A method, apparatus, and equipment for processing multi-source heterogeneous procurement data.

Abstract

This invention provides a method, apparatus, and device for processing multi-source heterogeneous procurement data. The method includes: acquiring raw procurement data from multiple heterogeneous data sources; performing dynamic semantic mapping processing on the raw procurement data according to preset concept relationship mapping rules to obtain a first knowledge tripartite combination set; performing data filtering processing on the first knowledge tripartite combination set to obtain a second knowledge tripartite combination set; determining a procurement knowledge graph based on the second knowledge tripartite combination set; and obtaining procurement intelligent analysis results based on the procurement knowledge graph. This invention achieves automated semantic fusion of multi-source heterogeneous data, providing real-time, accurate, and interpretable data services for upper-level procurement applications.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method, apparatus and equipment for processing multi-source heterogeneous procurement data. Background Technology

[0002] In the context of digital procurement, enterprise data environments exhibit typical VUCA (Volatile, Uncertain, Complex, Ambiguous) characteristics. Traditional ETL (Extract, Transform, Load) and data warehousing paradigms, due to their strong pre-existing assumptions about data patterns, preference for structured data, and emphasis on non-real-time batch processing, face inherent limitations such as rigid architecture, semantic gaps, and response delays when handling multimodal, highly dynamic, and low-quality raw data streams.

[0003] Current mainstream practices mostly employ data integration middleware based on predefined patterns or traditional data lakes driven by batch processing. The former requires hard-coded transformation logic for each data source-target system pair, leading to a combinatorial explosion in system coupling and poor maintainability; while the latter can store raw data, it lacks effective and automated data governance and semantic enhancement capabilities, and cannot directly serve upper-layer intelligent applications. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a method, apparatus, and equipment for processing multi-source heterogeneous procurement data. It achieves automated semantic fusion of multi-source heterogeneous data, constructs a unified, high-quality data asset foundation, and decomposes complex business logic (such as supplier evaluation and risk identification) into elastically scalable microservices through knowledge graphs, thereby providing real-time, accurate, and interpretable data services for upper-level procurement applications.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, embodiments of the present invention provide a method for processing multi-source heterogeneous procurement data, including: Obtain raw procurement data from multiple heterogeneous data sources; According to the preset concept relationship mapping rules, the original procurement data is subjected to dynamic semantic mapping processing to obtain the first knowledge ternary combination set; The first knowledge ternary combination set is subjected to data filtering processing to obtain the second knowledge ternary combination set; Based on the second knowledge ternary combination set, determine the procurement knowledge graph; Based on the procurement knowledge graph, the procurement intelligent analysis results are obtained.

[0006] Optionally, raw procurement data from multiple heterogeneous data sources can be obtained, including: The original procurement data is obtained from multiple preset data sources through protocol plugins and connector driver architecture, including relational databases, enterprise resource planning systems, office automation systems, and file systems.

[0007] Optionally, based on a pre-defined concept relation mapping model, the original procurement data undergoes dynamic semantic mapping processing to obtain a first knowledge ternary combination set, including: The unstructured text in the original procurement data is converted to a new format to obtain the first structured data; The existing second structured data and the first structured data in the original procurement data are parsed to obtain source data pattern elements; Based on the source data pattern elements and the target elements in the concept relationship mapping model, determine the mapping similarity data; When the mapping similarity data is greater than or equal to a preset threshold, the first structured data and the second structured data are transformed to obtain a first knowledge triplet set. The first knowledge triplet set includes multiple first knowledge triplets, and each first knowledge triplet includes an entity, an attribute, and a relationship between entities.

[0008] Optionally, the first knowledge ternary combination set is subjected to data filtering processing to obtain a second knowledge ternary combination set, including: Perform quality rule verification on the first knowledge triplet set to determine the confidence level data of each first knowledge triplet in the first knowledge triplet set; Based on the confidence data, a second knowledge triplet set is determined, which includes a first knowledge triplet whose confidence data is greater than a preset threshold.

[0009] Optionally, based on the second knowledge ternary set, a procurement knowledge graph is determined, including: Based on the pre-defined concept relationship mapping rules, the entities in the second knowledge ternary combination set are used as nodes, and the relationships between the entities in the second knowledge ternary combination set are used as edges to determine the procurement knowledge graph.

[0010] Optionally, based on the procurement knowledge graph, the procurement intelligent analysis results are obtained, including: Based on the procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained; Based on the supplier risk warning information and the purchase price warning information, the intelligent procurement analysis results are obtained.

[0011] Optionally, based on the procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained, including: Based on the procurement knowledge graph, historical supply data and current supply data of suppliers can be obtained. By inputting the historical supply data of the suppliers into models with different analytical functions, supplier risk warning information can be obtained; By inputting the supplier's current supply data into the risk prediction model, a purchase price early warning information is obtained.

[0012] Secondly, embodiments of the present invention also provide a multi-source heterogeneous procurement data processing device, comprising: The acquisition module is used to acquire raw procurement data from multiple heterogeneous data sources; The processing module is used to perform dynamic semantic mapping processing on the original procurement data according to preset concept relationship mapping rules to obtain a first knowledge ternary combination set; perform data filtering processing on the first knowledge ternary combination set to obtain a second knowledge ternary combination set; determine the procurement knowledge graph based on the second knowledge ternary combination set; and obtain the procurement intelligent analysis result based on the procurement knowledge graph.

[0013] Thirdly, embodiments of the present invention also provide an electronic device, including one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described above.

[0014] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a program that, when executed by a processor, implements the method described above.

[0015] The above-described solution of the present invention has at least the following beneficial effects: The above-described solution of the present invention acquires raw procurement data from multiple heterogeneous data sources; performs dynamic semantic mapping processing on the raw procurement data according to preset concept relationship mapping rules to obtain a first knowledge tripartite combination set; performs data filtering processing on the first knowledge tripartite combination set to obtain a second knowledge tripartite combination set; determines a procurement knowledge graph based on the second knowledge tripartite combination set; and obtains procurement intelligent analysis results based on the procurement knowledge graph. The present invention achieves automated semantic fusion of multi-source heterogeneous data, constructs a unified, high-quality data asset foundation, and decomposes complex business logic (such as supplier evaluation and risk identification) into elastically scalable microservices, thereby providing real-time, accurate, and interpretable data services for upper-layer procurement applications. Attached Figure Description

[0016] Figure 1This is a flowchart of an embodiment of the multi-source heterogeneous procurement data processing method of the present invention; Figure 2 This is a schematic diagram of an embodiment of a multi-source heterogeneous procurement data processing device according to the present invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] like Figure 1 As shown, an embodiment of the present invention proposes a method for processing multi-source heterogeneous procurement data, including: Step 11: Obtain raw procurement data from multiple heterogeneous data sources; Step 12: According to the preset concept relationship mapping rules, perform dynamic semantic mapping processing on the original procurement data to obtain the first knowledge ternary combination set; Step 13: Perform data filtering on the first knowledge ternary combination set to obtain the second knowledge ternary combination set; Step 14: Determine the procurement knowledge graph based on the second knowledge ternary combination set; Step 15: Obtain the procurement intelligent analysis results based on the procurement knowledge graph.

[0019] For step 11, obtain raw procurement data from multiple heterogeneous data sources, including: The original procurement data is obtained from multiple preset data sources through protocol plugins and connector driver architecture, including relational databases, enterprise resource planning systems, office automation systems, and file systems.

[0020] This embodiment employs a protocol plug-in and connector-driven architecture to construct a lightweight adapter to acquire system data from different node systems within an enterprise and file data from various files. Specifically, this adapter is responsible for protocol negotiation, initial data parsing, and time-based publishing. It supports breakpoint resume, incremental capture, and Change Data Capture (CDC) mechanisms to ensure the reliability and timeliness of data accessed from sources such as ERP, OA systems, and manually uploaded files. This lays the foundation for subsequent automatic ledger generation and real-time data updates, solving the problems of "delayed procurement ledgers" and "untimely procurement archiving."

[0021] System data and file data can be obtained from system databases, ERP systems, group OA systems, API streams, file systems, and IoT time-series data.

[0022] System data refers to data from the company's internal ERP system, such as the procurement system, supplier database system, historical pricing system, and contract storage system. Document data includes procurement plans, technical specifications, tender documents, review reports, and award notices.

[0023] In an optional embodiment of the present invention, in step 12, the original procurement data is subjected to dynamic semantic mapping processing according to a preset concept relationship mapping model to obtain a first knowledge ternary combination set, including: Step 121: Convert the unstructured text in the original procurement data to obtain the first structured data; Step 122: Perform pattern parsing on the existing second structured data and the first structured data in the original procurement data to obtain source data pattern elements; Step 123: Determine the mapping similarity data based on the source data pattern elements and the target elements in the concept relationship mapping model; Step 124: When the mapping similarity data is greater than or equal to a preset threshold, the first structured data and the second structured data are transformed to obtain a first knowledge triplet set. The first knowledge triplet set includes multiple first knowledge triplets, and each first knowledge triplet includes an entity, an attribute, and a relationship between entities.

[0024] In this embodiment, for step 121, the unstructured text originates from various business node systems within the enterprise, such as the ERP procurement system, supplier database system, historical price system, and contract storage system. The data formats are a mixture of structured database table data (such as database tables, Excel, ERP database tables, OA form ledgers), semi-structured data (such as JSON, XML, document messages, web page tables, segmented review reports, and procurement item lists, which lack strict database table constraints, have loose formats, and inconsistent fields), and unstructured data (such as text in web pages).

[0025] First, a full-domain preprocessing operation is performed on unstructured text, sequentially completing four basic processes: data cleaning, duplicate data removal, field conflict resolution, and intelligent inference of missing values. For key numerical fields such as procurement amount, budget amount, cost-saving rate, and service man-days, data format specifications are standardized, and numerical precision, unit system, and filling logic are verified. For business data with missing fields, inference is performed by combining procurement business rules and historical data from the same source. For data with inconsistent field definitions and conflicting content from multiple systems, unified standardization correction is performed according to procurement business standards and specifications.

[0026] Table 1 Supplier List

[0027] Table 2 Procurement Items

[0028] Then, a regular expression is determined based on the fields of the second structured data. This regular expression is used to match semi-structured and unstructured data in the unstructured text, transforming it into the first structured data (such as Tables 1 and 2). The steps for determining the regular expression include: 1. Extract all field names from the existing structured tables (supplier table, procurement project table) to form a standard field set. Extract fields from Table 1 and Table 2 such as "supplier code", "supplier qualification", and "product name". 2. Build a thesaurus and aliases for each standard field. For example, supplier code = supplier number, supplier code, supplier ID; 3. Define the data format of the field based on the field value of the structured data. For example, supplier code: letters + numbers (such as GY001, S001). 4. Generate a regular expression template based on the field name + alias + data format, for example, (?:field name|alias1|alias2)[::\s\t]{0,3}(. The regular expression ?)(?=\n|,|。|$|\s) matches "field name: value", "field name: value", and "field name value". It is a non-greedy match that matches until a newline / comma / period / end, and automatically adapts to various delimiters in the text. 5. Connect the regular expressions generated for each structured field with "|" to form a whole extraction regular expression, which can extract all field values ​​from semi / unstructured data at once.

[0029] Taking the field "Product Name" as an example, the regular expression for this field is: (?:Supplier Code|Supplier Number|Supplier Code|Supplier ID)[::\s]{0,3}([A-Za-z0-9]+); Taking the field "Supplier Code" as an example, the regular expression for this field is: (?:Product Name|Goods Name|Product Name|Goods)[::\s]{0,3}([\u4e00-\u9fa5a-zA-Z0-9]+).

[0030] In this embodiment, for step 122, the values ​​of the fields in the second structured data and the first structured data are used as source data pattern elements.

[0031] In this embodiment, for step 123, according to Determine the mapping similarity data; in, This represents the comprehensive similarity between the i-th source data pattern element and the j-th target element, with a value range of... The closer the value is to 1, the higher the matching degree. This represents the weight of literal similarity, with a value of 0.4, which can be determined by the relevant rules of the company's internal procurement operations. This indicates mapping similarity data. , express and Edit distance (soon) Transform into (Minimum number of insert, delete, and replace operations required) and They are respectively and The length of the characters; express and Semantic similarity can be calculated by transforming core concepts into vectors using the Word2Vec model and then calculating the cosine similarity of these vectors, using the formula: ,in, express and The vector dot product, and They are respectively and The vector magnitude.

[0032] In this embodiment, for step 124, the first structured data and the second structured data are transformed to obtain a first knowledge ternary combination set, including: Each table in the first and second structured data sets is treated as an entity table. Based on the unique primary key of each entity table, the entity is obtained. For example, in the supplier table (e.g., Table 1), procurement project table (e.g., Table 2), contract table, and price table, if the primary keys of Table 1 are "Supplier Code 01", "Supplier Code 02", and "Supplier Code 03", then the entities obtained from Table 1 include "Supplier Code 01", "Supplier Code 02", and "Supplier Code 03". Similarly, if the primary keys of Table 2 are "Project A", "Project B", and "Project C", then the entities obtained from Table 2 include "Project A", "Project B", and "Project C".

[0033] If two entity tables have a common key, a mapping relationship is established between the entities in the two tables. Based on this mapping, a pre-defined association rule is used to determine the primary relationship between the two entities. For example, if Table 1 and Table 2 have a common key "Product Name", then mapping relationships (Supplier Code 01, Project A), (Supplier Code 02, Project B), and (Supplier Code 03, Project C) are established based on this common key. The mapping relationship between entities is determined according to the pre-defined association rule. For example, querying the pre-defined association rule, the primary relationship between "Supplier Code 01" and "Project A" is "Participation", i.e., "Supplier Code 01 participates in Project A", resulting in (Supplier Code 01, Participation, Project A). Similarly, (Supplier Code 02, Participation, Project B) and (Supplier Code 03, Participation, Project C) can be obtained. Information for each entity (primary key) in the entity tables is queried, and this information is integrated to obtain the primary attribute of each entity. For example, if the primary entity data is "Supplier Code 01", the primary attributes obtained are "Qualified", "Product 1", and "Long-term Contract".

[0034] Based on the entity, the first attribute (which includes the attributes of each entity), and the first relation (which is the relationship between two entities), a first knowledge ternary combination set is obtained.

[0035] In an optional embodiment of the present invention, step 13 involves performing data filtering processing on the first knowledge ternary combination set to obtain a second knowledge ternary combination set, including: Step 131: Perform quality rule verification on the first knowledge triplet set to determine the confidence data of each first knowledge triplet in the first knowledge triplet set; Step 132: Based on the confidence data, determine the second knowledge triplet set, which includes the first knowledge triplet whose confidence data is greater than a preset threshold.

[0036] according to Determine the confidence score data for each first knowledge triple. Among them, All quality rules have fixed weights, with a total weight of 1. The corresponding compliance score for a single rule.

[0037] Filter only by confidence level The first knowledge triplet of 0.7 is incorporated into the second knowledge triplet set.

[0038] After this quality enhancement step, the second knowledge ternary set eliminates low-quality ternaries with incomplete entities, illegal relationships, contradictory business logic, and unreliable sources, ensuring that the procurement knowledge graph constructed subsequently is logically consistent and the data is reliable, providing a high-quality knowledge foundation for intelligent procurement analysis.

[0039] In an optional embodiment of the present invention, step 14, determining the procurement knowledge graph based on the second knowledge ternary combination set, includes: Step 141: Based on the preset concept relationship mapping rules, the entities in the second knowledge ternary combination set are used as nodes, and the relationships between the entities in the second knowledge ternary combination set are used as edges to determine the procurement knowledge graph.

[0040] In this embodiment, the specific process of determining the procurement knowledge graph includes: 1. Entity Import: Assign a unique entity ID (format: "entity type_serial number", such as "supplier_001", "project_001", "review_001") to each entity in the second knowledge ternary combination set, label the entity type (such as supplier, project, review, document, etc., consistent with the global ontology), and store the entity ID with the standard name of the entity to ensure the uniqueness and identifiability of the entity; 2. Attribute import: The "entity-attribute-attribute value" triple corresponding to each entity is stored as the entity's attribute information, and the source of the attribute is recorded (e.g., "price 95-100 yuan" is marked as "ERP system + tender document").

[0041] 3. Based on R3 (relation) in the triple, connect the corresponding two entities (C3_subject, C3_object) through relation edges, label the relation edges with the standard name of R3, and record the generation time of the relation. Simultaneously, combine the hierarchical relationships of entities in the global ontology (e.g., "procurement project" includes "engineering project" and "daily operation project," "supplier" includes "qualified supplier" and "supplier under investigation") to classify entities hierarchically, constructing a hierarchical structure for the graph. This ensures the logic and standardization of the graph and provides support for subsequent intelligent statistics (e.g., statistics by project type).

[0042] 4. Persistently store the knowledge graph in a graph database to obtain the knowledge graph.

[0043] In an optional embodiment of the present invention, step 15, obtaining the procurement intelligent analysis result based on the procurement knowledge graph, includes: Step 151: Based on the procurement knowledge graph, obtain supplier risk warning information and procurement price warning information; Step 152: Based on the supplier risk warning information and the purchase price warning information, obtain the intelligent procurement analysis results.

[0044] In an optional embodiment of the present invention, step 151, obtaining supplier risk warning information and procurement price warning information based on the procurement knowledge graph, includes: Step 1511: Obtain the supplier's historical supply data and current supply data by querying the procurement knowledge graph. Step 1512: Input the supplier's historical supply data into a model with different analysis functions to obtain supplier risk warning information; Step 1513: Input the current supply data of the supplier into the risk prediction model to obtain procurement price early warning information.

[0045] In this embodiment, the supplier's historical supply data includes: supplier name (e.g., Bidding Entity A), entity ID (e.g., Supplier_001), registered address, qualification level (e.g., Excellent), cooperation start time (e.g., 2023-01-15), etc. The supplier's current supply data includes: Project F (e.g., currently being fulfilled), Project G (e.g., pending fulfillment), the current progress of Project F (e.g., 80%), the quantity supplied (e.g., 60 units), the remaining quantity supplied (e.g., 40 units), etc.

[0046] Models with different analytical functions can include qualification level assessment models based on logistic regression algorithms, performance capability analysis models based on LSTM neural networks, and performance growth analysis models based on linear regression algorithms. Supplier profiling analysis results include: qualification level assessment models output qualification scores (0-100 points) and levels (Excellent ≥90, Good 80-89, Satisfactory 60-79, Unsatisfactory <60); performance capability analysis models output performance capability scores (0-100 points) and performance risk levels (Low Risk ≥85, Medium Risk 70-84, High Risk <70); and performance growth analysis models output performance growth rate (%) and growth levels (High Growth ≥15%, Medium Growth 5%-14%, Low Growth <5%, Negative Growth).

[0047] The risk prediction model can be an XGBoost-based model, which can effectively handle multi-dimensional features (such as performance risk, price risk, qualification risk, and supply chain risk) and accurately identify core risks throughout the procurement process. The procurement risk situation analysis results are obtained by weighting the scores of the aforementioned multi-dimensional features to arrive at a comprehensive score. Based on a preset risk threshold, the comprehensive score determines the procurement risk situation analysis result (e.g., low risk, high risk).

[0048] After obtaining supplier risk warnings and purchase price warnings, two microservice interfaces should be designed following RESTful API design specifications. The interface URLs, request methods, request parameters, response parameters, and return formats should be clearly defined to ensure the interfaces are standardized and scalable. Then, the supplier risk warnings and purchase price warnings should be encapsulated into the business logic layer (Service layer) of the interfaces, as follows: (1) Data query call: Call the Cypher query method in the Service layer to extract the corresponding data from the knowledge graph according to the request parameters. If no supplier data is found, return a 404 status code.

[0049] (2) Model analysis call: The historical / current supply data retrieved will be used to call models with different analysis functions and risk prediction models respectively to obtain analysis results. If the model call fails, a 500 status code and error message will be returned.

[0050] (3) Result standardization: The analysis results returned by the model are converted into the response parameter format defined in the interface design, and the field names, data types and return formats are unified to ensure consistency under different calling scenarios.

[0051] As an example, the interface design details for supplier risk warning information are as follows: Interface name: supplierPortraitAnalysis API URL: / api / procurement / supplier / portrait / analysis Request method: GET (Query interface, retrieves profile analysis results for the specified supplier) Request parameters (path parameters + request parameters): (1) Path parameter: supplierId (supplier entity ID, required, string type, such as "supplier_001"); (2) Request parameters (Query parameters): timeRange (historical data time range, optional, default 36 months, integer type, unit: month), format (return format, optional, default JSON, supports JSON / XML).

[0052] Response parameters (JSON format): include five primary fields: basic information, multi-dimensional score summary, core analysis conclusions, risk warnings, and data source. Each primary field contains corresponding secondary fields. The fields are in JSON format to ensure data consistency.

[0053] Interface status codes: 200 (Success), 400 (Incorrect request parameters), 404 (Supplier does not exist), 500 (Internal server error).

[0054] Interface permissions: Only procurement administrators and supplier administrators can call it (implemented through the permission control module of "Service Mesh and API Governance Layer").

[0055] As an example, the interface details for procurement price warning information are as follows: Interface name: procurementRiskSituationAnalysis API URL: / api / procurement / risk / situation / analysis Request method: GET (query-type interface, retrieves risk profile analysis results for a specified supplier) Request parameters (path parameters + request parameters): (1) Path parameter: supplierId (supplier entity ID, required, string type, such as "supplier_001"); (2) Request parameters (Query parameters): timeRange (current data time range, optional, default 3 months, integer type, unit: month), format (return format, optional, default JSON, supports JSON / XML), riskLevel (risk level filter, optional, supports low risk / low-medium risk / medium-high risk / high risk).

[0056] Response parameters (JSON format): It includes five primary fields: analysis subject, risk quantification results, core risk description, risk cause analysis, and risk management recommendations. Each primary field contains corresponding secondary fields, and the field type is JSON format.

[0057] Interface status codes: consistent with the supplier profile interface (200 / 400 / 404 / 500).

[0058] Interface permissions: Only procurement administrators and risk control personnel can access it (implemented through the permission control module of "Service Mesh and API Governance Layer").

[0059] Supplier risk warning information and purchase price warning information can be displayed via API calls.

[0060] This embodiment realizes the process from knowledge graph data query and model analysis to interface encapsulation and invocation. It constructs the supplier profile analysis results and the procurement risk situation analysis results into microservice interfaces and exposes them to the outside world through an API gateway, thereby realizing efficient invocation of analysis results.

[0061] Furthermore, front-end applications can be allowed to access supplier profiling and procurement risk analysis results via standard SQL or GraphQL query languages, without being aware of the complexity of their physical storage. To ensure data security and compliance during integration and service provisioning, a privacy-enhancing computing module can be implemented. Homomorphic encryption or differential privacy technologies are implemented along critical data flow paths. Access control employs attribute-based access control policies to achieve fine-grained, dynamic authorization based on data sensitivity levels, user roles, and access contexts. All data operations are immutably recorded in audit logs, meeting traceability requirements.

[0062] Figure 2 This is a schematic diagram of an embodiment of a multi-source heterogeneous procurement data processing device according to the present invention. The device 200 includes: The acquisition module 201 is used to acquire raw procurement data from multiple heterogeneous data sources; The processing module 202 is used to perform dynamic semantic mapping processing on the original procurement data according to the preset concept relationship mapping rules to obtain a first knowledge ternary combination set; perform data filtering processing on the first knowledge ternary combination set to obtain a second knowledge ternary combination set; determine the procurement knowledge graph according to the second knowledge ternary combination set; and obtain the procurement intelligent analysis result according to the procurement knowledge graph.

[0063] Optionally, raw procurement data from multiple heterogeneous data sources can be obtained, including: The original procurement data is obtained from multiple preset data sources through protocol plugins and connector driver architecture, including relational databases, enterprise resource planning systems, office automation systems, and file systems.

[0064] Optionally, based on a pre-defined concept relation mapping model, the original procurement data undergoes dynamic semantic mapping processing to obtain a first knowledge ternary combination set, including: The unstructured text in the original procurement data is converted to a new format to obtain the first structured data; The existing second structured data and the first structured data in the original procurement data are parsed to obtain source data pattern elements; Based on the source data pattern elements and the target elements in the concept relationship mapping model, determine the mapping similarity data; When the mapping similarity data is greater than or equal to a preset threshold, the first structured data and the second structured data are transformed to obtain a first knowledge triplet set. The first knowledge triplet set includes multiple first knowledge triplets, and each first knowledge triplet includes an entity, an attribute, and a relationship between entities.

[0065] Optionally, the first knowledge ternary combination set is subjected to data filtering processing to obtain a second knowledge ternary combination set, including: Perform quality rule verification on the first knowledge triplet set to determine the confidence level data of each first knowledge triplet in the first knowledge triplet set; Based on the confidence data, a second knowledge triplet set is determined, which includes a first knowledge triplet whose confidence data is greater than a preset threshold.

[0066] Optionally, based on the second knowledge ternary set, a procurement knowledge graph is determined, including: Based on the pre-defined concept relationship mapping rules, the entities in the second knowledge ternary combination set are used as nodes, and the relationships between the entities in the second knowledge ternary combination set are used as edges to determine the procurement knowledge graph.

[0067] Optionally, based on the procurement knowledge graph, the procurement intelligent analysis results are obtained, including: Based on the procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained; Based on the supplier risk warning information and the purchase price warning information, the intelligent procurement analysis results are obtained.

[0068] Optionally, based on the procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained, including: Based on the procurement knowledge graph, historical supply data and current supply data of suppliers can be obtained. By inputting the historical supply data of the suppliers into models with different analytical functions, supplier risk warning information can be obtained; By inputting the supplier's current supply data into the risk prediction model, a purchase price early warning information is obtained.

[0069] It should be noted that this device is the same as the method described above. All implementations in the above method embodiments are applicable to the embodiments of this device and can achieve the same technical effect.

[0070] The present invention also includes a computing device comprising: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the method as described above.

[0071] In another aspect, the present invention also provides a computer-readable storage medium storing a program that, when executed by a processor, implements the method described above.

[0072] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0074] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for processing multi-source heterogeneous procurement data, characterized in that, include: Obtain raw procurement data from multiple heterogeneous data sources; According to the preset concept relationship mapping rules, the original procurement data is subjected to dynamic semantic mapping processing to obtain the first knowledge ternary combination set; The first knowledge ternary combination set is subjected to data filtering processing to obtain the second knowledge ternary combination set; Based on the second knowledge ternary combination set, determine the procurement knowledge graph; Based on the procurement knowledge graph, the procurement intelligent analysis results are obtained.

2. The multi-source heterogeneous procurement data processing method according to claim 1, characterized in that, Obtain raw procurement data from multiple heterogeneous data sources, including: The original procurement data is obtained from multiple preset data sources through protocol plugins and connector driver architecture, including relational databases, enterprise resource planning systems, office automation systems, and file systems.

3. The multi-source heterogeneous procurement data processing method according to claim 1, characterized in that, Based on a pre-defined concept relation mapping model, the original procurement data undergoes dynamic semantic mapping processing to obtain a first knowledge ternary combination set, including: The unstructured text in the original procurement data is converted to a new format to obtain the first structured data; The existing second structured data and the first structured data in the original procurement data are parsed to obtain source data pattern elements; Based on the source data pattern elements and the target elements in the concept relationship mapping model, determine the mapping similarity data; When the mapping similarity data is greater than or equal to a preset threshold, the first structured data and the second structured data are transformed to obtain a first knowledge triplet set. The first knowledge triplet set includes multiple first knowledge triplets, and each first knowledge triplet includes an entity, an attribute, and a relationship between entities.

4. The multi-source heterogeneous procurement data processing method according to claim 1, characterized in that, Data filtering processing is performed on the first knowledge ternary combination set to obtain the second knowledge ternary combination set, which includes: Perform quality rule verification on the first knowledge triplet set to determine the confidence level data of each first knowledge triplet in the first knowledge triplet set; Based on the confidence data, a second knowledge triplet set is determined, which includes a first knowledge triplet whose confidence data is greater than a preset threshold.

5. The multi-source heterogeneous procurement data processing method according to claim 1, characterized in that, Based on the second knowledge ternary combination set, a procurement knowledge graph is determined, including: Based on the pre-defined concept relationship mapping rules, the entities in the second knowledge ternary combination set are used as nodes, and the relationships between the entities in the second knowledge ternary combination set are used as edges to determine the procurement knowledge graph.

6. The multi-source heterogeneous procurement data processing method according to claim 1, characterized in that, Based on the aforementioned procurement knowledge graph, the procurement intelligent analysis results are obtained, including: Based on the procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained; Based on the supplier risk warning information and the purchase price warning information, the intelligent procurement analysis results are obtained.

7. The multi-source heterogeneous procurement data processing method according to claim 6, characterized in that, Based on the aforementioned procurement knowledge graph, supplier risk warning information and procurement price warning information are obtained, including: Based on the procurement knowledge graph, historical supply data and current supply data of suppliers can be obtained. By inputting the historical supply data of the suppliers into models with different analytical functions, supplier risk warning information can be obtained; By inputting the supplier's current supply data into the risk prediction model, a purchase price early warning information is obtained.

8. A multi-source heterogeneous procurement data processing device, characterized in that, include: The acquisition module is used to acquire raw procurement data from multiple heterogeneous data sources; The processing module is used to perform dynamic semantic mapping processing on the original procurement data according to preset concept relationship mapping rules to obtain a first knowledge ternary combination set; perform data filtering processing on the first knowledge ternary combination set to obtain a second knowledge ternary combination set; and determine the procurement knowledge graph based on the second knowledge ternary combination set. Based on the procurement knowledge graph, the procurement intelligent analysis results are obtained.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images