to identify entities from the text analysis results; searching a graph database based on the identified entities to obtain a sub-knowledge graph associated with the logistics user; determining from the sub-knowledge graph the relationship nodes that have a causal relationship with the logistics user, and obtaining node attribute data of each relationship node; inputting the relationship nodes and node attribute data into a preset model, and displaying the reason for overdue payment output by the preset model as reply information; wherein the preset model constructs context information based on the association relationship between relationship nodes and the node attribute data, and determines the reason for overdue payment based on the context information.

[0006] Optionally, the process of determining the relationship node that has a causal relationship with the logistics user is: obtaining the retrieval depth, taking the node where the logistics user is located as the starting point, and determining the first-level relationship node in the sub-knowledge graph that is directly related to the logistics user; taking the first-level relationship node as the starting point, determining the second-level relationship node in the sub-knowledge graph that is directly related to the first-level relationship node and indirectly related to the logistics user; repeating the above-mentioned operations of determining the starting point and relationship node until the relationship node of the retrieval depth is determined.

[0007] Optionally, the sub-knowledge graph is a part of the knowledge graph in the graph database, and the process of constructing the knowledge graph is: determine the configuration file for the import tool; wherein the configuration file uses the data warehouse as the data source and the graph database as the data storage target; read data from each entity table in the data warehouse through the data framework access interface of the import tool; wherein the entity table is used to store entity triple relationships and entity attribute data; through the mapping relationship in the configuration file, convert the entity triple relationships into labels and edges that meet the requirements of the graph database, and convert the entity attribute data into label attribute data; import the labels, edges and label attribute data into the graph database, so that the graph database uses labels as nodes and label attribute data as node attribute data, combined with edges, to construct a knowledge graph.

[0008] Optionally, the process of constructing an entity table in a data warehouse is as follows: obtaining a set of overdue payment data of logistics users, identifying entities and entity data of each entity from the overdue payment data of each logistics user; receiving an identifier configured for each entity, and obtaining the relationship configured between entities, generating an entity triplet relationship based on the entity identifier and the relationship between entities; obtaining attributes pre-configured for each entity, obtaining attribute values ​​corresponding to the attributes from each entity data, and constructing entity attribute data based on the entity identifier, attribute and attribute value; creating an entity table for each entity in the data warehouse, and storing the entity triplet relationship and entity attribute data of each entity in the entity table of each entity.

[0009] Optionally, the obtaining of a logistics user's overdue payment collection data set includes: collecting overdue payment collection data associated with each logistics user from multiple data sources; using a preset duplicate data detection strategy, an erroneous data identification strategy, and an irrelevant data identification strategy to process the overdue payment collection data of each logistics user to obtain a logistics user's overdue payment collection data set.

[0010] Optionally, the determination of the configuration file for the import tool configuration includes: receiving the link information of the graph database configured in the configuration file; receiving the label configuration information configured in the configuration file; wherein the label configuration information includes the label name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the field list of the entity table, the field list of the label attribute, and the mapping relationship between the entity table fields and the label attribute fields; receiving the edge configuration information configured in the configuration file; wherein the edge configuration information includes the edge name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the starting label name of the edge, and the ending label name.

[0011] Optionally, the link information includes a service address, a database name, a user name and a password, and the method further includes: determining the graph database through the service address and database name of the graph database in the configuration file, and then using the user name and password to establish a communication link with the graph database.

[0012] To achieve the above-mentioned purpose, according to another aspect of an embodiment of the present invention, there is provided a data processing device, including: an entity recognition module, used to receive consultation information regarding overdue payment of a logistics user, perform text analysis on the consultation information, and use a preset entity library to identify entities from the text analysis results; a graph retrieval module, used to search a graph database according to the identified entities to obtain a sub-knowledge graph associated with the logistics user; a node determination module, used to determine, from the sub-knowledge graph, relationship nodes that have a causal relationship with the logistics user, and to obtain node attribute data of each relationship node; a cause analysis module, used to input relationship nodes and node attribute data into a preset model, and display the cause of overdue payment output by the preset model as reply information; wherein the preset model constructs context information according to the association relationship between relationship nodes and node attribute data, and determines the cause of overdue payment according to the context information.

[0013] Optionally, the node determination module is used to: obtain the retrieval depth, taking the node where the logistics user is located as the starting point, and determining the first-level relationship node in the sub-knowledge graph that is directly associated with the logistics user; taking the first-level relationship node as the starting point, determine the second-level relationship node in the sub-knowledge graph that is directly associated with the first-level relationship node and indirectly associated with the logistics user; repeat the above-mentioned operations of determining the starting point and relationship node until the relationship node of the retrieval depth is determined.

[0014] Optionally, the sub-knowledge graph is a part of the knowledge graph in the graph database, and the device also includes a knowledge graph construction module, which is used to: determine a configuration file for configuring an import tool; wherein the configuration file uses the data warehouse as the data source and the graph database as the data storage target; read data from each entity table in the data warehouse through the data framework access interface of the import tool; wherein the entity table is used to store entity triple relationships and entity attribute data; through the mapping relationship in the configuration file, the entity triple relationships are converted into labels and edges that meet the requirements of the graph database, and the entity attribute data is converted into label attribute data; the labels, edges and label attribute data are imported into the graph database, so that the graph database uses labels as nodes, label attribute data as node attribute data, and combines edges to construct a knowledge graph.

[0015] Optionally, the device also includes a module for constructing an entity table, which is used to: obtain a set of data on overdue payment collection from logistics users, identify entities and entity data of each entity from the overdue payment collection data of each logistics user; receive an identifier configured for each entity, and obtain the relationship configured between entities, and generate an entity triplet relationship based on the entity identifier and the relationship between entities; obtain attributes pre-configured for each entity, obtain attribute values ​​corresponding to the attributes from each entity data, and construct entity attribute data based on the entity identifier, attribute and attribute value; create an entity table for each entity in the data warehouse, and store the entity triplet relationship and entity attribute data of each entity in the entity table of each entity.

[0016] Optionally, the entity table construction module is used to: collect overdue payment data associated with each logistics user from multiple data sources; use preset duplicate data detection strategy, error data identification strategy and irrelevant data identification strategy to process the overdue payment data of each logistics user to obtain a set of logistics user overdue payment data.

[0017] Optionally, the knowledge graph construction module is used to: receive link information of the graph database configured in the configuration file; receive label configuration information configured in the configuration file; wherein the label configuration information includes the label name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the field list of the entity table, the field list of the label attribute, and the mapping relationship between the entity table fields and the label attribute fields; receive edge configuration information configured in the configuration file; wherein the edge configuration information includes the edge name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the starting label name of the edge, and the ending label name.

[0018] Optionally, the link information includes a service address, a database name, a user name and a password. The knowledge graph construction module is also used to: determine the graph database through the service address and database name of the graph database in the configuration file, and then use the user name and password to establish a communication link with the graph database.

[0019] To achieve the above objective, according to another aspect of an embodiment of the present invention, a data processing electronic device is provided.

[0020] The electronic device of an embodiment of the present invention includes: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned data processing methods.

[0021] To achieve the above objective, according to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, any of the above data processing methods is implemented.

[0022] To achieve the above object, according to another aspect of an embodiment of the present invention, a computer program product is provided. A computer program product of an embodiment of the present invention includes a computer program, and when the program is executed by a processor, the data processing method provided by the embodiment of the present invention is implemented.

[0023] According to the solution provided by the present invention, one embodiment of the above invention has the following advantages or beneficial effects: In order to solve the problems of low efficiency, timeliness and dynamism in the existing reliance on manual tracking and communication with logistics users on the reasons for overdue payment, this solution adopts a root cause analysis method for overdue payment of logistics users based on a large language model, and combines cause-and-effect graph technology to conduct in-depth analysis and root cause tracing of the problem of non-payment, so as to facilitate the subsequent provision of targeted solutions to logistics users, thereby improving the business collection rate.

[0024] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0026] Figure 1 is a main flow chart of a data processing method according to an embodiment of the present invention;

[0027] Figure 2 is a flowchart of an optional data processing method according to an embodiment of the present invention;

[0028] Figure 3is a flowchart of another optional data processing method according to an embodiment of the present invention;

[0029] Figure 4 is an example diagram of a knowledge graph according to an embodiment of the present invention;

[0030] Figure 5 is a flowchart of a specific data processing method according to an embodiment of the present invention;

[0031] Figure 6 is a schematic diagram of main modules of a data processing device according to an embodiment of the present invention;

[0032] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0033] Figure 8 It is a schematic diagram of the structure of a computer system of a mobile device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0035] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0036] In supply chain management, the relationship between merchants (i.e., logistics users), customers, and logistics service providers is interdependent and closely linked. Specifically, the relationship between the three can be described as follows:

[0037] 1. Merchants and customers. Merchants: companies or individuals that produce or sell goods and are responsible for providing products or services to customers. Customers: end users or companies that purchase or use merchant products or services. The relationship between merchants and customers: merchants earn income by selling products or services to meet customer needs. Customers pay for goods, merchants provide products or services, and the two parties establish a business relationship through transactions. Customer satisfaction directly affects the sales and brand reputation of merchants.

[0038] 2. Merchants and logistics service providers. Logistics service providers: companies that provide logistics services such as cargo transportation, warehousing, and distribution. Relationship between merchants and logistics service providers: merchants rely on logistics service providers to transport products from production sites or warehouses to customers. The service quality of logistics service providers (such as transportation speed, cargo integrity, etc.) directly affects the merchant's customer satisfaction and brand image. Merchants pay logistics fees, and logistics service providers provide corresponding logistics services.

[0039] 3. Relationship between customers and logistics service providers: Customers expect to receive the purchased goods on time and in good condition, so they have direct experience and evaluation of the services provided by logistics service providers. Logistics service providers directly contact customers during the transportation process, such as through delivery personnel or customer service. Customer satisfaction with logistics services will also affect their overall evaluation of the merchant.

[0040] The combined impact of the relationship between merchants, customers and logistics service providers:

[0041] 1. Information flow: Merchants need to pass on order information, delivery information, etc. to logistics service providers, and logistics service providers need to feed back transportation progress, delivery information, etc. to merchants and customers. Timely and accurate transmission of information is crucial to the smooth operation of the entire supply chain.

[0042] 2. Cash flow: Customers pay merchants for goods, and merchants pay logistics fees to logistics service providers. The timeliness of payment and smooth cash flow have an important impact on the operation and financial status of merchants and logistics service providers.

[0043] 3. Logistics Flow: Logistics service providers are responsible for the actual transportation and distribution of goods, ensuring that the process of goods from merchants to customers is efficient and error-free.

[0044] The root causes of overdue payment from logistics users include but are not limited to the following:

[0045] 1. Delayed payment by customers: The consignee fails to pay for the goods in time, resulting in obstruction of the capital chain.

[0046] 2. Disputes over logistics warehousing and distribution services: For example, problems such as discrepancies in incoming / outgoing inventory, damaged or lost goods, and delayed delivery may cause disputes between customers and merchants and affect the progress of payment collection.

[0047] 3. Complex or inefficient settlement process: cumbersome settlement procedures and poor coordination between multiple departments or links lead to longer collection time.

[0048] 4. Untimely information communication: Information transmission between merchants, customers and logistics service providers is untimely and inaccurate, resulting in misunderstanding of collection time or delayed processing.

[0049] 5. Financial reconciliation issues: The financial accounts of both parties are inconsistent, which requires time to check and resolve, such as write-off issues in the settlement process of the logistics service provider.

[0050] 6. The company is short of funds: The customer has difficulty in cash flow and is unable to pay on time.

[0051] 7. Changes in the market environment: such as economic downturn, customers’ financial pressure increases, and their willingness or ability to collect payments decreases.

[0052] 8. Unclear contract terms: The specific requirements and time points for payment collection are vague, which can easily lead to disagreements and delays.

[0053] 9. Lack of credit management: Inaccurate credit assessment of customers and insufficient estimation of collection risk.

[0054] 10. Conflict of interest: Merchants use the logistics services of the same company’s mall retail, the logistics company’s mall retail does not rebate the merchants, there are no abnormalities in the merchant’s store, procurement and sales cannot rebate in time, resulting in freight overdue, etc., the merchants are unwilling to collect the money.

[0055] In order to clarify the payment status of each logistics user and the specific reasons for non-payment, so as to formulate strategies and methods for subsequent payment collection and improve the payment collection rate of the business. At present, with the help of manual tracking, every month or quarter, the financial staff will identify the accounts that have expired or the merchants with a history of overdue payments, and then formulate a schedule for the visit plan. The sales department will designate a person to communicate with the logistics service provider about the actual reasons for the overdue payment, and the financial department will then plan the subsequent payment collection plan for overdue merchants. However, since the situation of each merchant is different, the reasons for non-payment are also different, and the corresponding data needs to be obtained for verification, manual tracking has many obvious shortcomings, such as the following two points:

[0056] 1. Manual tracking of overdue reasons is inefficient: Due to the large number of merchants of different types, complex product structures and diverse collection periods, a large number of financial personnel and sales personnel are required to analyze the collection status of each merchant, and sales personnel are required to visit and communicate the specific reasons for the overdue collection. It is difficult to quickly and comprehensively cover the specific details of each merchant. It is very cumbersome and inefficient, and some key information or situations may be missed.

[0057] 2. Lack of dynamism and timeliness: The sales department relies on designated personnel to communicate with merchants periodically, and is unable to discover problems and make adjustments and take countermeasures in real time based on the latest situation of the merchants.

[0058] See also Figure 1 , which shows a main flow chart of a data processing method provided by an embodiment of the present invention, comprising the following steps:

[0059] S101: receiving consultation information regarding overdue payment from a logistics user, performing text analysis on the consultation information, and identifying entities from the text analysis results using a preset entity library;

[0060] S102: searching a graph database according to the identified entity to obtain a sub-knowledge graph associated with the logistics user;

[0061] S103: determining, from the sub-knowledge graph, the relationship nodes that have a causal relationship with the logistics user, and obtaining node attribute data of each relationship node;

[0062] S104: inputting the relationship nodes and node attribute data into the preset model, and displaying the reason for overdue payment output by the preset model as reply information; wherein the preset model constructs context information according to the association relationship between the relationship nodes and the node attribute data, and determines the reason for overdue payment according to the context information.

[0063] In the above implementation, based on the specific task of analyzing the root cause of overdue payment from logistics users, this solution uses Nebula Graph database combined with LLM (Large Language Model) retrieval enhancement technology. By collecting contextual information related to overdue payment from logistics users, and using a machine learning model with cognitive capabilities for contextual learning, the answer to the task is synthesized, that is, knowledge graph combined with LLM retrieval enhancement technology (Graph Retrieval Argumented Generation). Among them, Nebula Graph is a high-performance distributed graph database suitable for processing large-scale graph data and complex graph queries. Nebula Graph is not the only choice, you can also choose Neo4j, JanusGraph, HugeGraph, DGraph and NebulaGraph.

[0064] For step S101, the staff clicks on the "X merchant payment overdue reason task" and receives the overdue reason question of the logistics user input by the staff, such as please tell me the reason for the overdue payment of X merchant. Use the LLM model to extract key entities from the question, such as performing text analysis on "Please tell me the reason for the overdue payment of X merchant" to identify noun words, such as the specific "logistics user" name X, "payment", "overdue", and some related concepts may also be extracted, such as "contract", "customer", "transportation" and other entities. Since there are multiple text analysis results, in order to distinguish which are entities and which are non-entities, they can be filtered based on the entity library. The entity library here can be generated or continuously updated based on the entities identified by the big data platform, or it can be generated in advance, and there is no restriction here.

[0065] This solution automatically receives consulting information and performs text analysis, reducing the need for human intervention; it quickly and accurately identifies entities related to consulting information through a preset entity library, improving the speed and accuracy of information processing, and avoiding misunderstanding of consulting information through precise entity recognition, ensuring the correctness of subsequent processing.

[0066] For step S102, the LLM model retrieves subgraphs through LlamaIndex based on these entities and goes deep into a certain depth. A sub-knowledge graph is a subset of a large knowledge graph. When it is necessary to analyze the reasons for overdue payment of a specific logistics user (such as merchant X), it is necessary to extract the part of the graph related to merchant X from the entire knowledge graph. This part of the graph includes merchant X's orders, customers, payments, logistics information, overdue records and other related data.

[0067] Retrieval is a common operation. The above LlamaIndex is just an example. It is a data framework used by LLM applications to ingest, structure, and access private or domain-specific data, showing a possible implementation method. In fact, other tools and methods can be used to retrieve subgraphs from knowledge graphs and analyze them in conjunction with LLM. For example, you can directly use the query language of the graph database itself, such as Neo4j's Cypher, Nebula Graph's nGQL, JanusGraph's Gremlin, etc., and you can directly use these query languages ​​to retrieve subgraphs.

[0068] For step S103, after the subgraph is retrieved, it is necessary to determine the relationship nodes that have a causal relationship with the logistics user. Assuming the search depth is 2, taking "logistics user X" as the core node, going to a depth of 2, the multi-level relationship nodes that have a causal relationship with "logistics user X" are obtained:

[0069] 1. First-level relationship, for example 1) Find the contract node related to the logistics user X, and check the specific agreement on payment collection in the contract, such as payment time, conditions, and methods; 2) Find the transaction record node to check whether there are any abnormal transactions or unsettled payments; 3) Associate the customer information node to understand the customer's credit status, previous payment behavior, etc.

[0070] 2. Secondary relationships, such as 1) extending from the contract node to view the background, process and other relevant information of the contract signing; 2) under the transaction record node, further view the specific logistics business links, such as delivery time, conditions during transportation, etc.; 3) expanding from the customer information node to analyze the possible impact of the customer's industry characteristics, market environment, etc. on the collection of payments; 4) for the transportation link, find the relevant transportation methods, routes, transportation companies and other nodes for analysis.

[0071] It can be seen that the first-level relationship node is the node directly associated with the "logistics user" node, and the second-level relationship node is the node indirectly associated with the "logistics user" node, that is, the node associated through the first-level relationship node. The causal graph technology is used here to obtain the first and second-level relationship nodes related to the logistics user X. In addition, each node in the knowledge graph has node attribute data, so while determining the multi-level relationship nodes, the node attribute data of each relationship node can also be obtained from the subgraph.

[0072] The above is just an example. In reality, there are also third-level nodes and beyond. The third-level node is a node that is indirectly associated with the logistics user through the second-level node. For example, supply chain information: if the first-level node is a transaction record, the second-level node is a specific logistics link, and the third-level node may be an upstream supplier or downstream distributor in the supply chain. By expanding the depth of the retrieval subgraph and the logic of sorting the subgraph information, the third-level and deeper nodes can be processed, which makes the analysis more comprehensive and can cover more factors that may affect the overdue payment of logistics users.

[0073] In the above operation, the sub-knowledge graph related to the logistics user is directly located through the entity, which improves the efficiency and accuracy of the query, and can quickly obtain all information closely related to the logistics user; the sub-knowledge graph provides rich contextual information, which helps to more deeply understand and analyze the reasons for the overdue collection; by analyzing the relationship nodes in the sub-knowledge graph, it can be clarified which factors are causally related to the overdue collection, help find the root cause, and obtain detailed attribute data for each relationship node, further deepen the understanding of the problem, and help discover hidden clues to the problem.

[0074] For step S104, all information obtained through the subgraph is comprehensively sorted out, such as summarizing the analysis results of the contract terms, sorting out the problems reflected by the transaction records and customer information, and combining other aspects such as transportation. The sorted information is input into the LLM, which makes inferences and judgments based on this information and generates specific root causes that may lead to overdue payment, such as "the payment time agreement in the contract is not clear, and the goods are damaged during transportation, resulting in delayed payment by the customer" and other reasons.

[0075] Here we explain how LLM infers the reasons for overdue payment from logistics users, especially when it has not received direct training in related fields. LLM's inference process is mainly based on the following aspects:

[0076] 1. Contextual information: LLM uses the input contextual information to make inferences. For example, contextual information includes contracts, transaction records, customer information, logistics information, etc. The model will look for clues that may lead to overdue payment based on this information.

[0077] 2. Knowledge base: LLM is trained on large-scale text data that covers a wide range of fields and topics. Although the model itself is not trained specifically for a certain field, it has been exposed to a large amount of general knowledge and cases during training. It can use this knowledge to make inferences.

[0078] 3. Pattern Recognition: LLM is good at identifying and understanding patterns and relationships in text. For example, the model can identify potential risks in contract terms, abnormal patterns in transaction records, anomalies in customer behavior, etc.

[0079] In the above operations, by analyzing the attribute data of different relationship nodes, we can comprehensively consider the problem from multiple angles to improve the comprehensiveness and effectiveness of the solution; using the preset model to automatically analyze the data reduces the workload of manual analysis and improves efficiency; the model can give the reason for the overdue payment based on the context information, which enhances the credibility and interpretability of the results; the model can be continuously optimized according to new data and feedback to ensure the accuracy and timeliness of the analysis results; the output reasons for the overdue payment can be directly used to guide the subsequent strategy formulation and help enterprises make decisions faster

[0080] The method provided in the above embodiment introduces Graph RAG technology, and realizes efficient and automated processing of overdue payment problems of logistics users through entity recognition, graph database retrieval, causal relationship analysis and model intelligent inference of causes. It not only reduces labor costs, but also improves the accuracy and speed of problem solving, and solves the existing problems of "low efficiency, lack of dynamics and timeliness".

[0081] See also Figure 2 , shows a schematic flow chart of an optional data processing method according to an embodiment of the present invention, comprising the following steps:

[0082] S201: Collect overdue payment data associated with each logistics user from multiple data sources;

[0083] S202: Using the preset duplicate data detection strategy, error data identification strategy and irrelevant data identification strategy, the overdue payment data of each logistics user is processed;

[0084] S203: Obtain a collection of overdue payment data of logistics users, and identify entities and entity data of each entity from each collection of overdue payment data of logistics users;

[0085] S204: receiving an identifier configured for each entity, and obtaining a relationship configured between entities, and generating an entity triple relationship according to the entity identifier and the relationship between entities;

[0086] S205: Acquire the attributes pre-configured for each entity, acquire the attribute value corresponding to the attribute from each entity data, and construct entity attribute data according to the entity identifier, attribute and attribute value;

[0087] S206: Create an entity table for each entity in the data warehouse, and store the entity triple relationship and entity attribute data of each entity in the entity table of each entity.

[0088] In the above implementation, for step S201, this step is used to collect data, such as data related to merchants and overdue payment from multiple sources such as the internal retail ordering system, waybill system, billing and settlement system, and credit reporting system of the logistics department, including but not limited to the following aspects:

[0089] 1. Logistics user information: including the basic information, business scope, credit rating, credit limit, etc. of the merchants in the internal logistics system.

[0090] 2. Product data: product name, quantity, specifications, unit price, etc., or specific content and billing standards of the service.

[0091] 3. Contract information: including the contract terms, service content, price, etc. signed with the logistics user.

[0092] 4. Collection plan: including the time, amount, method, etc. of collection.

[0093] 5. Merchant waybill information: waybill number, shipping address, receiving address, shipping time, estimated delivery time, actual delivery time, freight amount, transportation status, etc.

[0094] 6. Mall rebate: Merchants who use the same company’s logistics services for mall retail need to obtain data on the rebates given by the mall retail to the merchants, such as sales data, rebate ratio or amount data, rebate calculation cycle data, rebate settlement status data, promotional activity data, etc.

[0095] 7. Overdue situation: including overdue time, amount, etc.

[0096] 8. Collection records: including the time, method, and results of collection.

[0097] 9. Logistics warehousing and distribution service status: data such as differences in incoming / outgoing / inventory, damaged goods, loss, delivery delays, etc., which will cause disputes between customers and merchants and affect payment collection.

[0098] 10. Billing and settlement: transaction data (amount, time, order number), review data, settlement status data, settlement order number, settlement rule data such as settlement cycle, settlement ratio, deduction items and standards, etc.

[0099] 11. Reconciliation data: verification information of transaction details and actual receipts and payments.

[0100] 12. Cost data: various cost data such as handling fees, commissions, freight, etc.

[0101] 13. Industry credit investigation: mainly includes the following types of information, which can help financial institutions, government agencies and enterprises to have a more comprehensive understanding of the credit status of target enterprises, so as to make more accurate decisions. Enterprise information: basic information of the enterprise, such as name, registration time, registered capital, business scope, etc. Enterprise business data: business status and financial data of the enterprise, such as sales, profits, balance sheet, etc. Government information: legal proceedings information involving the enterprise. Financial / tax / trade information: financial statements, tax information, trade data, etc. of the enterprise. Risk public opinion monitoring: risk events and public opinion information related to the enterprise. Relationship map: the relationship between the enterprise and other enterprises and individuals.

[0102] For step S202, the collected data on overdue payment of logistics users is cleaned to remove duplicate, erroneous or irrelevant information, and then integrated to ensure the accuracy and consistency of the data. For duplicates, two pieces of data are identified as the same, that is, duplicates. For identification errors and irrelevant situations, specific strategies are required. The cleaned data is more reliable and provides a solid foundation for subsequent analysis.

[0103] 1. Identifying erroneous data usually relies on data characteristics, business rules, and statistical methods, such as: 1) Data type and format check. Data type: Make sure the data type of the field is correct, for example, date fields should be in date format and numeric fields should be numeric types. Format: Check whether the data format is consistent, for example, phone numbers should conform to a specific format, and dates should use a unified format (such as YYYY-MM-DD). 2) Value range and logical consistency. Value range: Check whether the numeric field is within a reasonable range, for example, the age field should not be negative or exceed a reasonable upper limit. Logical consistency: Check the logical relationship between the data, for example, the order date should not be later than the shipping date, and the end time should not be earlier than the start time. 3) Business rules and outlier detection. Business rules: Check the data according to business logic and rules, for example, the price field should be positive. Outlier detection: Use statistical methods (such as standard deviation, box plots) to detect outliers, for example, sales that are much higher or lower than the average may be erroneous data.

[0104] 2. Irrelevant data refers to information that is useless for current analysis or business needs. The basis for identifying irrelevant data includes business needs, data correlation and statistical analysis. For example: 1) Business needs. Clarify the goal: According to the goal of data analysis, determine which fields and records are necessary and which are redundant. Demand analysis: Communicate with the business team to understand which data is valuable for business decisions. 2) Data correlation analysis: Use statistical methods (such as Pearson correlation coefficient, chi-square test) to analyze the correlation between fields. Fields with low correlation may be irrelevant data. 3) Data redundancy. Duplicate records: Identify and delete duplicate records through unique identifiers (such as ID). Redundant fields: Delete duplicate or redundant fields. For example, multiple fields containing the same information.

[0105] For steps S203 and S204, after obtaining the relevant data on overdue payment from the logistics user, key entities such as merchants, orders, goods, transportation tools, locations, etc. are identified. Each entity should have a unique identifier, including the following entities:

[0106] 1. Logistics users: represent enterprises or individuals that provide logistics services.

[0107] 2. Commodities: specific items or services.

[0108] 3. Contract: The contract signed between the merchant and the logistics user.

[0109] 4. Waybill: The waybill number certificate for logistics transportation.

[0110] 5. Mall: a platform for retail transactions.

[0111] 6. Billing and settlement platform: Merchants using logistics services perform billing settlement and reconciliation on this platform.

[0112] 7. Logistics users: entities that use logistics services or shop in the mall.

[0113] 8. Industry credit reporting related entities: such as enterprises, courts, etc.

[0114] After that, we can define the triple relationship <entity 1, relationship, entity 2> between entities. For example, there is an "order" relationship between customers and orders, a "contains" relationship between orders and goods, and a "transported" relationship between goods and transportation tools. These relationships can be pre-defined. Since overdue data is constantly updated, these pre-defined triple relationships can be used directly to reduce the dependence on staff.

[0115] 1. “Logistics user - providing logistics services - goods”: means that the logistics service provider provides transportation and other services for the goods.

[0116] 2. “Logistics user - contract signing - other parties”: reflects the relationship between merchants in signing contracts.

[0117] 3. “Contract - including terms - service content, price, etc.”: describes the specific contents of the contract.

[0118] 4. “Logistics User-Owned-Waybill”: indicates the ownership relationship between the waybill and the logistics user.

[0119] 5. “Mall - Rebate to - Logistics User”: describes the mall’s rebate behavior to logistics users.

[0120] 6. “Logistics user - overdue situation - overdue time and amount”: displays the merchant’s overdue related information.

[0121] 7. “Logistics user - collection record - collection time, method, and result”: related collection data.

[0122] 8. “Logistics users-involved-logistics warehousing and distribution service situation”: reflects the connection between merchants and warehousing and distribution services.

[0123] 9. “Transaction-association-billing and settlement”: explains the relationship between transactions and billing and settlement.

[0124] 10. “Transaction-Generate-Reconciliation Data”: indicates that the transaction results in the generation of reconciliation data.

[0125] 11. “Transaction-Including-Fee Data”: The transaction includes various fees.

[0126] 12. “Logistics users-corresponding-industry credit”: connect logistics users with relevant credit information.

[0127] 13. “Industry credit investigation - including - corporate information, operating data, etc.”: indicates the specific categories of information included in the credit investigation.

[0128] For step S205, obtain the attributes pre-configured for each entity, such as the logistics user entity includes attributes such as credit score and historical transaction records; the order entity includes attributes such as amount and order time. Then, based on the attribute, obtain the attribute value from the entity data. This is because the attribute value already exists in the data, but it may not be known which attribute it corresponds to. By setting the attribute and obtaining the attribute value from the relevant data of the entity based on the attribute, and constructing <entity, attribute, attribute value>, data can be managed more systematically and accurately. Considering that the overdue payment data in the database is constantly updated, in order to reduce dependence on manual work, attributes can be set for each entity in advance and updated regularly.

[0129] For step S206, this solution is mainly applied to the big data platform. The big data platform first creates an entity table for each entity in the data warehouse. For example, if the entity has Customer, Order, Payment, and OverdueRecord, it is necessary to create a customer table, an order table, a payment table, and an overdue record table. For each entity table, the entity triple relationship and entity attribute data of the entity need to be stored in the entity table, for example, the customer triple relationship and attribute data need to be stored in the entity table.

[0130] The method provided in the above embodiment ensures the comprehensiveness and accuracy of information by collecting overdue payment data from multiple data sources. The multi-angle data coverage helps to have a deeper understanding of the overdue payment problem. By identifying the entities in the overdue payment data and establishing triple relationships between entities, the connection between the entities can be clearly displayed, which helps to deeply analyze the root causes of overdue payment. Configuring attributes according to entity characteristics and obtaining corresponding attribute values ​​from entity data enhances the flexibility and practicality of the knowledge graph. An entity table is created for each entity in the data warehouse, and the triple relationship and attribute data of the entity are stored. This structured storage facilitates data query and maintenance and improves data analysis efficiency. The overall system can not only effectively sort out and refine the complex factors related to overdue payment, but also facilitate the construction of a practical and clearly structured knowledge graph, providing strong support for solving the problem of overdue payment.

[0131] See also Figure 3 , shows a schematic flow chart of another optional data processing method according to an embodiment of the present invention, comprising the following steps:

[0132] S301: Determine a configuration file for the import tool; wherein the configuration file uses the data warehouse as the data source and the graph database as the data storage target;

[0133] S302: Read data from each entity table of the data warehouse through the data framework access interface of the import tool; wherein the entity table is used to store entity triple relationships and entity attribute data;

[0134] S303: Convert the entity triple relationship into labels and edges that meet the requirements of the graph database through the mapping relationship in the configuration file, and convert the entity attribute data into label attribute data;

[0135] S304: Import labels, edges and label attribute data into the graph database, so that the graph database uses labels as nodes and label attribute data as node attribute data, combined with edges, to construct a knowledge graph.

[0136] In the above implementation, after generating the entity table of overdue payment for logistics users, the related entities, relationships, and attribute data involved in the overdue payment are stored in the Nebula Graph database for subsequent Graph RAG. Since the logistics industry involves data sources from many links, from order placement, distribution to billing, after-sales production systems, etc., each link generates a large amount of data. Direct batch query of each production system will generate a large number of IO operations that affect the system. Therefore, the production system will be extracted from the library data to the big data platform (hive table) at the T+1 time point, and the entity, relationship, and attribute knowledge graph data will be sorted on the big data platform and reflected using the hive table. For example, for Customer, Order, Payment, OverdueRecord, it is necessary to create a customer table, order table, payment table, and overdue record table to include entity triple relationships and entity attribute data.

[0137] For steps S301 to S303, create a Spark Writer configuration file, specify the label configuration information and edge configuration information to be imported, and the mapping relationship with the Hive table. Spark Writer is used to write data in Hive to Nebula Graph. For example, the configuration file contains the Nebula Graph service address, user name and password, database name, Spark related configuration (such as application name and running mode), and label and edge configuration information. The service address includes the address of the Graph service and the Meta service. Graph is the address of the graph storage service, and meta is the address of the metadata service. Then, the graph database is determined by the service address and database name, and then the user name and password are used to establish a communication link with the graph database.

[0138] In this solution, there is a one-to-one correspondence between tags and entities. They are essentially the same, but the names are different in different environments. As described above, the entities include Customer, Order, Payment, and OverdueRecord, so the tags include Customer, Order, Payment, and OverdueRecord. The corresponding edges are CustomerOrder (Customer -> Order), OrderPayment (Order -> Payment), and OrderOverdue (Order -> OverdueRecord).

[0139] In Nebula Graph, labels represent vertices, and edges represent the relationship between vertices. You need to specify these labels and edges in the configuration file, as well as their mapping relationship with the Hive table. With such a configuration file, you can write the data in Hive into Nebula Graph according to the defined labels and edge relationships, and use it to build the data structure of the graph database.

[0140] For example, in the tag configuration information, the name, data source and target, field mapping relationship, and unique identifier of each tag are defined. For example, the configuration of the four tags Customer, Order, Payment, and OverdueRecord mentioned above. Including but not limited to, name: the name of the vertex tag. type: specifies the data source and target, here is to read data from Hive and import it into Nebula Graph. fields: the list of fields in the Hive table. nebula.fields: the corresponding list of fields in Nebula Graph, such as the field list of tag attributes. The mapping relationship between entity table fields and tag attribute fields.

[0141] For example, in the edge configuration information, the name, data source and target, field mapping relationship, and the association relationship between the source vertex and the target vertex of each edge are defined. For example, the configuration of the three edges CustomerOrder, OrderPayment, and OrderOverdue are included. Including but not limited to, name: edge name. type: also specifies the data source and target. fields: a list of fields in the Hive table. nebula.fields: a list of fields in the corresponding Nebula Graph. source and target: define the starting label name and the end label name of the edge. ranking: the field used to sort the edges, usually the end field of the edge. The starting label name and the end label name can be generated based on the entity triple relationship.

[0142] Use Spark's DataFrame API to read data from Hive tables. The DataFrame API makes it simple and efficient to read data from data warehouses. According to the mapping relationship in the configuration file, convert the data in the DataFrame into Nebula Graph labels and edges. In Nebula Graph, the data model is based on graph theory, where labels represent entities (such as customers, orders, etc.) and edges represent the relationship between labels (such as customers placing orders, order payments, etc.). Therefore, the conversion process includes:

[0143] 1. Vertex conversion: Select the columns in the DataFrame that correspond to the vertex attributes and construct the vertex. For example, for the "Customer" vertex, the "CustomerID", "Name", "ContactInfo", and "CreditRating" columns in the DataFrame will be used to construct the vertex.

[0144] 2. Edge conversion: Select the columns in the DataFrame that correspond to the edge attributes and construct the edge object. For example, the "CustomerOrder" edge contains "CustomerID" and "OrderID" columns, which will be used to define the source and target vertices of the edge.

[0145] You can use Spark's various transformation methods, such as map() or flatMap(), or more advanced DataFrame methods such as withColumn() and groupBy() to ensure that the data format meets Nebula Graph's requirements. Typically, this conversion involves writing some custom logic or using specialized libraries, such as NebulaGraph's Spark Connector, which provides convenient tools and functions to simplify the process, allowing you to directly convert DataFrames into the format of nodes and edges that Nebula Graph can understand, and then use Spark's write function to batch import data into the Nebula Graph database.

[0146] For step S304, use Spark Writer to send the converted data to Nebula Graph for insertion. The graph database then uses labels as nodes and relationships between entities as edges. The nodes in the knowledge graph are connected through relationship nodes to form a large graph. For some of the graphs, see Figure 4 The specific entities and relationship details are described as follows:

[0147] 1. The central node is "Logistics System", from which extends the "Logistics User" node, which includes "Merchant Name" (such as XX Merchant Company), "Merchant Address" (specific address), "Contact Information" (telephone, email, etc.), "Establishment Time", "Enterprise Size" (large, medium, small, etc.), "Enterprise Culture" (corporate values, mission, etc.), "Technical Strength" (advancedness of logistics management system, etc.), "Quota" and other sub-nodes;

[0148] 2. The "Contract" node is connected to the "Logistics User", including the "Contract Number" (specific number), "Contracting Party" (specific company or individual), and "Contract Terms" (such as "Service Content" including detailed service methods such as land transportation, air transportation, warehousing, and distribution, "Price" including detailed pricing such as basic freight and additional fees, and "Responsibility Division" such as the different risk bearers during transportation, etc.);

[0149] 3. "Logistics User" is also connected to the "Waybill" node, including "Waybill Number" (specific number), "Shipping Time" (specific date and time), "Transportation Route" (detailed information such as starting point, waypoint, end point, and whether it is a dedicated line, etc.), "Cargo Status" (shipped, in transit, arrived, signed for, etc.), and is also associated with the "Logistics Vehicle" node, which includes "Vehicle Model" (specific model), "License Plate Number" (specific license plate number), "Vehicle Status" (normal, repair, maintenance, etc.), "Vehicle Ownership" (self-owned, leased, etc.), "Vehicle Fuel Consumption" (specific value), "Vehicle Maintenance Record" (maintenance time, maintenance items, etc.);

[0150] 4. It is also connected to the "Logistics Personnel" node, with "Personnel Name" (specific name), "Work Number" (specific number), "Position Responsibilities" (driver, loader, dispatcher, etc.), "Work Experience" (specific years), and "Training Records" (training content, time, etc.);

[0151] 5. The "Logistics User" is connected to the "Commodity" node, and has "Commodity ID" (specific code), "Commodity Name" (specific name), "Commodity Specification" (size, weight, volume, etc.), "Commodity Attribute" (fragile, moisture-proof, flammable, etc.), "Commodity Value" (high, medium, low, etc.), "Commodity Packaging Requirements" (special packaging methods, etc.);

[0152] 6. Connected to the "mall" node, including "mall name" (XX mall), "mall scale" (land area, number of stores, average daily customer flow, etc.), "rebate relationship" connects the two, including "rebate amount" (specific value), "rebate time" (specific date), "rebate policy" (such as full amount rebate, specific product rebate, etc.); connected to the "order" node, including "amount", "order time", etc.;

[0153] 7. "Logistics User" is connected to the "Overdue Record" node, including "Overdue Time" (specific time point), "Overdue Amount" (specific amount), "Overdue Reason" (such as broken capital chain, poor management, etc.), and is related to the "Collection Record" node, including "Collection Time" (specific time), "Collection Method" (phone, letter, door-to-door, etc.), "Collection Result" (repaid, partially repaid, not repaid, negotiated repayment, etc.);

[0154] 8. "Logistics User" is connected to the "Logistics Warehousing and Distribution Service" node, including "Warehouse Information" ("Warehouse Location" (specific address), "Warehouse Area" (specific value), "Warehouse Layout" (regional division, etc.), "Warehouse Facilities" (shelves, forklifts, etc.), "Warehouse Temperature and Humidity Control" (whether there are temperature control equipment, etc.), "Delivery Information" ("Delivery Area" (detailed regional division), "Delivery Time" (promised time in different regions), "Delivery Method" (express delivery, logistics dedicated line, etc.), such as "Warehouse Location" (specific address), "Delivery Area" (division by city, region, etc.), etc.;

[0155] 9. "Logistics User" is connected to the "Transaction" node, including "Transaction ID" (specific number), "Transaction Time" (specific time), "Transaction Amount" (specific amount), "Billing and Settlement" node related, including "Billing Method" (by weight, volume, number of tickets, etc.), "Settlement Time" (date of the month, day of the week, etc.), "Reconciliation Data" node connected, including "Reconciliation Time" (specific time), "Reconciliation Difference" (specific difference content and amount), "Expense Data" node connected, including "Freight" (specific amount), "Packaging Fee" (specific amount), "Loading and Unloading Fee" (specific amount), "Storage Fee" (specific amount), etc.;

[0156] 10. "Logistics users" are connected to the "Industry Credit" node, under which are "Enterprise Credit Data" (such as "Credit Rating" (AAA, AA, A, etc.), "Operating Status" (profit, loss, break-even), "Court Judgment Information" (such as "Judgment Result" win, loss, etc., "Judgment Time" specific date), and "Other Related Information" (such as "Related Enterprise" name, "Related Person" name, etc.).

[0157] It should be noted that the process of generating knowledge graphs provided by this solution is similar to the existing technology, such as identifying entities, identifying the relationship between entities, and building a graph structure. However, there are some unique differences in building a logistics user graph, especially in terms of data sources and data acquisition. The logistics industry involves many links, from ordering, distribution to billing, after-sales, etc., and each link generates a large amount of data. These data are not only of various types, but also come from various sources, including order systems, waybill systems, transportation management systems, billing systems, after-sales systems, etc. To build a comprehensive logistics user knowledge graph, it is necessary to integrate and process data from these different sources to provide a more comprehensive and detailed display of each link in the logistics chain.

[0158] The method provided in the above embodiment defines data migration rules by configuring the configuration file of the import tool, simplifies the data migration process, and can easily respond by changing the configuration file when the demand changes. It is also easy to expand and maintain. The data format conversion is completed automatically through the configuration file, which reduces the errors and workload of manual operations and improves the speed and accuracy of data processing. The graph database constructs an efficient and easy-to-use knowledge graph in the form of labels and edges, which can support complex query requirements. The overall provides strong support for subsequent data analysis and decision support.

[0159] See also Figure 5 , shows a schematic flow chart of a specific data processing method according to an embodiment of the present invention, comprising the following steps:

[0160] 1. Collect overdue payment data associated with each logistics user from multiple data sources; the data here includes structured data, unstructured data, and semi-structured data, which are the three main data types in the field of data management and analysis.

[0161] Structured data usually refers to data organized in a tabular form with a fixed schema and structure, such as an order table, a customer table, an Excel table, etc. Semi-structured data is between fully structured data and fully unstructured data. It has a certain structure but is not as strict as structured data. Semi-structured data usually contains tags or names to represent different parts of the data, such as XML files, JSON files, HTML documents. Unstructured data has no predefined schema or structure and is usually difficult to process and analyze directly. For example, text files such as emails, news reports, and social media posts. Image, audio and video files. Scanned documents such as PDF files, especially those that contain handwritten or non-text elements.

[0162] 2. Use the preset duplicate data detection strategy, error data identification strategy and irrelevant data identification strategy to process the overdue payment data of each logistics user and obtain the overdue payment data set of the logistics user;

[0163] 3. Identify entities and entity data of each entity from the overdue payment data of each logistics user;

[0164] 4. Receive the identifier configured for each entity, and obtain the relationship configured between the entities, and generate the entity triple relationship according to the entity identifier and the relationship between the entities;

[0165] 5. Obtain the attributes pre-configured for each entity, obtain the attribute value corresponding to the attribute from each entity data, and construct entity attribute data based on the entity identifier, attribute and attribute value;

[0166] 6. Create an entity table for each entity in the data warehouse, and store the entity triple relationship and entity attribute data of each entity in the entity table of each entity;

[0167] 7. Use the import tool to import the data in the data warehouse entity table into the graph database to generate a knowledge graph in the graph database;

[0168] 8. The staff inputs the consultation information about the overdue payment of the logistics user;

[0169] 9. Perform text analysis on the consulting information, and use a preset entity library to identify entities from the text analysis results;

[0170] 10. Search the graph database according to the identified entity to obtain a sub-knowledge graph associated with the logistics user;

[0171] 11. Determine, from the sub-knowledge graph, the relationship nodes that have a causal relationship with the logistics user, and obtain node attribute data of each relationship node;

[0172] 12. Input the relationship nodes and node attribute data into the preset model; wherein the preset model constructs context information based on the association relationship between the relationship nodes and the node attribute data, and determines the reason for the overdue payment based on the context information;

[0173] 13. The reason for overdue payment output by the preset model will be used as the reply information and displayed.

[0174] This solution models the technical problems to be solved. The root cause analysis method of overdue payment for logistics users based on LLM and knowledge graph is modeled to replace the manual optimization problem. It aims to find as many root causes of overdue payment for logistics users as possible under the same time limit T and with different business lines and product line customer requirements, and find the best application strategy to improve the accuracy and efficiency of the root cause analysis of overdue payment for logistics users.

[0175] Assumptions (T) represents the comprehensive efficiency function of this scheme within T time, which is a function of time; (T) represents the comprehensive efficiency function of the traditional manual tracking of overdue reasons within the same time T, and the efficiency ratio is used to represent the degree of optimization. The optimization index is defined as C = (T) / (T), when C is greater than 1 and the value is larger, it means that the efficiency of this solution is more obvious than that of manual work.

[0176] The comprehensive efficiency function of this solution includes the following two considerations: E(T)=A(T) × S(T), where A(T) is the accuracy of the root cause analysis method for overdue payment collection by logistics users within time T, such as the correct judgment ratio for the root cause analysis of overdue payment collection by logistics users. S(T) is the speed of analysis within time, such as the number of cases that can be processed per hour. For example, A(T)= 0.9 (indicating an accuracy rate of 90%), S(T)= 50 (indicating 50 cases processed per hour), then the value of the comprehensive efficiency function can be calculated to be 45.

[0177] From the efficiency comparison table, we can see that in terms of analysis accuracy, the sales staff’s inquiry method using Graph RAG is slightly higher than that of manual sales visits, at 0.95 to 0.9. In terms of analysis speed, the sales staff’s inquiry method using Graph RAG has a clear advantage, and can analyze 150 merchants per day, while manual sales visits can only visit 10 merchants per day. In terms of comprehensive efficiency evaluation, the large model reached 142.5, far higher than the 9 of manual sales visits, which shows that this solution Graph RAG has a greater advantage than manual sales in terms of efficiency in handling the root cause of overdue payment from logistics users.

[0178]

[0179] Assuming that the average number of overdue merchants per month is 5,700, according to "sales personnel have obvious advantages with the inquiry method of Graph RAG, and can analyze 150 merchants every day", it can be known that it takes 5,700 / 150=38 days. The days here are different from the days we usually understand. The calculation is that it takes 38 days for a salesperson to handle these 5,700 overdue issues with the help of Graph RAG. In actual operations, there is usually not only one salesperson. For example, there are two people, A and B. A takes 19 days to handle, and B takes 19 days to handle, and the total is 38 days. The sales labor requires 5,700 / 10=570 days. In terms of years, Graph RAG takes 38*12=456 days to handle the reasons for overdue merchants. Similarly, the days here are the cumulative processing time of all sales personnel. Assuming that the cost of processing one day is 1,000 yuan, the estimated labor cost savings through this solution are: 6,840*1,000 - 456*1,000 = 6.384 million yuan / year.

[0180]

[0181] The method provided by the embodiment of the present invention can solve the problems of inefficiency and untimeliness in the traditional manual processing of overdue payment data analysis. Specifically: Graph RAG technology is used to build a complete knowledge graph of overdue payment for logistics users and integrate it into the LLM technology stack. The level of contextual learning is improved with the help of retrieval and generation core technologies to assist the large model in outputting the root causes of overdue payment for merchants, effectively reducing the heavy workload of manual visits and data analysis, so as to formulate strategies and methods for subsequent merchant payment collection and improve the business payment collection rate.

[0182] See also Figure 6 , showing a schematic diagram of the main modules of a data processing device 600 provided by an embodiment of the present invention, including: an entity recognition module 601, used to receive consultation information about overdue payment of a logistics user, perform text analysis on the consultation information, and use a preset entity library to identify entities from the text analysis results; a graph retrieval module 602, used to search the graph database according to the identified entities to obtain a sub-knowledge graph associated with the logistics user; a node determination module 603, used to determine the relationship nodes that have a causal relationship with the logistics user from the sub-knowledge graph, and obtain the node attribute data of each relationship node; a cause analysis module 604, used to input the relationship nodes and node attribute data into a preset model, and use the overdue payment reason output by the preset model as reply information and display it; wherein the preset model constructs context information according to the association relationship between the relationship nodes and the node attribute data, and determines the overdue payment reason according to the context information.

[0183] In the implementation device of the present invention, the node determination module 603 is used to: obtain the retrieval depth, taking the node where the logistics user is located as the starting point, determine the first-level relationship node directly associated with the logistics user in the sub-knowledge graph; taking the first-level relationship node as the starting point, determine the second-level relationship node in the sub-knowledge graph that is directly associated with the first-level relationship node and indirectly associated with the logistics user; repeat the above-mentioned operations of determining the starting point and relationship node until the relationship node of the retrieval depth is determined.

[0184] In the implementation device of the present invention, the sub-knowledge graph is a part of the knowledge graph in the graph database, and the device also includes a knowledge graph construction module, which is used to: determine a configuration file for configuring an import tool; wherein the configuration file uses a data warehouse as a data source and a graph database as a data storage target; read data from each entity table in the data warehouse through the data framework access interface of the import tool; wherein the entity table is used to store entity triple relationships and entity attribute data; through the mapping relationship in the configuration file, the entity triple relationships are converted into labels and edges that meet the requirements of the graph database, and the entity attribute data is converted into label attribute data; the labels, edges and label attribute data are imported into the graph database, so that the graph database uses labels as nodes and label attribute data as node attribute data, combined with edges, to construct a knowledge graph.

[0185] The implementation device of the present invention also includes a module for constructing an entity table, which is used to: obtain a set of overdue payment data of logistics users, identify entities and entity data of each entity from the overdue payment data of each logistics user; receive an identifier configured for each entity, and obtain the relationship configured between entities, and generate an entity triplet relationship based on the entity identifier and the relationship between entities; obtain the attributes pre-configured for each entity, obtain the attribute value corresponding to the attribute from each entity data, and construct entity attribute data based on the entity identifier, attribute and attribute value; create an entity table for each entity in the data warehouse, and store the entity triplet relationship and entity attribute data of each entity in the entity table of each entity.

[0186] In the implementation device of the present invention, the entity table construction module is used to: collect overdue payment data associated with each logistics user from multiple data sources; use preset duplicate data detection strategies, error data identification strategies and irrelevant data identification strategies to process the overdue payment data of each logistics user to obtain a set of logistics user overdue payment data.

[0187] In the implementation device of the present invention, the knowledge graph construction module is used to: receive the link information of the graph database configured in the configuration file; receive the label configuration information configured in the configuration file; wherein the label configuration information includes the label name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the field list of the entity table, the field list of the label attribute, and the mapping relationship between the entity table fields and the label attribute fields; receive the edge configuration information configured in the configuration file; wherein the edge configuration information includes the edge name, the name of the data warehouse as the data source, the name of the graph database as the data import target, the starting label name of the edge, and the ending label name.

[0188] In the implementation device of the present invention, the link information includes the service address, database name, user name and password. The knowledge graph construction module is also used to: determine the graph database through the service address and database name of the graph database in the configuration file, and then use the user name and password to establish a communication link with the graph database.

[0189] In addition, the specific implementation content of the device described in the embodiment of the present invention has been described in detail in the method described above, so the repeated content will not be described again here.

[0190] Figure 7 An exemplary system architecture 700 to which embodiments of the present invention may be applied is shown, including terminal devices 701 , 702 , 703 , a network 704 and a server 705 (only an example).

[0191] Terminal devices 701, 702, 703 can be various electronic devices with display screens and supporting web browsing, and various communication client applications are installed. Users can use terminal devices 701, 702, 703 to interact with server 705 through network 704 to receive or send messages, etc.

[0192] The network 704 is used to provide a medium for communication links between the terminal devices 701, 702, 703 and the server 705. The network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0193] The server 705 may be a server that provides various services, such as a backend management server that provides support for shopping websites browsed by users using terminal devices 501, 502, and 503 (for example only). The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - for example only) to the terminal device. It should be noted that the method provided in the embodiment of the present invention is generally executed by the server 705, and accordingly, the device is generally set in the server 705.

[0194] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0195] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing an embodiment of the present invention. Figure 8 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0196] like Figure 8As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0197] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0198] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0199] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0200] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0201] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor including an entity recognition module, a graph retrieval module, a node determination module, and a cause analysis module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the cause analysis module may also be described as a "cause analysis module."

[0202] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device executes any of the above data processing methods.

[0203] The computer program product of the present invention comprises a computer program, and when the computer program is executed by a processor, the data processing method in the embodiment of the present invention is implemented.

[0204] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data processing method, characterized in that: include: Receive consultation information regarding overdue payment from a logistics user, perform text analysis on the consultation information, and use a preset entity library to identify entities from the text analysis results; Searching a graph database according to the identified entity to obtain a sub-knowledge graph associated with the logistics user; Determine, from the sub-knowledge graph, the relationship nodes that have a causal relationship with the logistics user, and obtain node attribute data of each relationship node; The relationship nodes and node attribute data are input into the preset model, and the reason for overdue payment output by the preset model is used as the reply information and displayed; wherein the preset model constructs context information according to the association relationship between the relationship nodes and the node attribute data, and determines the reason for overdue payment according to the context information.

2. The method according to claim 1, characterized in that The process of determining the relationship node that has a causal relationship with the logistics user is as follows: Obtaining the search depth, taking the node where the logistics user is located as the starting point, determining the first-level relationship node directly associated with the logistics user in the sub-knowledge graph; Taking the first-level relationship node as a starting point, determining the second-level relationship node in the sub-knowledge graph that is directly associated with the first-level relationship node and indirectly associated with the logistics user; Repeat the above operations of determining the starting point and relationship nodes until the relationship node of the search depth is determined.

3. The method according to claim 1, characterized in that The sub-knowledge graph is a part of the knowledge graph in the graph database. The process of constructing the knowledge graph is as follows: Determine the configuration file for the import tool; in which the configuration file uses the data warehouse as the data source and the graph database as the data storage target; Read data from each entity table in the data warehouse through the data framework access interface of the import tool; the entity table is used to store entity triple relationships and entity attribute data; Through the mapping relationship in the configuration file, the entity triple relationship is converted into labels and edges that meet the requirements of the graph database, and the entity attribute data is converted into label attribute data; Import labels, edges and label attribute data into the graph database, so that the graph database uses labels as nodes and label attribute data as node attribute data, combined with edges, to build a knowledge graph.

4. The method according to claim 3, characterized in that The process of building an entity table in a data warehouse is: Obtain a collection of overdue payment data from logistics users, and identify entities and entity data of each entity from each logistics user's overdue payment data; Receive an identifier configured for each entity, and obtain a relationship configured between entities, and generate an entity triple relationship according to the entity identifier and the relationship between entities; Obtain the attributes pre-configured for each entity, obtain the attribute value corresponding to the attribute from each entity data, and construct entity attribute data based on the entity identifier, attribute and attribute value; An entity table is created for each entity in the data warehouse, and the entity triple relationship and entity attribute data of each entity are stored in the entity table of each entity.

5. The method according to claim 4, characterized in that The acquisition of the overdue payment data set of the logistics user includes: Collect overdue payment data associated with each logistics user from multiple data sources; The preset duplicate data detection strategy, error data identification strategy and irrelevant data identification strategy are used to process the overdue payment data of each logistics user and obtain a set of overdue payment data of the logistics users.

6. The method according to claim 3, characterized in that The configuration file for determining the configuration of the import tool includes: Receive the link information of the graph database configured in the configuration file; Receive tag configuration information configured in a configuration file; wherein the tag configuration information includes a tag name, a data warehouse name as a data source, a graph database name as a data import target, a field list of an entity table, a field list of tag attributes, and a mapping relationship between entity table fields and tag attribute fields; Receive edge configuration information configured in a configuration file; wherein the edge configuration information includes an edge name, a data warehouse name as a data source, a graph database name as a data import target, a start label name and an end label name of the edge.

7. The method according to claim 6, characterized in that The link information includes a service address, a database name, a user name and a password, and the method further includes: Determine the graph database through the service address and database name of the graph database in the configuration file, and then use the username and password to establish a communication link with the graph database.

8. A data processing device, characterized in that: include: An entity recognition module is used to receive consultation information about overdue payment from logistics users, perform text analysis on the consultation information, and use a preset entity library to identify entities from the text analysis results; A graph retrieval module is used to search the graph database according to the identified entity to obtain a sub-knowledge graph associated with the logistics user; A node determination module, used to determine, from the sub-knowledge graph, the relationship nodes that have a causal relationship with the logistics user, and to obtain node attribute data of each relationship node; The reason analysis module is used to input the relationship nodes and node attribute data into the preset model, and display the reason for the overdue payment output by the preset model as reply information; wherein the preset model constructs context information based on the association relationship between the relationship nodes and the node attribute data, and determines the reason for the overdue payment based on the context information.

9. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images