Enterprise credit dynamic evaluation and visualization system based on multi-source data fusion

Abstract

The application discloses an enterprise credit dynamic evaluation and visualization system based on multi-source data fusion, and relates to the technical field of enterprise credit evaluation.The system comprises a multi-source data acquisition module, a unified semantic analysis module, a feature dynamic fusion module, a credit dynamic evaluation module and a holographic visualization module.The application eliminates data modal differences through multi-source data acquisition and unified semantic analysis, generates holographic credit features, breaks through traditional evaluation barriers and builds a precise evaluation foundation.Relying on a lightweight computing model and a risk event monitoring mechanism, the application realizes multi-dimensional credit real-time calculation and emergency update, improves risk response efficiency, and, in combination with the holographic visualization module, intuitively displays credit information, simplifies interpretation processes, reduces decision-making thresholds and comprehensively improves the efficiency and effectiveness of enterprise credit management and risk control.

Description

Technical Field

[0001] This invention relates to the field of enterprise credit assessment technology, specifically to an enterprise credit dynamic assessment and visualization system based on multi-source data fusion. Background Technology

[0002] With the deep integration of the digital economy and industrial finance, corporate credit assessment has become a core support for credit risk control, supply chain cooperation, and market supervision. Currently, corporate operational data is characterized by multiple sources, heterogeneity, and real-time processing, encompassing structured data such as financial and business registration information, as well as multimodal unstructured data such as news, public opinion, audio and video, and court announcements. Traditional, single-source, static credit assessment models are no longer suitable for the complex business environment. Corporate credit risk has dynamic and sudden characteristics, urgently requiring a credit assessment system that can integrate multi-source data, perform real-time dynamic calculations, and provide holographic visualization to improve the timeliness, comprehensiveness, and decision-making reference value of credit assessments.

[0003] Existing corporate credit assessment technologies have many shortcomings: data collection dimensions are limited, mostly confined to structured data such as financial and business registration information, failing to effectively integrate multimodal unstructured data such as text, images, and audio, resulting in insufficient data value mining; feature fusion uses fixed weight ratios, unable to dynamically adjust according to data timeliness and importance, making it difficult to highlight the assessment weights of the latest data and key features, and the depth of feature fusion is insufficient; assessment models are mostly static offline calculations, lacking an emergency response mechanism for major risk events, resulting in serious lag in risk warnings and score updates; the visualization interface is monotonous, lacking interactive query, drill-down, and information traceability functions, and the assessment results are not intuitive or comprehensive, making it difficult to support refined decision analysis.

[0004] In summary, existing corporate credit assessment technologies have significant shortcomings in data fusion, dynamic calculation, risk response, and visualization, failing to meet the current practical needs of dynamic corporate credit assessment. Therefore, this invention proposes a dynamic corporate credit assessment and visualization system based on multi-source data fusion. Through multi-source data collection, multimodal semantic parsing, dynamic feature fusion, real-time credit assessment, and holographic visualization interaction, it overcomes the problems of data barriers, rigid weighting, assessment lag, and limited display in existing technologies. This significantly improves the accuracy, real-time performance, and ease of use of corporate credit assessment, providing reliable support for financial risk control and business decision-making. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a dynamic assessment and visualization system for enterprise credit based on multi-source data fusion. It can acquire comprehensive data through a multi-source data acquisition module, eliminate modal differences through unified semantic parsing, dynamically fuse features using an attention mechanism to generate holographic credit features, calculate multi-dimensional credit scores in real time based on a lightweight model, and achieve emergency updates through a built-in risk monitoring mechanism. The holographic visualization module supports interactive queries and information traceability, intuitively displaying credit status and improving the efficiency and effectiveness of enterprise credit management and risk control.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a dynamic enterprise credit assessment and visualization system based on multi-source data fusion, the system comprising: a multi-source data acquisition module, a unified semantic parsing module, a feature dynamic fusion module, a credit dynamic assessment module, and a holographic visualization module;

[0007] The multi-source data acquisition module collects structured and unstructured data from the Internet, government open platforms, corporate websites, news media, and social media channels in real time. After data cleaning and format conversion, it directly extracts structured feature vectors from the structured data and inputs the unstructured data into the unified semantic parsing module.

[0008] The unified semantic parsing module uses a multimodal large model to perform end-to-end semantic parsing on the input data of different modalities, generates unstructured credit feature vectors in a unified format, updates the parsing results in real time, and inputs them into the feature dynamic fusion module.

[0009] The feature dynamic fusion module employs an attention mechanism to dynamically adjust the weights of structured and unstructured features based on the timeliness and importance of different data types. This deeply integrates traditional financial structured indicators with unstructured credit features to generate a holographic credit feature representation of the enterprise. The fusion process is performed in real time, and the feature vector is updated immediately when new data is input and then input into the credit dynamic assessment module.

[0010] The credit dynamic assessment module: based on holographic credit feature representation, it calculates the enterprise's comprehensive credit score and specific credit scores for each dimension in real time through a lightweight gradient boosting tree model; it has a built-in rule base for major risk events, and when a major risk event is detected, it immediately triggers an emergency update of the credit score and inputs the assessment results into the holographic visualization module;

[0011] The holographic visualization module receives holographic credit feature representations, comprehensive credit scores, and specific credit score data, and generates a multi-dimensional visualization interface that supports interactive user queries and drill-downs.

[0012] Furthermore, the multi-source data acquisition module deploys multiple nodes to crawl public data in parallel using distributed crawling technology, obtains authoritative data by connecting to government official platforms through API interfaces, and obtains third-party professional data through data subscription technology. The collected structured data includes financial statements, business registration information, tax data, and annual report operating data, while the unstructured data includes annual report texts, news reports, court announcements, product images, and audio recordings of executive interviews.

[0013] Furthermore, the unified semantic parsing module uses a pre-trained and finely tuned multimodal large model to perform end-to-end semantic parsing on data from different modalities. For text data, it extracts entities, events, sentiment tendencies, and risk signals; for image data, it identifies product defects, production safety hazards, and trademark infringements; for audio data, it converts it into text and analyzes tone, emotion, and credibility of promises; all parsing results are converted into unstructured credit feature vectors of a unified dimension, and are immediately updated and transmitted to the feature dynamic fusion module when new data arrives.

[0014] Furthermore, the feature dynamic fusion module employs an attention mechanism to calculate the attention weight of each feature, specifically:

[0015] The initial feature vector is obtained by concatenating the structured feature vector output by the multi-source data acquisition module and the unstructured credit feature vector output by the unified semantic parsing module.

[0016] The attention weight of the i-th feature is calculated according to the formula. :

[0017]

[0018] in, Let be the attention weight for the i-th feature; This is a timeliness weighting coefficient; The timeliness factor is the data corresponding to the i-th feature; This refers to the importance weighting coefficient; Let i be the importance factor of the i-th feature;

[0019] The initial feature vector is weighted feature by feature based on attention weights to generate a weighted holographic credit feature representation.

[0020] Furthermore, the timeliness factor in the feature dynamic fusion module It is negatively correlated with the time difference between data release time and system time; the smaller the time difference, the better. The larger the value, the higher the value, ranging from 0 to 1; the importance factor The degree of influence of characteristics based on historical data statistics on credit scores is determined by the magnitude of the influence. The larger the value, the higher the value, ranging from 0 to 1; the timeliness weight coefficient And importance weight coefficient satisfy It can be adjusted according to the characteristics of the industry.

[0021] Furthermore, the credit dynamic assessment module, based on holographic credit feature representation, calculates specialized credit scores for five dimensions using a lightweight gradient boosting tree model, with the following formula:

[0022]

[0023] in, For the j-th dimension, a specific credit score is given. This is a lightweight gradient boosting tree model for the corresponding dimension; This represents the holographic credit feature.

[0024] The comprehensive credit score is calculated based on the specific credit scores of each dimension, using the following formula:

[0025]

[0026] in, For comprehensive credit scoring; is the weight coefficient for the j-th dimension, where j ranges from 1 to 5, corresponding to the five dimensions of finance, operations, compliance, reputation, and performance, respectively. For the j-th dimension, a specific credit score is given.

[0027] The specific credit scores and comprehensive credit scores for the five dimensions all range from 0 to 100 points, with higher scores indicating better creditworthiness in the corresponding dimension or overall. The weighting coefficients for the five dimensions satisfy the following... The system has built-in default weight configurations for different industries, and users can customize and adjust them according to their actual needs.

[0028] Furthermore, the credit dynamic assessment module has a built-in rule base for major risk events, which continuously monitors risk signals in the holographic credit feature representation. When a major risk event is detected, such as an enterprise being listed as a dishonest executor, a major safety accident, a large debt default, the revocation of its business license, or a major administrative penalty, all ordinary update tasks are immediately suspended, computing resources are prioritized to trigger an emergency credit score update process, and all score calculations are completed and pushed to the holographic visualization module within 10 seconds.

[0029] Furthermore, the visualization interface generated by the holographic visualization module includes a credit scoring dashboard, a feature contribution heatmap, a risk event timeline, a multi-dimensional credit radar chart, and an information source traceability chart. The credit scoring dashboard displays the company's comprehensive credit score and specific credit scores for each dimension in real time. The feature contribution heatmap shows the degree of contribution of different features to the comprehensive credit score. The risk event timeline displays all risk events that have occurred to the company in chronological order. The multi-dimensional credit radar chart provides an intuitive comparison of the company's credit performance across various dimensions. The information source traceability chart displays the original data source for each credit feature.

[0030] Furthermore, the holographic visualization module supports multi-dimensional interactive queries and drill-down operations; users can select different target companies through the search box, view historical credit data and score change trends for any time period through the time selector; click on any feature in the feature contribution heatmap to view detailed information and data sources of the feature; and click on any event in the risk event timeline to view the detailed content and impact of the event.

[0031] Compared with existing technologies, this enterprise credit dynamic assessment and visualization system based on multi-source data fusion has the following beneficial effects:

[0032] I. This invention covers various structured and multimodal unstructured information of enterprises through multi-source data collection. It obtains comprehensive and authoritative raw data through distributed collection, official interface docking, and professional data subscription. Then, it completes end-to-end processing of different modal data through unified semantic parsing, accurately extracts credit-related features and transforms them into unified format vectors, effectively eliminating the parsing difficulties caused by data modality differences. It realizes dynamic allocation of feature weights by relying on attention mechanism, and achieves deep integration of two types of features by combining data timeliness and importance, generating holographic credit features that fit the actual situation of enterprises. It breaks through the data barriers and rigid weight problems of traditional assessment, making feature mining more thorough and representation more complete. It lays a solid foundation for accurate assessment from the data level and greatly improves the comprehensiveness and reliability of assessment basis.

[0033] Second, this invention utilizes a lightweight computational model based on holographic credit features to perform real-time calculations of multi-dimensional specific and comprehensive credit information. It supports flexible adaptation of assessment weights according to industry and scenario, overcoming the lag of traditional static offline assessments. The built-in risk event monitoring mechanism can quickly identify key risk signals and prioritize resource allocation to complete emergency updates of credit scores, significantly improving risk response and early warning efficiency. Combined with a holographic visualization module, it presents multi-dimensional credit information, supports interactive queries, data drilling, and information tracing, and intuitively displays credit scores, feature contributions, risk context, and data sources. This simplifies the credit information interpretation process, lowers the threshold for decision-making, and makes assessment results easier to implement, comprehensively improving the efficiency and effectiveness of enterprise credit management and risk control.

[0034] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0036] Figure 1 This is a block diagram of the modules of a dynamic enterprise credit assessment and visualization system based on multi-source data fusion.

[0037] Figure 2 The flowchart shows the execution process of the multi-source data acquisition module in a dynamic enterprise credit assessment and visualization system based on multi-source data fusion.

[0038] Figure 3 This is a flowchart of the dynamic credit assessment process for a corporate credit assessment and visualization system based on multi-source data fusion. Detailed Implementation

[0039] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0040] Example 1

[0041] Multi-source data acquisition module: Upon entering operational mode, this module utilizes distributed crawler technology to achieve parallel deployment across multiple nodes, enabling efficient crawling of enterprise credit-related data from publicly available internet channels. It seamlessly connects with official government data platforms via standardized API interfaces, stably acquiring authoritative government data from sectors such as industry and commerce, taxation, and judiciary. Simultaneously, it leverages data subscription technology to connect with third-party professional data institutions, obtaining enterprise operation and credit data with industry reference value. The module comprehensively acquires various credit-related data of target enterprises in real-time. Structured data includes enterprise financial statements, business registration information, tax payment data, and annual report operating data; unstructured data includes enterprise annual report texts, industry news reports, court announcements, and product photos. The system includes audio interviews with senior executives. After data collection, the module automatically performs data cleaning and format conversion, removing duplicate, erroneous, missing, and redundant information. It also standardizes data storage formats and encoding standards. For structured data, it directly extracts features and generates standardized structured feature vectors. For unstructured data, it is stably transmitted to a unified semantic parsing module after preprocessing. Through multi-dimensional and multi-channel data collection and standardized processing, a high-quality credit data source covering all dimensions of enterprise operations is built, ensuring the authenticity, timeliness, completeness, and authority of the data. This lays a solid data foundation for subsequent data analysis, feature fusion, and credit assessment, allowing credit assessment to break free from the limitations of a single data source and fully reflect the actual business and credit status of the enterprise.

[0042] The unified semantic parsing module, upon receiving unstructured data from the multi-source data acquisition module, immediately activates a pre-trained, fine-tuned multimodal large model to perform end-to-end semantic parsing processing on text, image, and audio data. For text data, it accurately extracts enterprise entity information, key business events, market sentiment, and potential credit risk signals. For image data, it efficiently identifies product quality defects, production safety hazards, and trademark infringement violations. For audio data, it first performs accurate speech-to-text conversion and then deeply analyzes the tone, emotional characteristics, and credibility levels of the promises made. All parsing results are converted into unstructured information according to a unified dimensional standard. Using feature vectors, when new unstructured data is transmitted to the module, the parsing process immediately starts and completes the real-time update of the feature vectors. The updated unstructured credit feature vectors are then synchronously transmitted to the feature dynamic fusion module. Through the end-to-end parsing capability of the multimodal large model, the processing barriers of different types of unstructured data are completely broken down. Text, image, and audio information that could not be directly quantified are transformed into standardized features that can be used for credit assessment. The credit value and risk hidden in unstructured data are deeply mined, and the comprehensiveness, accuracy, and real-time performance of credit feature extraction are comprehensively improved. This allows all kinds of unstructured information to become effective evidence to support corporate credit assessment, making up for the shortcomings of traditional credit assessment that only relies on structured data.

[0043] Feature dynamic fusion module: After receiving structured feature vectors and unstructured credit feature vectors, it first concatenates the two types of feature vectors in an ordered manner to generate an initial feature vector. Then, through attention weight calculation, it dynamically adjusts the weight coefficient of each feature based on the timeliness and importance of different data types. The formula is as follows: ,in, Let be the attention weight for the i-th feature; This is a timeliness weighting coefficient; The timeliness factor is the data corresponding to the i-th feature; This refers to the importance weighting coefficient; Let be the importance factor of the i-th feature. Traditional financial structured indicators are deeply integrated with multi-dimensional unstructured credit features to generate a holographic credit feature representation that comprehensively and accurately reflects the overall credit status of a company. The entire integration process operates in real-time. When new feature data is input into the module, the feature vector is immediately updated and optimized. The updated holographic credit feature representation is transmitted to the dynamic credit assessment module in real time. Attention weight calculation is used to dynamically adapt feature weights, giving higher weights to features that are more timely and critical to credit assessment. This fully leverages the complementary advantages of structured and unstructured data, constructing a dynamic, comprehensive, and high-precision corporate credit feature system. This effectively avoids assessment biases caused by single data types or fixed weights, allowing credit features to accurately match the company's real-time operating status and credit level, providing the most valuable feature foundation for subsequent credit scoring calculations.

[0044] The credit dynamic assessment module: After obtaining the holographic credit feature representation, it immediately performs credit score calculation through a lightweight gradient boosting tree model. It first calculates specific credit scores for five dimensions: financial, operational, compliance, reputation, and performance, using the following formula: ,in, For the j-th dimension, a specific credit score is given. This is a lightweight gradient boosting tree model for the corresponding dimension; This is a holographic representation of credit characteristics; then, through a comprehensive credit scoring method, combined with the weight coefficients of each dimension, the enterprise's comprehensive credit score is obtained, as shown in the formula: ,in, For comprehensive credit scoring; is the weight coefficient for the j-th dimension, where j ranges from 1 to 5, corresponding to the five dimensions of finance, operations, compliance, reputation, and performance, respectively. This module provides a specific credit score for the j-th dimension. It incorporates a rule base for major risk events, continuously monitoring risk signals in the holographic credit feature representation around the clock. When a major risk event is detected, such as a company being listed as a dishonest executor, experiencing a major production safety accident, defaulting on a large debt, having its business license revoked, or receiving a major administrative penalty, all routine data updates and calculations are immediately suspended. System computing resources are prioritized to trigger an emergency credit score update process, completing the recalculation of all credit scores within a specified time. The latest specific credit score and comprehensive credit score are then pushed to the holographic visualization module in real time. A lightweight gradient boosting tree model enables rapid and accurate quantification of credit scores, clearly defining a company's credit performance across various dimensions and its overall credit rating. Relying on the rule base for major risk events, it achieves real-time risk identification and emergency response, rapidly updating credit scores. This provides financial institutions with intuitive, reliable, and real-time quantitative data for loan approval, credit limit adjustments, post-loan risk warnings, and risk disposal decisions, significantly improving the response speed and accuracy of risk assessment in financial institutions' credit risk control and effectively reducing potential losses in credit business.

[0045] Holographic Visualization Module: Upon receiving holographic credit feature representations, comprehensive credit scores, and specific credit scores for each dimension, the module immediately generates a multi-dimensional, interactive visualization interface. This interface comprises five core sections: a credit score dashboard, a feature contribution heatmap, a risk event timeline, a multi-dimensional credit radar chart, and an information source traceability chart. The credit score dashboard displays the dynamic values ​​of the company's comprehensive credit score and the five specific credit scores in real time. The feature contribution heatmap clearly presents the degree of contribution of different credit features to the comprehensive credit score. The risk event timeline comprehensively displays the occurrence and handling of various risk events for the company in chronological order. The multi-dimensional credit radar chart intuitively compares the company's credit performance across different dimensions, and the information source traceability chart clearly marks the original data collection channels for each credit feature. The module also supports multi-dimensional interactive queries and drill-down operations. Users can quickly select target companies through the search box, retrieve historical credit data and score trends for any time period using the time selector, view detailed information and data sources by clicking on features in the feature contribution heatmap, and view specific content and impact by clicking on events in the risk event timeline. This transforms abstract and complex credit data and assessment results into intuitive, easy-to-understand, and interactive visualizations. It comprehensively presents the overall picture of corporate credit, risk details, feature value, and data traceability, simplifying the data analysis and decision-making process for financial institution staff. It facilitates quick access, comparison, and assessment of corporate credit status, enabling efficient, visualized, and intelligent full-process management of loan approval and post-loan risk control.

[0046] This embodiment fully verifies the practical value of a multi-source data fusion-based dynamic enterprise credit assessment and visualization system in the entire credit risk control process of financial institutions. The system achieves comprehensive collection and standardized processing of enterprise credit data through a multi-source data acquisition module, laying a solid data foundation for assessment. The unified semantic parsing module relies on a multimodal large model to complete end-to-end parsing of unstructured data, transforming diverse information into standard credit features and filling the data dimension limitations of traditional assessments. The feature dynamic fusion module achieves dynamic adaptation and deep fusion of features through attention weight calculation, constructing a holographic feature system that accurately reflects the true credit of enterprises. The credit dynamic assessment module uses a lightweight gradient boosting tree model to complete multi-dimensional scoring calculations and implements emergency risk response through a major risk event rule base, ensuring real-time and reliable scoring. The holographic visualization module presents assessment results through a multi-dimensional interactive interface, simplifying the risk control decision-making process, such as... Figure 1 As shown, the entire system enables dynamic assessment, real-time early warning, and visualized management of corporate credit, effectively improving the efficiency of credit approval and risk control capabilities of financial institutions, reducing credit default risk, and providing core support for the intelligent transformation of financial credit business.

[0047] Example 2

[0048] After the system is launched and connected to the core enterprise management system of the supply chain, the multi-source data acquisition module immediately initiates the data acquisition process. This module utilizes distributed crawler technology to achieve parallel deployment across multiple nodes, efficiently crawling credit-related data of upstream and downstream partner companies in the supply chain from publicly available internet channels. It also connects with official government platforms through standardized API interfaces to obtain authoritative government data from partner companies, including business registration, tax, and judicial information. Simultaneously, it leverages data subscription technology to acquire industry and operational data from third-party professional institutions. The module comprehensively collects both structured and unstructured data from partner companies in real-time. Structured data includes financial statements, business registration information, tax payment data, and annual report operational data. Unstructured data includes annual report texts, industry news reports, court announcements, product photos, and executive interview audio. After data acquisition, the module automatically performs data cleaning and format conversion, removing invalid information, standardizing data formats, and regulating data standards. For structured data, it directly extracts features and generates standardized structured feature vectors. For unstructured data, after preprocessing, it stably transmits the data to the unified semantic parsing module. Figure 2 As shown, through full-channel and full-dimensional data collection and purification, a comprehensive credit data source covering all aspects of cooperative enterprises' operations, compliance, performance, and reputation is constructed to ensure that the data quality meets standards and the information is comprehensive and authentic. This provides high-quality data support for subsequent data analysis, feature fusion, and credit assessment, enabling core enterprises to fully grasp the true operating status and credit level of upstream and downstream cooperative enterprises.

[0049] After receiving unstructured data from the multi-source data acquisition module, the unified semantic parsing module uses a pre-trained and finely tuned multimodal large-scale model to perform end-to-end semantic parsing of text, image, and audio data. For text data, it accurately extracts enterprise entities, business events, market sentiment, and credit risk signals. For image data, it effectively identifies product defects, production safety hazards, and trademark infringement. For audio data, it performs speech-to-text conversion and analyzes tone, emotion, and credibility of promises. All parsing results are converted into unstructured credit feature vectors according to a unified dimensional standard. When new unstructured data is input into the module, parsing and feature vector updates are immediately completed, and the updated unstructured credit feature vectors are transmitted in real time to the feature dynamic fusion module. Through the end-to-end parsing capabilities of the multimodal large-scale model, various types of unstructured information are transformed into standardized credit features. This allows for in-depth mining of implicit credit information and potential risks of upstream and downstream partners in terms of operations, compliance, reputation, and performance, comprehensively improving the depth and breadth of credit feature extraction. Unstructured data becomes an important basis for assessing the credit of partner companies, overcoming the shortcomings of traditional supply chain credit management that relies solely on structured data such as financial and business registration information.

[0050] Upon receiving structured and unstructured credit feature vectors, the feature dynamic fusion module sequentially concatenates these two types of vectors to generate an initial feature vector. Through attention weight calculation, it dynamically adjusts the weight coefficient of each feature based on the timeliness and importance of the data, achieving deep fusion of traditional financial structured indicators and multi-dimensional unstructured credit features. This generates a holographic credit feature representation that comprehensively reflects the credit status of supply chain partners. The entire fusion process operates in real-time; feature vectors are updated immediately upon new data input, and the updated holographic credit feature representation is transmitted to the credit dynamic assessment module in real-time. Dynamic optimization of feature weights is achieved through attention weight calculation, ensuring that the latest and most critical credit information dominates the assessment results. By fully integrating the advantages of structured and unstructured data, a holographic credit feature system tailored to the operational characteristics of supply chain companies is constructed. This system accurately reflects the performance capabilities, operational stability, compliance levels, and credit risks of partner companies, providing precise feature support for core enterprises to assess cooperation value and manage cooperation risks.

[0051] After acquiring the holographic credit feature representation, the credit dynamic assessment module uses a lightweight gradient boosting tree model to calculate specific credit scores for five dimensions of the partner company: finance, operations, compliance, reputation, and performance. Then, a comprehensive credit score is derived by combining the weight coefficients of each dimension. The module has a built-in rule base for major risk events, monitoring the credit risk signals of partner companies around the clock. When a major risk event is detected, such as being subject to enforcement for breach of trust, a major safety accident, a large debt default, license revocation, or a major administrative penalty, the regular update task is immediately suspended, and system computing resources are prioritized to trigger an emergency credit score update process. This quickly recalculates the credit score and pushes the latest assessment results to the holographic visualization module in real time. Figure 3 As shown, a lightweight gradient boosting tree model is used to accurately quantify the credit level of partner companies, clearly distinguishing credit performance in various dimensions and overall credit rating. Relying on a major risk event rule base, it enables real-time risk identification and emergency response, promptly identifying sudden performance risks of partner companies. This provides core enterprises with real-time, accurate, and reliable scoring criteria for conducting cooperation access reviews, adjusting cooperation strategies, and managing supply chain risks, effectively mitigating credit risks in supply chain cooperation and ensuring the security and stability of the overall supply chain operation.

[0052] After receiving holographic credit feature representations, comprehensive credit scores, and specific credit scores, the holographic visualization module generates a multi-dimensional visualization interface that includes a credit score dashboard, a feature contribution heatmap, a risk event timeline, a multi-dimensional credit radar chart, and an information source traceability chart. The credit score dashboard displays the partner company's comprehensive credit score and five-dimensional specific credit scores in real time. The feature contribution heatmap shows the degree of contribution of each credit feature to the score. The risk event timeline outlines the partner company's risk events throughout its entire lifecycle. The multi-dimensional credit radar chart compares the partner company's credit performance across different dimensions, and the information source traceability chart clarifies the original data sources of the credit features. Meanwhile, the module supports core enterprises in conducting multi-dimensional interactive queries and drill-down operations. Target partners can be selected via the search box, historical credit data and score changes can be viewed using the time selector, detailed information and data sources can be viewed by clicking on features, and specific content and impact can be viewed by clicking on risk events. This transforms the abstract credit data of supply chain partners into intuitive, clear, and interactive visualizations, comprehensively showcasing the partner's credit advantages, risk weaknesses, characteristic value, and data traceability. This simplifies the core enterprise's supply chain credit management workflow, enabling staff to quickly complete tasks such as partner credit access review, performance capability assessment, and cooperation risk control, improving the efficiency and accuracy of supply chain credit management and helping core enterprises build a stable, secure, and efficient supply chain cooperation system.

[0053] This embodiment fully demonstrates the effectiveness of a multi-source data fusion-based enterprise credit dynamic assessment and visualization system in supply chain credit management. The system comprehensively acquires structured and unstructured credit data from upstream and downstream enterprises through a multi-source data acquisition module. After cleaning and transformation, standardized feature vectors are generated, ensuring comprehensive and accurate data. A unified semantic parsing module performs deep analysis of multimodal unstructured data, extracting implicit credit and risk information to enrich the dimensions of credit assessment. A feature dynamic fusion module integrates two types of features through attention weight calculation, generating holographic credit features tailored to the characteristics of supply chain enterprises, accurately reflecting their performance capabilities and risk levels. A credit dynamic assessment module completes specific and comprehensive credit score calculations, relying on a risk rule base to achieve rapid early warning and score updates for sudden risks. A holographic visualization module presents credit data and assessment results with an intuitive interface, supporting interactive querying and drill-down. The system helps core enterprises efficiently complete partner credit access, performance monitoring, and risk control, ensuring supply chain stability and security, and promoting the intelligent and refined upgrading of supply chain credit management.

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dynamic enterprise credit assessment and visualization system based on multi-source data fusion, characterized in that, The system includes: a multi-source data acquisition module, a unified semantic parsing module, a feature dynamic fusion module, a credit dynamic evaluation module, and a holographic visualization module; The multi-source data acquisition module collects structured and unstructured data from the Internet, government open platforms, corporate websites, news media, and social media channels in real time. After data cleaning and format conversion, it directly extracts structured feature vectors from the structured data and inputs the unstructured data into the unified semantic parsing module. The unified semantic parsing module uses a multimodal large model to perform end-to-end semantic parsing on the input data of different modalities, generates unstructured credit feature vectors in a unified format, updates the parsing results in real time, and inputs them into the feature dynamic fusion module. The feature dynamic fusion module employs an attention mechanism to dynamically adjust the weights of structured and unstructured features based on the timeliness and importance of different data types. This deeply integrates traditional financial structured indicators with unstructured credit features to generate a holographic credit feature representation of the enterprise. The fusion process is performed in real time, and the feature vector is updated immediately when new data is input and then input into the credit dynamic assessment module. The credit dynamic assessment module: based on holographic credit feature representation, it calculates the enterprise's comprehensive credit score and specific credit scores for each dimension in real time through a lightweight gradient boosting tree model; it has a built-in rule base for major risk events, and when a major risk event is detected, it immediately triggers an emergency update of the credit score and inputs the assessment results into the holographic visualization module; The holographic visualization module receives holographic credit feature representations, comprehensive credit scores, and specific credit score data, and generates a multi-dimensional visualization interface that supports interactive user queries and drill-downs.

2. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The multi-source data acquisition module deploys multiple nodes in parallel to crawl publicly available data using distributed crawler technology, connects to government official platforms through API interfaces to obtain authoritative data, and acquires third-party professional data through data subscription technology. The collected structured data includes financial statements, business registration information, tax data, and annual report operating data, while the unstructured data includes annual report texts, news reports, court announcements, product images, and audio recordings of executive interviews.

3. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The unified semantic parsing module uses a pre-trained and finely tuned multimodal large model to perform end-to-end semantic parsing on data from different modalities. For text data, it extracts entities, events, sentiment, and risk signals; for image data, it identifies product defects, production safety hazards, and trademark infringement; for audio data, it converts it into text and analyzes tone, emotion, and credibility of promises; all parsing results are converted into unstructured credit feature vectors of a unified dimension, and are immediately updated and transmitted to the feature dynamic fusion module when new data arrives.

4. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The feature dynamic fusion module uses an attention mechanism to calculate the attention weight of each feature, specifically: The initial feature vector is obtained by concatenating the structured feature vector output by the multi-source data acquisition module and the unstructured credit feature vector output by the unified semantic parsing module. The attention weight of the i-th feature is calculated according to the formula. : in, Let be the attention weight for the i-th feature; This is a timeliness weighting coefficient; The timeliness factor is the data corresponding to the i-th feature; This refers to the importance weighting coefficient; Let i be the importance factor of the i-th feature; The initial feature vector is weighted feature by feature based on attention weights to generate a weighted holographic credit feature representation.

5. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1 or 4, characterized in that, The timeliness factor in the feature dynamic fusion module It is negatively correlated with the time difference between data release time and system time; the smaller the time difference, the better. The larger the value, the higher the value, ranging from 0 to 1; the importance factor The degree of influence of characteristics based on historical data statistics on credit scores is determined by the magnitude of the influence. The larger the value, the higher the value, ranging from 0 to 1; the timeliness weight coefficient And importance weight coefficient satisfy It can be adjusted according to the characteristics of the industry.

6. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The credit dynamic assessment module, based on holographic credit feature representation, calculates specialized credit scores for five dimensions using a lightweight gradient boosting tree model. The formula is as follows: in, For the j-th dimension, a specific credit score is given. This is a lightweight gradient boosting tree model for the corresponding dimension; This represents the holographic credit feature. The comprehensive credit score is calculated based on the specific credit scores of each dimension, using the following formula: in, For comprehensive credit scoring; is the weight coefficient for the j-th dimension, where j ranges from 1 to 5, corresponding to the five dimensions of finance, operations, compliance, reputation, and performance, respectively. For the j-th dimension, a specific credit score is given. The specific credit scores and comprehensive credit scores for the five dimensions all range from 0 to 100 points, with higher scores indicating better creditworthiness in the corresponding dimension or overall. The weighting coefficients for the five dimensions satisfy the following... The system has built-in default weight configurations for different industries, and users can customize and adjust them according to their actual needs.

7. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The credit dynamic assessment module has a built-in rule base for major risk events and continuously monitors risk signals in the holographic credit feature representation. When a major risk event is detected, such as a company being listed as a dishonest executor, experiencing a major safety accident, defaulting on a large amount of debt, having its business license revoked, or being subject to a major administrative penalty, all ordinary update tasks are immediately suspended, computing resources are prioritized to trigger an emergency credit score update process, and all score calculations are completed within 10 seconds and pushed to the holographic visualization module.

8. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The visualization interface generated by the holographic visualization module includes a credit scoring dashboard, a feature contribution heatmap, a risk event timeline, a multi-dimensional credit radar chart, and an information source traceability chart. The credit scoring dashboard displays the company's comprehensive credit score and specific credit scores for each dimension in real time. The feature contribution heatmap shows the degree of contribution of different features to the comprehensive credit score. The risk event timeline displays all risk events that have occurred to the company in chronological order. The multi-dimensional credit radar chart provides an intuitive comparison of the company's credit performance across different dimensions. The information source traceability chart shows the original data source for each credit feature.

9. The enterprise credit dynamic assessment and visualization system based on multi-source data fusion according to claim 1, characterized in that, The holographic visualization module supports multi-dimensional interactive queries and drill-down operations. Users can select different target companies through the search box, view historical credit data and score change trends for any time period through the time selector, click on any feature in the feature contribution heatmap to view detailed information and data sources of the feature, and click on any event in the risk event timeline to view the detailed content and impact of the event.

Images