A knowledge graph construction system for commercial banking business association and matching

Abstract

The present invention discloses a knowledge graph construction system for commercial banking business association and matching, including: a plurality of data source interface modules, a construction recognition module, a knowledge graph construction module, a data analysis module, a recommendation module, a resource optimization module, a visualization interface module, a feedback mechanism module, and a security control module. It summarizes the financial reports, patents, and news of global commercial banks in real time from multiple information sources, extracts the association relationships between commercial banking business scenarios and specific technologies, and uses association mining methods to infer and complement the relationships between nodes, which can fully ensure the real-time nature of data, the accuracy of cleaning, and the recognition accuracy of triples, and maintain the up-to-date status of the knowledge graph. On the one hand, starting from the idea of business and technology matching, this system explores the data value of public information to support the configuration decision-making of commercial banking business and technology resources. On the other hand, by introducing association inference methods and real-time multi-source information, the accuracy and practicality of the knowledge graph are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge graph construction, and particularly to a knowledge graph construction system for business association and matching in commercial banking. Background Art

[0002] In recent years, with the continuous improvement of the complexity and refinement of commercial bank services, emerging technologies have been continuously integrated into the industry development to help financial institutions provide high-quality services to customers. However, with the rapid iteration of technologies, when financial institutions judge how to select appropriate supporting technologies in combination with business characteristics, the difficulty and cost of decision-making are also rapidly increasing.

[0003] After retrieval, a Chinese patent with the application number CN109977419A discloses a knowledge graph construction system, which proposes to display the updated knowledge in various fields in the form of a knowledge graph to facilitate people's browsing and communication of relevant information;

[0004] A Chinese patent with the application number CN116401379A discloses a financial product data pushing method, device, equipment and storage medium, which points out the problem that the existing financial product data pushing method has a low matching degree with the actual needs of users when pushing financial product data, and discloses using a graph modeling method to increase data matching.

[0005] In the existing methods, it is a great challenge to summarize the experience of business and technology integration in the industry from public information. The importance of the association between business and technology is not well considered in the existing technologies, the practical value of the associated data cannot be explored, and it is often impossible to integrate news, patents and financial information globally in real time and accurately.

[0006] In summary, how to update the association and matching between business and technology resources in real time from news, patents and financial information globally to support decision-making in the directions of technology and business development is not only an important problem urgently to be solved by commercial banks, but also by various industries. Therefore, it is necessary to study a knowledge graph construction system for providing the association and matching relationship between commercial banking business and technology, so as to summarize the development experience for the industry and provide good support for corporate decision-making. Summary of the Invention

[0007] The purpose of the present invention is to solve the defects existing in the prior art, and to propose a knowledge graph construction system for business association and matching in commercial banking.

[0008] To achieve the above purpose, the present invention adopts the following technical solutions:

[0009] A knowledge graph construction system for commercial banking business association and matching, comprising: a number of data source interface modules for collecting text data, automatically extracting, cleaning, and transforming data from various data sources to ensure data consistency and availability;

[0010] A construction recognition module for performing named entity recognition on the preprocessed text data and constructing triples to generate a structured knowledge representation;

[0011] A knowledge graph construction module for integrating the financial data and technical resource data into a hierarchical knowledge graph, which includes the key business areas, technical resources of commercial banks, and the relationships between the areas and resources. The knowledge graph is implemented in the form of a graph database, allowing efficient data retrieval and query.

[0012] A data analysis module for automatically analyzing the business requirements, technical resource characteristics of commercial banks, and their matching degree using machine learning algorithms, providing predictive analysis, and generating corresponding evaluation results. The evaluation results include multiple dimensions such as resource utilization rate, risk analysis, potential growth opportunities, and cost-benefit analysis;

[0013] A recommendation module for providing intelligent suggestions for business and technical resource matching by analyzing the data in the knowledge graph, including resource priorities, potential partners, and strategic directions;

[0014] A resource optimization module for further automatically or semi-automatically optimizing the suggestions of the recommendation module according to the business objectives and resource status data of commercial banks to achieve the best resource allocation;

[0015] A visualization interface module for graphically presenting the knowledge graph, matching suggestions, and optimization results to help commercial bank managers better understand and make decisions on resource allocation;

[0016] A feedback mechanism module that allows commercial bank users to provide judgments and feedback on the situation after the knowledge graph data is cleaned and participate in the process of matching suggestions and optimization decisions;

[0017] A security control module to ensure the secure storage and access control of the data and resource matching information in the knowledge graph.

[0018] Among them, the construction recognition module and the data analysis module are respectively connected to the data source interface module. The data source interface module is connected to the knowledge graph construction module through the construction recognition module. The knowledge graph construction module is respectively connected to the data analysis module and the recommendation module. Both the data analysis module and the recommendation module are connected to the visualization interface module through the resource optimization module. The knowledge graph construction module, the construction recognition module, the data analysis module, the resource optimization module, and the recommendation module are all connected to the feedback mechanism module and the security control module.

[0019] Further, the data source interface module is used to collect multi-source financial data including transaction data, customer data, market data, competitor data, and technical resource data. Among them, different data interface modules are assigned different data types and collection methods for the data to be collected.

[0020] Further, the specific methods for the data source interface module to obtain and preprocess data are as follows:

[0021] The data source interface module is assigned a data collection method and a collection address, and recognition data is obtained from corresponding websites, papers, reports, and forums. Among them, each data source interface module collects data from at least one of the following addresses, specifically including any relevant industry news website, any relevant field patent website module, any relevant field social forum module, and the business description of the financial reports of any relevant industry listed companies.

[0022] Preprocess the obtained text data, including removing HTML tags, splitting sentences, word segmentation, etc., to obtain clear text content.

[0023] Further, the specific methods for the construction recognition module to perform named entity recognition and construct triples are as follows:

[0024] Perform named entity recognition on the preprocessed text data to identify entity descriptions in the text, including company names, technology types, business types, and industry terms.

[0025] Assign appropriate tags to each recognized entity and incorporate it into the corresponding category in the knowledge graph.

[0026] Based on the recognized entity descriptions, perform relationship extraction to identify the association relationships between entities.

[0027] Construct triples, taking entities and relationships as the subject, verb, and object to generate a structured knowledge representation.

[0028] Further, the specific step process of the knowledge graph construction module is as follows:

[0029] Import the constructed triples into the graph database to construct the nodes and edges of the knowledge graph;

[0030] Utilize the query and analysis functions of the graph database for association mining;

[0031] Compare the results through the link prediction algorithm to obtain the potential associations between entities;

[0032] Such as the co-occurrence relationship of technologies among different companies, the application frequency and situation of technologies in business domains, etc.

[0033] Based on the association inference, form a dotted line with association probability parameters in the knowledge graph to prompt the user that this association is generated by the inference algorithm.

[0034] Furthermore, the feedback mechanism module, data source interface module, and knowledge graph construction module are used to establish multiple question tools that can perform training and automatically collect data. The specific step process is as follows: during the feedback and adjustment process, establish multiple learning models based on different data collection sources;

[0035] Classify the learning models. The models search for relevant data through the feedback questions of each user, record and count this Q&A, and the models ask questions to train each other and provide data support;

[0036] The model provides keywords for the data to be obtained, and another matching model conducts a search. After obtaining the results, it feeds back to the system.

[0037] Furthermore, the specific step process of the visualization interface module is as follows:

[0038] Construct a user interface or API to enable users to access and query the information in the knowledge graph through the query interface;

[0039] According to the user's query, display relevant triple information, show the relationships between entities, and the application fields of technologies;

[0040] Provide a visual display of the graph to intuitively show the association relationship between commercial banking business and technology resources.

[0041] Furthermore, the intelligent suggestions of the recommendation module specifically adopt a deep learning model, optimize the suggestions according to real-time data and user feedback, and are divided into long-term strategic suggestions and short-term emergency suggestions according to the changing needs of commercial banks to meet the needs within different time ranges.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] It can store, clean, identify, and associate this information in real time, thereby constructing a comprehensive and real-time updated knowledge graph of the association between commercial banking business and technology resources;

[0044] It can fully guarantee the real-time nature of data, the accuracy of cleaning, and the precision of triple recognition, update information globally in real time, keep the knowledge graph in the latest state, extract associated information from multiple information sources to form a comprehensive knowledge graph, use association mining methods to infer and complete the relationships between nodes, enhance the integrity and accuracy of the knowledge graph, and through the construction of triple relationships, it can infer and complete the relationships between nodes in the knowledge graph, improving the accuracy and practicality of the knowledge graph. Brief Description of the Drawings

[0045] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention.

[0046] Figure 1 It is a schematic flowchart of a knowledge graph construction system for commercial banking business association and matching proposed by the present invention. Detailed Embodiments

[0047] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0048] Referring to Figure 1 , a knowledge graph construction system for commercial banking business association and matching includes: a number of data source interface modules for collecting text data, automatically extracting, cleaning, and transforming data from various data sources to ensure data consistency and availability;

[0049] A construction recognition module for performing named entity recognition on the preprocessed text data and constructing triples to generate a structured knowledge representation;

[0050] A knowledge graph construction module for integrating the financial data and technical resource data into a hierarchical knowledge graph, which includes the key business areas of commercial banks, technical resources, and the relationships between the areas and resources. The knowledge graph is implemented in the form of a graph database, allowing efficient data retrieval and query.

[0051] A data analysis module for automatically analyzing the business requirements of commercial banks, the characteristics of technical resources, and their matching degree using machine learning algorithms, providing predictive analysis, and generating corresponding evaluation results. The evaluation results include multiple dimensions such as resource utilization rate, risk analysis, potential growth opportunities, and cost-benefit analysis;

[0052] Recommendation module, which provides intelligent recommendations for business and technical resource matching by analyzing data in the knowledge graph, including resource priorities, potential partners, and strategic directions;

[0053] Resource optimization module, which is used to further automatically or semi-automatically optimize the recommendations of the recommendation module according to the business objectives and resource status data of commercial banks to achieve the best resource allocation;

[0054] Visualization interface module, which is used to graphically present the knowledge graph, matching recommendations, and optimization results to help commercial bank managers better understand and make decisions on resource allocation;

[0055] Feedback mechanism module, which allows commercial bank users to provide judgments and feedback on the situation after the knowledge graph data is cleaned and participate in the process of matching recommendations and optimization decisions;

[0056] Security control module, which ensures the secure storage and access control of the data and resource matching information in the knowledge graph.

[0057] Among them, the construction recognition module and the data analysis module are respectively connected to the data source interface module. The data source interface module is connected to the knowledge graph construction module through the construction recognition module. The knowledge graph construction module is respectively connected to the data analysis module and the recommendation module. Both the data analysis module and the recommendation module are connected to the visualization interface module through the resource optimization module. The knowledge graph construction module, the construction recognition module, the data analysis module, the resource optimization module, and the recommendation module are all connected to the feedback mechanism module and the security control module.

[0058] In a specific embodiment of the present application, the data source interface module is used to collect multi-source financial data including transaction data, customer data, market data, competitor data, and technical resource data. Among them, different data interface modules are assigned different data types and collection methods for collection.

[0059] In a specific embodiment of the present application, the specific method for the data source interface module to obtain and preprocess data is as follows:

[0060] The data source interface module is assigned a data collection method and a collection address to obtain identification data from corresponding websites, papers, reports, and forums. Each data source interface module collects data from at least one of the following addresses, specifically including any relevant industry news website, any relevant field patent website module, any relevant field social forum module, and any relevant industry listed company financial report business description;

[0061] Preprocess the obtained text data, including removing HTML tags, sentence segmentation, word segmentation, etc. to obtain clear text content.

[0062] In a specific embodiment of the present application, the specific method for constructing the recognition module for named entity recognition and constructing triples is as follows:

[0063] Perform named entity recognition on the preprocessed text data to identify entity descriptions in the text, including company names, technology categories, business types, and industry terms;

[0064] Assign appropriate labels to each recognized entity and incorporate it into the corresponding category in the knowledge graph;

[0065] Based on the recognized entity descriptions, perform relation extraction to identify the association relationships between entities; for example, the relationship between specific technologies used by a certain commercial bank, the association of specific technologies applied in certain businesses, etc.

[0066] Construct triples, taking entities and relationships as the subject, verb, and object to generate a structured knowledge representation.

[0067] In a specific embodiment of the present application, the specific step process of the knowledge graph construction module is as follows:

[0068] Import the constructed triples into the graph database to construct the nodes and edges of the knowledge graph;

[0069] Utilize the query and analysis functions of the graph database to perform association mining;

[0070] Perform result comparison through the link prediction algorithm to obtain potential associations between entities;

[0071] Such as the co-occurrence relationship of technologies between different companies, the application frequency and situation of technologies in the business field, etc. It should be further noted that the link prediction algorithms include:

[0072] ① Co-occurrence matrix method: This method calculates the similarity between nodes based on the co-occurrence situation between nodes. Common methods include cosine similarity, Jaccard similarity, etc.

[0073] ② Random walk algorithm: Includes Random Walk and Personalized PageRank, etc. These methods simulate the random walk process between nodes and calculate the similarity between nodes based on the access frequency of nodes.

[0074] ③ Graph structure-based method: This type of method focuses on the neighbor structure of nodes and predicts the connection relationship between nodes by calculating the similarity between the neighbors of nodes. For example, the Common Neighbors method calculates the number of common neighbors of two nodes.

[0075] ④Matrix factorization method: Using matrix factorization technology, nodes and edges are represented as matrices, and the association relationships between nodes are obtained by decomposing the matrices.

[0076] ⑤Graph Neural Networks (GNN): GNN is a method that has emerged in recent years and can learn complex relationships between nodes for link prediction through the embedded representations of nodes.

[0077] ⑥Negative sampling method: This method conducts training by randomly sampling some negative examples to learn the similarity model between nodes, thereby predicting the connection relationships between nodes.

[0078] Classic link prediction algorithms can be selected and adjusted according to the application scenario and data characteristics. Among them, the random walk algorithm, graph structure-based methods, and graph neural networks perform relatively well in processing complex network data and large-scale knowledge graphs.

[0079] Based on association inference, dotted lines with association probability parameters are formed in the knowledge graph to prompt users that the association is generated by the inference algorithm.

[0080] In the specific embodiments of this application, the feedback mechanism module, data source interface module, and knowledge graph construction module are used to establish multiple question tools capable of training and automatically collecting data. The specific step process is as follows: In the process of feedback and adjustment, multiple learning models based on different data collection sources are established;

[0081] The learning models are classified. The models search for relevant data through the feedback questions of each user, record and count this Q&A, and the models ask questions to each other for training and provide data support;

[0082] The models provide keywords for the data to be obtained, and another matching model conducts a search. After obtaining the results, they are fed back into the system.

[0083] In the specific embodiments of this application, the specific step process of the visualization interface module is as follows:

[0084] Construct a user interface or API to enable users to access and query the information in the knowledge graph through the query interface;

[0085] According to the user's query, display relevant triple information, showing the relationships between entities and the application fields of technologies;

[0086] Provide a visual display of the graph to intuitively show the association relationships between commercial banking operations and technical resources.

[0087] In a specific embodiment of the present application, the intelligent suggestions of the recommendation module specifically adopt a deep learning model, optimize the suggestions according to real-time data and user feedback, and are divided into long-term strategic suggestions and short-term emergency suggestions according to the changing needs of commercial banks to meet the needs within different time ranges.

[0088] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention should cover equivalent replacements or changes made according to the technical solution and inventive concept of the present invention within the protection scope of the present invention.

Claims

1. A knowledge graph construction system for commercial banking business association and matching, characterized in that, Including: A number of data source interface modules, which are used to collect text data, automatically extract, clean, and transform data from various data sources to ensure data consistency and availability; A construction recognition module, which is used to perform named entity recognition on the preprocessed text data and construct triples to generate a structured knowledge representation; A knowledge graph construction module, which is used to integrate financial data and technical resource data into a hierarchical knowledge graph. The knowledge graph includes the key business areas, technical resources of commercial banks, and the relationships between business and technology. The knowledge graph is implemented in the form of a graph database, allowing efficient data retrieval and query; A data analysis module, which is used to automatically analyze the business requirements, technical resource characteristics of commercial banks, and their matching degree using machine learning algorithms, provide predictive analysis, and generate corresponding evaluation results. The evaluation results include resource utilization rate, risk analysis, potential growth opportunities, and cost-benefit analysis; A recommendation module, which provides intelligent suggestions for business and technical resource matching by analyzing the data in the knowledge graph, including resource priorities, potential partners, and strategic directions; A resource optimization module, which is used to further automatically or semi-automatically optimize the suggestions of the recommendation module according to the business objectives and resource status data of commercial banks to achieve the best resource allocation; A visualization interface module, which is used to graphically present the knowledge graph, matching suggestions, and optimization results to help commercial bank managers better understand and make decisions on resource allocation; A feedback mechanism module, which allows users to provide manual judgment and feedback on the situation after data cleaning of the knowledge graph, and allows users to participate in the process of matching suggestions and optimization decisions; A security control module, which ensures the secure storage and access control of the data and resource matching information in the knowledge graph; Among them, the construction recognition module and the data analysis module are respectively connected to the data source interface module. The data source interface module is connected to the knowledge graph construction module through the construction recognition module. The knowledge graph construction module is respectively connected to the data analysis module and the recommendation module. The data analysis module and the recommendation module are both connected to the visualization interface module through the resource optimization module. The knowledge graph construction module, the construction recognition module, the data analysis module, the resource optimization module, and the recommendation module are all connected to the feedback mechanism module and the security control module.

2. The knowledge graph construction system for commercial banking business association and matching according to claim 1, wherein The data source interface module is used to collect multi-source financial data including transaction data, customer data, market data, competitor data, and technical resource data. Among them, different data interface modules are assigned different data types and collection methods for collection.

3. The knowledge graph construction system for commercial banking business association and matching according to claim 2, wherein The specific method for the data source interface module to obtain and preprocess data is: The data source interface module assigns data collection methods and collection addresses, and obtains identification data from corresponding websites, papers, reports, and forums. Among them, each data source interface module collects data from at least one of the following addresses, specifically including any relevant industry news website, any patent website module in the relevant field, any social forum module in the relevant field, and the business descriptions of financial reports of listed companies in any relevant industry; Preprocess the obtained text data, including removing HTML tags, clause segmentation, and word segmentation, to obtain clear text content.

4. The knowledge graph construction system for commercial banking business association and matching according to claim 1, characterized in that The specific methods for constructing the entity recognition and triple construction in the recognition module are as follows: Perform named entity recognition on the preprocessed text data to identify entity descriptions in the text, including company names, technology types, business types, and industry terms; Assign appropriate tags to each recognized entity and incorporate it into the corresponding category in the knowledge graph; Based on the recognized entity descriptions, perform relationship extraction to identify the association relationships between entities; Construct triples, using entities and relationships as the subject, verb, and object to generate a structured knowledge representation.

5. The knowledge graph construction system for commercial banking business association and matching according to claim 4, characterized in that, The specific step process of the knowledge graph construction module is as follows: Import the constructed triples into the graph database to construct the nodes and edges of the knowledge graph; Utilize the query and analysis functions of the graph database to perform association mining; Perform result comparison through the link prediction algorithm to obtain the potential associations between entities; Based on association inference, form a dotted line with association probability parameters in the knowledge graph to prompt the user that this association is generated through the inference algorithm.

6. The knowledge graph construction system for commercial banking business association and matching according to claim 5, wherein The feedback mechanism module, data source interface module, and knowledge graph construction module are used to establish multiple question tools that can perform training and automatically collect data. The specific step process is as follows: During the feedback and adjustment process, establish multiple learning models based on different data collection sources; Classify the learning models. The models search for relevant data through the feedback questions of each user, record and count this Q&A, and the models ask questions and provide data support for each other; The model provides keywords for the data to be obtained, and another matching model conducts a search. After obtaining the results, it is fed back into the system.

7. The knowledge graph construction system for commercial banking business association and matching according to claim 6, characterized in that The specific step process of the visualization interface module is as follows: Construct a user interface or API to enable users to access and query the information in the knowledge graph through the query interface; According to the user's query, display relevant triple information, showing the relationships between entities and the application fields of technologies; Provide a visual display of the graph to intuitively show the association relationship between commercial banking business and technical resources.

8. The knowledge graph construction system for commercial banking business association and matching according to claim 7, wherein The intelligent suggestions of the recommendation module specifically adopt a deep learning model, optimize the suggestions according to real-time data and user feedback, and are divided into long-term strategic suggestions and short-term emergency suggestions according to the changing needs of commercial banks to meet the needs within different time ranges.

Images