An AI model-based multi-source data standardization and credit investigation report generation method and system

Abstract

The present application relates to the technical field of data processing, and more particularly to a multi-source data regularization and credit reporting generation method and system based on an AI model, which comprises: acquiring a multi-source business data set, identifying a business change event, determining an event influence path and associated business objects, determining a credit data calling range, performing event dimension regularization, and generating a dynamic credit report. The present application acquires a multi-source business data set of a target enterprise in multiple heterogeneous data sources, actively identifies a business change event in the supply chain based on a model, analyzes the influence chain of the event using a relationship graph, accurately and efficiently extracts and regularizes data strongly related to the event from the multi-source data, and finally generates a dynamic credit report focusing on the influence of the event, thereby achieving automatic identification of business events and on-demand regularization of multi-source data, significantly improving the response speed of post-loan management and the pertinence of risk analysis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for multi-source data normalization and credit report generation based on an AI model. Background Technology

[0002] In supply chain finance, financial institutions provide financing services to companies in the upstream and downstream of the supply chain, which is an important means of supporting the development of the real economy. After the loan is issued, financial institutions need to conduct continuous post-loan management of the financing company's operations to monitor and prevent credit risk.

[0003] Currently, the most common method for updating credit data in post-loan management is periodic. For example, financial institutions will re-collect the financial data, transaction records, and publicly available credit records of borrowing companies at fixed intervals. These data are then comprehensively analyzed by risk control models or manual processes to generate a comprehensive credit report. This method can serve a certain monitoring function when the company's operating status is relatively stable.

[0004] However, supply chain operations are highly dynamic and interconnected. For example, the operating conditions of downstream suppliers to core enterprises are closely related to the core enterprises' orders. When a core enterprise adjusts its production plan due to market changes and suddenly cancels a batch of purchase orders from a certain supplier, this "order cancellation" event will directly affect the supplier's future sales revenue, corresponding accounts receivable, and related contract performance status.

[0005] Existing technologies have significant shortcomings in such scenarios. First, credit report updates are based on fixed cycles rather than being triggered by key business events, leading to delayed risk monitoring. When an order cancellation occurs, the system cannot respond immediately and must wait until the next reporting cycle to detect the data change. Second, there is a lack of business-level correlation between multi-source data. Order data, contract data, and fund settlement data are usually stored in different heterogeneous systems. When an order changes, it is difficult to automatically and quickly locate its associated contracts, performance records, and accounts receivable, resulting in inefficient data processing and a high risk of omissions. Finally, the generated credit reports are usually comprehensive but lack specificity, failing to clearly reveal "which specific changes in credit-related data were caused by a particular business change event," placing an additional interpretive burden on risk analysts. Summary of the Invention

[0006] To address this, the present invention provides a method and system for multi-source data normalization and credit report generation based on an AI model. This method and system can normalize multi-source credit data by identifying business change events and normalizing the event dimensions based on the event impact path. This overcomes the problems of delayed risk identification and lack of specificity in credit reports caused by fixed-period data updates and the weak correlation between scattered and heterogeneous multi-source data in the prior art.

[0007] To achieve the above objectives, this invention provides a method for multi-source data normalization and credit report generation based on an AI model, comprising:

[0008] Acquire business status data of the target enterprise from multiple heterogeneous business data sources to form a multi-source business data set. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data.

[0009] Extract time-series data reflecting changes in business status from the multi-source business data set and input it into a preset business event recognition model to identify business change events corresponding to the target enterprise;

[0010] Based on the event entity identifier in the business change event, according to the preset supply chain business relationship map, the path is traced according to the business association relationship corresponding to the business change event to determine the event impact path, and the associated business object corresponding to the business change event is determined according to the event impact path.

[0011] Based on the event type of the business change event, the event impact path, and the associated business object, the corresponding credit data retrieval scope is determined from the multi-source business data set, and the business data within the credit data retrieval scope is extracted to obtain event-related credit data.

[0012] The event-related credit data is normalized according to the event dimension to obtain an event normalized credit data package. The event dimension normalization includes arranging the related data according to the occurrence time of the business change event and uniformly mapping the subjects from different data sources according to the related business objects.

[0013] Based on the event-structured credit data package, a dynamic credit report is generated that presents the impact of business change events.

[0014] Furthermore, the preset business event recognition model is a time series-based machine learning model. The model is obtained by training on a historical business state sequence including at least one of order quantity, order amount, delivery quantity, and payment amount, and is used to output the event type, event occurrence time, and event entity identifier of the business change event based on the input business state time series of the target enterprise.

[0015] Furthermore, the business change events include: order cancellation events, order reduction events, fulfillment abnormal events, or payment abnormal events.

[0016] Furthermore, the preset supply chain business relationship graph includes enterprise nodes, order nodes, contract nodes, product nodes, fulfillment nodes, and funding nodes. Different nodes are connected through enterprise relationships, order contract relationships, contract fulfillment relationships, and fulfillment settlement relationships.

[0017] Furthermore, the process of determining the impact path of the event includes: taking the business object corresponding to the event entity identifier in the business change event as the starting node, and tracing the order node, the contract node, the performance node, and the funding node in the preset supply chain business relationship graph according to the business association relationship corresponding to the business change event.

[0018] Furthermore, the process of determining and extracting the corresponding credit data access scope includes: when the business change event is an order cancellation event, accessing the unexecuted order data, contract remaining amount data, and unsettled accounts data associated with the order; when the business change event is a performance abnormality event, accessing the historical delivery records and contract performance status data associated with the order; and when the business change event is a payment abnormality event, accessing the accounts receivable details, historical payment flow, and contractual payment period data associated with the payment abnormality event.

[0019] Furthermore, the event dimension normalization includes: using the occurrence time of the business change event as a time benchmark, arranging the event-related credit data by time to form a data sequence before the event and a data sequence after the event; and mapping the same enterprise entity from different data sources to the same business entity according to the unified enterprise code or subject mapping relationship corresponding to the related business object.

[0020] Furthermore, the dynamic credit report includes: the type and timing of the business change event, a description of the affected business scope, and a comparative analysis of changes in key data.

[0021] Furthermore, the dynamic credit report further includes an event impact topology diagram, which visually presents the chain of event impacts triggered by the business change event and propagated between the order node, the contract node, the performance node, and the funding node.

[0022] On the other hand, the present invention provides a multi-source data normalization and credit report generation system based on an AI model, comprising:

[0023] The data acquisition module is used to acquire business status data of the target enterprise from multiple heterogeneous business data sources to form a multi-source business data set. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data.

[0024] The event recognition module is used to extract time-series data reflecting changes in business status from the multi-source business data set and input it into a preset business event recognition model to identify business change events corresponding to the target enterprise.

[0025] The path generation module is used to perform path tracing based on the event entity identifier in the business change event, according to the preset supply chain business relationship map, and according to the business association relationship corresponding to the business change event, to determine the event impact path, and to determine the associated business object corresponding to the business change event based on the event impact path.

[0026] The scope determination module is used to determine the corresponding credit data retrieval scope from the multi-source business data set based on the event type of the business change event, the event impact path, and the associated business object, and to extract the business data within the credit data retrieval scope to obtain event-related credit data.

[0027] The data normalization module is used to normalize the event-related credit data in terms of event dimensions to obtain an event normalized credit data package. The event dimension normalization includes arranging the related data in terms of time according to the occurrence time of the business change event, and uniformly mapping the subjects from different data sources according to the related business objects.

[0028] The report generation module is used to generate dynamic credit reports that present the impact of business change events based on the credit data packets organized by the event.

[0029] Compared with existing technologies, the advantages of this invention are as follows: Addressing the dynamic fluctuations and fragmented, isolated nature of supply chain business status, this invention constructs a data foundation by centrally collecting multi-source heterogeneous business status data. Then, utilizing a pre-defined business event identification model, it accurately captures status transition events from time-series data, achieving a mechanism shift from passive full-data processing to proactive event triggering. Furthermore, to clarify the impact scope of isolated events within complex business networks, it traces paths in a pre-defined supply chain business relationship graph using event entity identifiers, expanding isolated events into a network of affected business objects. Subsequently, it dynamically extracts strongly correlated credit data based on event type and associated objects, avoiding redundant processing of irrelevant data. Further, it performs time-based arrangement and unified subject mapping of multi-source heterogeneous data around the event occurrence time, ensuring data consistency and usability across time and subject dimensions. Finally, it generates dynamic credit reports focusing on the event's impact based on the standardized data packets. This achieves event-driven, on-demand data standardization and report generation, effectively overcoming the technical problems of delayed response and weak data correlation leading to a lack of targeted reports in traditional periodic models. This provides timely and targeted decision support for post-loan risk management in supply chain finance. Attached Figure Description

[0030] Figure 1 This is a flowchart of the multi-source data normalization and credit report generation method based on an AI model in this embodiment;

[0031] Figure 2 This is a flowchart for identifying service change events in this embodiment;

[0032] Figure 3 This is a detailed flowchart illustrating the generation of dynamic credit reports based on business change events in this embodiment.

[0033] Figure 4 This is a schematic diagram of the AI-based multi-source data normalization and credit report generation system in this embodiment. Detailed Implementation

[0034] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0035] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0036] In the post-loan management scenario of supply chain finance, financial institutions need to continuously monitor the operating status of the financed enterprises. Traditional post-loan management often adopts a periodic and comprehensive data collection and report generation model, such as generating a comprehensive credit report monthly or quarterly. However, supply chain business is highly dynamic. When key business change events occur, such as the core enterprise canceling large orders, this periodic reporting model is often untimely and cannot immediately reveal the specific impact of the event on the enterprise's future cash flow and performance capabilities. At the same time, because order data, contract data, performance data, and settlement data are scattered in different business systems, the data correlation is weak, making it difficult to quickly and accurately locate the data range related to specific business events, and the generated reports therefore lack specificity. This invention proactively identifies business change events in the supply chain through a preset business event identification model, and uses a preset supply chain business relationship graph to analyze the impact chain of the event, thereby accurately and efficiently extracting and organizing data strongly related to the event from multi-source data, and finally generating a dynamic credit report focusing on the impact of the event.

[0037] To better understand this invention, the core terms involved will first be explained. The pre-defined supply chain business relationship graph refers to a model that pre-stores various business objects in the supply chain and their interrelationships in the form of a data structure of nodes and edges. The pre-stored business objects include enterprises, orders, contracts, products, fulfillment tasks, and accounts receivable, etc. This graph solidifies the business logic, enabling quick retrieval of all related objects starting from any business object. The event impact path, in this graph, is one or more paths triggered by an identified business change event, propagating along the pre-defined business relationships within the graph. This path reveals the scope of business objects affected by the event.

[0038] Event dimension normalization is a data processing method centered on specific business change events. Its key lies in extracting scattered event-related data from multiple data sources, unifying the subject identity and aligning the occurrence time around the event, forming a structured, context-clear data set—the event normalization credit data package. The credit data refers to business credit-related data reflecting the target enterprise's business performance capabilities, transaction stability, and fund repayment status, including order data, contract execution data, performance data, accounts receivable data, and cash flow data. This credit data can originate from the core enterprise's supply chain management system, enterprise resource planning system, fund settlement system, and other data sources related to the target enterprise's business activities. Through this design, the present invention enables rapid response and in-depth analysis of business changes.

[0039] Please see Figure 1The diagram shown is a flowchart of the multi-source data normalization and credit report generation method based on an AI model in this embodiment. The method in this embodiment includes:

[0040] Step S101: Obtain the target enterprise's business status data from multiple heterogeneous business data sources to form a multi-source business data set. This enables centralized management of business data scattered across different systems, resolving the data silo problem. Specifically, the heterogeneous business data sources include the core enterprise's supply chain management system, enterprise resource planning system, and bank's fund settlement system, etc. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data.

[0041] Step S102: Extract time-series data reflecting changes in business status from the multi-source business data set, and input the time-series data into a preset business event identification model to identify business change events corresponding to the target enterprise. This automated model identifies potential risk signals of status changes that may affect the enterprise's creditworthiness from massive, continuous business data streams. Specifically, the preset business event identification model is used to identify events with specific business meanings from continuous or discrete data changes. The business change events include at least one of order status change events, fulfillment status change events, and settlement status change events.

[0042] Step S103: Based on the event entity identifier contained in the business change event, and taking the business object corresponding to the event entity identifier as the starting node, in the preset supply chain business relationship graph, path tracing is performed according to the business association relationship corresponding to the business change event to generate one or more event impact paths, and the associated business object corresponding to the business change event is determined according to the event impact path. The event entity identifier includes the changed order number or contract number, etc.

[0043] Specifically, the business relationships corresponding to the business change events are used to define the graph edges that have a business transmission relationship with the current business change event. For example, for an order cancellation event, the business relationships at least include the order-contract relationship between the order and the contract, the contract performance relationship between the order and the fulfillment object, and the performance settlement relationship between the fulfillment object and the funding object; for an abnormal payment event, the business relationships at least include the funding relationship between payment receipts and accounts receivable, and the contract relationship between accounts receivable and the contract.

[0044] Specifically, the system starts from the starting node corresponding to the event entity identifier, traces the path only along the business association relationship corresponding to the business change event, and identifies the business objects passed through on the path as associated business objects, thereby limiting the scope of the event's impact and avoiding including other business objects that do not have a business transmission relationship with the current business change event in the subsequent data processing scope.

[0045] Specifically, this invention utilizes the association query capability of graphs to analyze the chain reactions that business change events may produce in complex supply chain business networks, clarify the transmission chain from the event trigger point to all potentially affected business objects, and expand an isolated event point into a network of affected business objects, thereby accurately defining the boundaries of the analysis.

[0046] Step S104: Based on the event type of the business change event, the event impact path, and the associated business object, determine the corresponding credit data retrieval range from the multi-source business data set, and extract the business data within the credit data retrieval range to obtain event-related credit data.

[0047] Specifically, the present invention determines the data types that need to be focused on based on the event type of the business change event, determines the business objects that need to be queried based on the event impact path, and then determines the corresponding data query range based on the data association relationship of the business objects in different data sources, thereby obtaining the event-related credit data corresponding to this event.

[0048] For example, for an order cancellation event, the system determines the contract, performance, and funding objects associated with the cancelled order based on the event impact path corresponding to the order cancellation event, and further calls the unexecuted order data, contract remaining amount data, and unsettled accounts data associated with the cancelled order; for a performance exception event, it calls the historical delivery records and contract performance status data associated with the corresponding order; for a payment exception event, it calls the accounts receivable details, historical payment flow, and contractual payment period data associated with the corresponding payment exception event.

[0049] This allows the scope of credit data access to correspond to current business change events and their impact paths, without the need to indiscriminately access and process all data in a multi-source business data set.

[0050] Step S105: Perform event dimension normalization on the event-related credit data to obtain an event normalized credit data package; specifically, the event dimension normalization uses the business change event as the data organization benchmark, including time dimension normalization and subject dimension normalization.

[0051] In the time-dimensional normalization process, the occurrence time of the business change event is used as the time benchmark to arrange the event-related credit data by time, and to form a data sequence before the event and a data sequence after the event according to the time sequence, so as to facilitate the corresponding comparison of the business status before and after the business change event.

[0052] In the subject-dimensional normalization process, based on the unified enterprise code or preset subject mapping relationship corresponding to the associated business objects, the same enterprise entities from different data sources are uniformly mapped to the same business entity. For example, the supplier name in the order system, the contract entity name in the contract system, and the payee entity in the fund settlement system are mapped to the same enterprise entity.

[0053] Specifically, through the above-mentioned regularization process, the present invention forms an event-regularized credit data package centered on the business change event, with a unified subject identifier and a unified time reference.

[0054] Step S106: Generate a dynamic credit report based on the event-organized credit data package, which presents the impact of the business change event. The dynamic credit report presents at least the business change event information, the related business object information, and the data change information corresponding to the business change event, so as to provide decision-makers with timely and targeted risk information support for the in-depth impact analysis of the specific business change event.

[0055] Specifically, this invention addresses the dynamic fluctuations and fragmented, isolated nature of supply chain business status. It constructs a data foundation by centrally collecting multi-source heterogeneous business status data, and then utilizes a pre-defined business event identification model to accurately capture status transition events from time-series data. This achieves a mechanism shift from passive full-data processing to proactive event triggering. Furthermore, to clarify the impact scope of isolated events within complex business networks, path tracing is performed using event entity identifiers within a pre-defined supply chain business relationship graph, expanding isolated events into a network of affected business objects. Subsequently, strongly correlated credit data is dynamically extracted based on event type and associated objects, avoiding redundant processing of irrelevant data. Further, the multi-source heterogeneous data is chronologically arranged and uniformly mapped to subjects based on event occurrence time, ensuring data consistency and usability across time and subject dimensions. Finally, dynamic credit reports focusing on event impact are generated based on the regulated data packets. This achieves event-driven, on-demand data regulation and report generation, effectively overcoming the technical problems of delayed response and weak data correlation leading to a lack of targeted reports in traditional periodic models. This provides timely and targeted decision support for post-loan risk management in supply chain finance.

[0056] Based on the above solution, in order to effectively identify business trend events that accumulate over time, as a preferred implementation method, the preset business event identification model is preferably a time-series-based machine learning model. The business trend events include non-instantaneous state changes such as a continuous decline in order volume.

[0057] Please see Figure 2 As shown, this is a flowchart for identifying business change events in this embodiment. In this embodiment, the process of constructing a preset business event identification model includes:

[0058] Time-series data reflecting continuous changes in business status are extracted from multi-source business data to form a historical business status sequence. The time-series data includes historical order quantity, total order amount, on-time delivery quantity, and actual payment amount, which are statistically analyzed on a daily or weekly basis. The historical business status sequence serves as the model input for a preset business event identification model. Each time point in the historical business status sequence corresponds to at least one business status feature, which includes at least one of the following: order quantity, order amount, delivery quantity, and actual payment amount.

[0059] The historical business state sequence is input into a pre-trained business event recognition model. Specifically, the pre-trained business event recognition model can adopt architectures such as long short-term memory networks, gated recurrent units, or Transformer-based time series models to capture long-term dependencies in the sequence data. The model training process is offline, using a large amount of historical business data, which includes samples of known specific business events. Through supervised learning of these historical sequences with event type labels, the model can learn the specific data patterns corresponding to different business change events.

[0060] The trained preset business event recognition model receives a new business state time series of the target enterprise as input, and outputs business change event information according to the state change pattern in the business state time series. The business change event information includes at least the event type, the event occurrence time, and the event entity identifier, so that the model can automatically identify the business change event corresponding to the target enterprise from the continuous data stream.

[0061] Finally, the model outputs structured business change event information, which includes event type, specific time of event occurrence, entity identifier directly associated with the event, and other relevant contextual information. Specifically, for business change events of state change mutation type, the model identifies the corresponding event based on the business state change between adjacent time nodes; for business change events of trend change type, the model identifies the corresponding event based on the continuous change trend of the business state sequence within a preset time range.

[0062] Specifically, the present invention can not only identify sudden events such as order cancellation, but also slow-moving risk events such as a continuous decrease in orders through the above-described method.

[0063] Furthermore, as a preferred implementation, the business change events can specifically include several typical events that significantly impact supply chain stability, to adapt to complex business scenarios within the supply chain. Specifically, the business change events include: order cancellation events, order reduction events, fulfillment anomaly events, or payment anomaly events. An order cancellation event refers to an order status changing from normal execution to cancellation; an order reduction event refers to an order size showing a continuous downward trend over a period of time; a fulfillment anomaly event may refer to frequent or severe delays in actual delivery dates compared to planned delivery dates; and a payment anomaly event refers to accounts receivable actually being collected significantly later than the payment cycle stipulated in the contract.

[0064] Furthermore, this invention demonstrates a high degree of specificity in determining the scope of credit data retrieval for different types of business change events. Specifically, when the business change event is an order cancellation event, the system prioritizes retrieving data on unexecuted orders associated with the cancelled order, the remaining amount data of the corresponding contract, and any accounts receivable that may have been generated but not yet settled. When the business change event is a performance anomaly event, the system prioritizes retrieving historical delivery records associated with the order, records of reasons for delays, and contractual clauses regarding performance status. Performance anomalies include events such as delivery delays.

[0065] Specifically, the present invention, through the aforementioned event-type-based differentiated data retrieval strategy, enables the system to extract highly relevant data for analysis based on the different impacts of different events on the company's future revenue, production plans, or cash flow health, thereby greatly improving the relevance and effectiveness of data analysis.

[0066] Please see Figure 3 As shown, it is a flowchart of generating a dynamic credit report based on business change events in this embodiment; in order to achieve accurate analysis from business change events to the scope of impact, this invention relies on a pre-constructed preset supply chain business relationship map.

[0067] In a preferred embodiment, the supply chain business relationship graph includes various types of nodes and edges. The nodes include enterprise nodes representing various company entities in the supply chain, as well as order nodes, contract nodes, product nodes, fulfillment nodes, and funding nodes. Fulfillment nodes include business objects such as delivery notes or acceptance slips. Funding nodes include business objects such as accounts receivable or payment records. Different nodes are connected by edges with clear business meanings. Specifically, enterprise nodes are connected by procurement or supply relationship edges; order nodes and contract nodes are connected by order-contract relationship edges; order nodes and fulfillment nodes are connected by contract-fulfillment relationship edges; and fulfillment nodes and funding nodes are connected by fulfillment-settlement relationship edges.

[0068] Based on the constructed graph, the system generates event impact paths. Specifically, the system obtains event entity identifiers from the identified business change event information. These event entity identifiers include the order number of the canceled order in an order cancellation event, the order number of the corresponding order in a fulfillment anomaly event, and the business identifier of the corresponding accounts receivable or payment receipt in a payment anomaly event. The system uses these event entity identifiers as the starting node in the supply chain business relationship graph. Subsequently, the system traces paths in the supply chain business relationship graph from the starting node according to the predefined business associations corresponding to the business change events.

[0069] Specifically, the path tracing employs a breadth-first search algorithm, a depth-first search algorithm, or other customized graph traversal algorithms. For example, for an order cancellation event, the system starts with the order node corresponding to the cancelled order, traces along the order contract relationship to the corresponding contract node, then along the order fulfillment relationship to the corresponding fulfillment node, and finally from the fulfillment node along the fulfillment settlement relationship to the corresponding funds node, thus forming an event impact path from order to contract and from order to fulfillment to funds. The system aggregates the business object nodes traversed in the above path to form a set of associated business objects corresponding to the order cancellation event.

[0070] Specifically, for abnormal payment events, the system uses the corresponding payment flow or the fund node corresponding to the accounts receivable as the starting node, traces the relationship between funds and contracts to the corresponding contract node, and determines the corresponding enterprise node and other objects that are business-related to the abnormal payment event based on the contract node.

[0071] Specifically, by using graphs for path tracing, this invention can clearly and completely reveal the transmission chain of a single business event in a complex and intertwined supply chain relationship network. Compared with directly performing indiscriminate association queries on multiple data tables, it can limit the scope of data association according to the business relationship corresponding to the business change event, thereby improving the efficiency and accuracy of locating event-related data.

[0072] The dynamic credit report generated by this invention is used to present the impact of business change events. In a preferred embodiment, the dynamic credit report includes at least the type and timing of the business change event. The report also includes a description of the affected business scope and a comparative analysis of changes in key data. This comparative analysis includes displaying changes in the target company's total outstanding orders and expected accounts receivable before and after the event, presented in tabular or chart form.

[0073] To further enhance the readability and intuitiveness of the report, the dynamic credit report also includes an event impact topology diagram. This diagram is a visual representation of the aforementioned event impact paths. Specifically, the data structure of the event impact paths is converted into topology diagram description data; the system draws a graph based on this data; different business objects are displayed as nodes with different icons or shapes in the topology diagram; these business objects include orders, contracts, and enterprises; the relationships between these objects are represented by arrowed line segments; thus clearly presenting the event impact path triggered by the business change event, along the order node, contract node, and performance node to the funding node; users can interact with the topology diagram; when a user selects any business node, the system displays detailed information about that node, including the remaining contract amount or the current order status.

[0074] Specifically, this invention, through the content structure of the aforementioned dynamic credit report, enables report readers to quickly grasp the core information, breadth of impact, and quantifiable effects of an event. Simultaneously, this invention presents complex business relationships to users in an intuitive way through an event impact topology diagram, thereby lowering the barrier to understanding and improving the report's decision support efficiency.

[0075] On the other hand, this embodiment also provides a multi-source data normalization and credit report generation system based on an AI model;

[0076] Please see Figure 4 As shown, this is a schematic diagram of the multi-source data normalization and credit report generation system based on an AI model in this embodiment. The system in this embodiment includes:

[0077] The data acquisition module is used to acquire business status data of the target enterprise from multiple heterogeneous business data sources to form a multi-source business data set. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data.

[0078] The event recognition module is used to extract time series data reflecting changes in business status from the multi-source business data set and input it into a preset business event recognition model to identify business change events corresponding to the target enterprise. The business change events include at least one of order status change events, performance status change events, and settlement status change events.

[0079] The path generation module is used to determine the event impact path based on the event entity identifier in the business change event and according to the preset supply chain business relationship graph, and to determine the associated business object corresponding to the business change event based on the event impact path.

[0080] The scope determination module is used to determine and extract the corresponding credit data retrieval scope from the multi-source business data set based on the event type of the business change event and the associated business object, so as to obtain event-related credit data.

[0081] The data normalization module is used to normalize the event-related credit data according to the event dimension to obtain the event normalized credit data package. The normalization includes arranging the related data according to the occurrence time of the business change event and uniformly mapping the subjects from different data sources according to the related business objects.

[0082] The report generation module is used to generate dynamic credit reports that present the impact of business change events based on the credit data packets organized by the event.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for multi-source data normalization and credit report generation based on an AI model, characterized in that, include: Acquire business status data of the target enterprise from multiple heterogeneous business data sources to form a multi-source business data set. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data. Extract time-series data reflecting changes in business status from the multi-source business data set and input it into a preset business event recognition model to identify business change events corresponding to the target enterprise; Based on the event entity identifier in the business change event, according to the preset supply chain business relationship map, the path is traced according to the business association relationship corresponding to the business change event to determine the event impact path, and the associated business object corresponding to the business change event is determined according to the event impact path. Based on the event type of the business change event, the event impact path, and the associated business object, the corresponding credit data retrieval scope is determined from the multi-source business data set, and the business data within the credit data retrieval scope is extracted to obtain event-related credit data. The event-related credit data is normalized according to the event dimension to obtain an event normalized credit data package. The event dimension normalization includes arranging the related data according to the occurrence time of the business change event and uniformly mapping the subjects from different data sources according to the related business objects. Based on the event-structured credit data package, a dynamic credit report is generated that presents the impact of business change events.

2. The method for multi-source data normalization and credit report generation based on an AI model according to claim 1, characterized in that, The preset business event recognition model is a time series-based machine learning model. The model is obtained by training on a historical business state sequence including at least one of order quantity, order amount, delivery quantity and payment amount, and is used to output the event type, event occurrence time and event entity identifier of the business change event based on the input business state time series of the target enterprise.

3. The method for multi-source data normalization and credit report generation based on an AI model according to claim 1, characterized in that, The business change events include: order cancellation events, order reduction events, performance abnormal events, or payment abnormal events.

4. The method for multi-source data normalization and credit report generation based on an AI model according to claim 1, characterized in that, The preset supply chain business relationship graph includes enterprise nodes, order nodes, contract nodes, product nodes, fulfillment nodes, and funding nodes. Different nodes are connected through enterprise relationships, order contract relationships, contract fulfillment relationships, and fulfillment settlement relationships.

5. The method for multi-source data normalization and credit report generation based on an AI model according to claim 4, characterized in that, The process of determining the impact path of an event includes: taking the business object corresponding to the event entity identifier in the business change event as the starting node, and tracing the order node, the contract node, the performance node, and the funding node in the preset supply chain business relationship graph according to the business association relationship corresponding to the business change event.

6. The method for multi-source data normalization and credit report generation based on an AI model according to claim 3, characterized in that, The process of determining and extracting the corresponding credit data call scope includes: when the business change event is an order cancellation event, calling the unexecuted order data, contract remaining amount data, and unsettled accounts data associated with the order; when the business change event is a performance abnormality event, calling the historical delivery records and contract performance status data associated with the order; and when the business change event is a payment abnormality event, calling the accounts receivable details, historical payment flow, and contractual payment period data associated with the payment abnormality event.

7. The method for multi-source data normalization and credit report generation based on an AI model according to claim 1, characterized in that, The event dimension normalization includes: using the occurrence time of the business change event as a time benchmark, arranging the event-related credit data by time to form a data sequence before the event and a data sequence after the event; and mapping the same enterprise entity from different data sources to the same business entity according to the unified enterprise code or entity mapping relationship corresponding to the related business object.

8. The method for multi-source data normalization and credit report generation based on an AI model according to claim 1, characterized in that, The dynamic credit report includes: the type and time of the business change event, a description of the affected business scope, and a comparative analysis of changes in key data.

9. The method for multi-source data normalization and credit report generation based on an AI model according to claim 8, characterized in that, The dynamic credit report further includes an event impact topology diagram, which visually presents the chain of event impacts triggered by the business change event and propagated between the order node, the contract node, the performance node, and the funding node.

10. A system for multi-source data normalization and credit report generation based on an AI model, based on the method for multi-source data normalization and credit report generation based on an AI model as described in claims 1-9, characterized in that, include: The data acquisition module is used to acquire business status data of the target enterprise from multiple heterogeneous business data sources to form a multi-source business data set. The business status data includes at least order status data, contract execution data, performance status data, and fund settlement status data. The event recognition module is used to extract time-series data reflecting changes in business status from the multi-source business data set and input it into a preset business event recognition model to identify business change events corresponding to the target enterprise. The path generation module is used to perform path tracing based on the event entity identifier in the business change event, according to the preset supply chain business relationship map, and according to the business association relationship corresponding to the business change event, to determine the event impact path, and to determine the associated business object corresponding to the business change event based on the event impact path. The scope determination module is used to determine the corresponding credit data retrieval scope from the multi-source business data set based on the event type of the business change event, the event impact path, and the associated business object, and to extract the business data within the credit data retrieval scope to obtain event-related credit data. The data normalization module is used to normalize the event-related credit data in terms of event dimensions to obtain an event normalized credit data package. The event dimension normalization includes arranging the related data in terms of time according to the occurrence time of the business change event, and uniformly mapping the subjects from different data sources according to the related business objects. The report generation module is used to generate dynamic credit reports that present the impact of business change events based on the credit data packets organized by the event.

Images