Business risk analysis method and device, electronic equipment and storage medium

Abstract

The invention discloses a business risk analysis method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence or other related fields, and the method comprises the steps: obtaining target multi-source data related to a target financial business from N financial data sources, N being a positive integer; the target multi-source data is input into a preset target model, a data analysis result in a natural language form is output, and the preset target model is a large model adjusted in advance according to the financial field direction and is used for analyzing risk data relevance of the target multi-source data and expressing the risk data relevance in the natural language form; and filling a preset report template with the data analysis result in the natural language form to obtain a business risk analysis report of the target financial business. Through the method and the device, the technical problem of low accuracy of a business risk report caused by insufficient data integration analysis in related technologies is solved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology or other related fields. Specifically, it relates to a method and apparatus for analyzing business risks, electronic devices, and storage media. Background Technology

[0002] With the rapid development of fintech, especially in the management of inclusive finance, the generation and analysis of post-business risk reports are at a critical stage of technological transformation. For a long time, financial institutions have relied on a combination of traditional template-filling systems and rule engines to monitor and assess the health of their loan portfolios. However, this inherent model, relying on a single structured data processing method, can no longer meet the increasingly diverse data types and complex business scenarios of today's financial institutions.

[0003] On the one hand, static template-based systems can only handle limited structured data and cannot delve into the inherent relationships between data, especially the key information hidden in unstructured text, resulting in a lack of in-depth insight in risk reports. On the other hand, while rule engines compensate for the shortcomings of template systems to some extent by performing preliminary data analysis based on preset logic, their rigid analytical framework struggles to cope with complex and ever-changing business scenarios, especially when dealing with cross-organizational and multi-product portfolio analyses. Moreover, rule engines are also helpless against risk signals in unstructured data, often resulting in information blind spots in risk reports and reducing the accuracy of business decisions.

[0004] Existing technologies have significant shortcomings in processing data integration and analysis, failing to effectively integrate multi-source data from both inside and outside financial institutions. This deficiency in data integration and analysis not only limits the depth and breadth of risk reports but also results in low report generation efficiency, heavy reliance on manual labor, and difficulty in reflecting market dynamics and changes in customer needs in real time.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] The main objective of this application is to provide a business risk analysis method, apparatus, electronic device, and storage medium to at least solve the technical problem of insufficient data integration and analysis in related technologies, which leads to low accuracy in business risk reports.

[0007] To achieve the above objectives, according to one aspect of this application, a method for analyzing business risks is provided. The method includes: acquiring target multi-source data related to a target financial business from N financial data sources, where N is a positive integer; inputting the target multi-source data into a preset target model, and outputting data analysis results in natural language form, wherein the preset target model is a large model pre-adjusted according to the financial field, used to analyze the risk data correlation of the target multi-source data, and expressing the risk data correlation in natural language form; and filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report for the target financial business.

[0008] Furthermore, the target multi-source data includes structured data and unstructured data. Before inputting the target multi-source data into the preset target model, the method further includes: performing a first preprocessing on the structured data, wherein the first preprocessing includes: data cleaning and data structure standardization; and performing a second preprocessing on the unstructured data, wherein the second preprocessing includes: natural language processing and entity recognition.

[0009] Furthermore, the financial data sources include: an internal database of the financial system, a data access interface within the permissions of the financial system, and an external data platform pre-authorized to the financial system. The steps of obtaining target multi-source data related to the target financial business from N financial data sources include: extracting first relevant data related to the target financial business from the internal database; accessing the data access interface within the permissions of the financial system to request second relevant data related to the target financial business within the latest update cycle; receiving a data push message pre-authorized to the external data platform of the financial system to obtain third relevant data related to the target financial business; and integrating the first relevant data, the second relevant data, and the third relevant data to obtain the target multi-source data related to the target financial business.

[0010] Furthermore, the preset target model is constructed through the following steps: obtaining a natural language processing framework as the base model and obtaining a financial domain corpus provided in advance by the financial system; selecting a model optimization algorithm and a dynamic learning rate adjustment strategy to obtain a model iteration adjustment strategy; training the base model based on the financial domain corpus and the model iteration adjustment strategy, and adjusting the model parameters during the iteration process; stopping iterative training when the accuracy of the base model in identifying the risk correlation of historical business data in the financial domain corpus exceeds a preset threshold, thereby obtaining the target model.

[0011] Further, the step of inputting the target multi-source data into a preset target model and outputting data analysis results in natural language includes: analyzing the risk correlation patterns of structured data in the target multi-source data using a multimodal data fusion algorithm preset in the target model, and decoding the potential risk characteristics of unstructured data in the target multi-source data to obtain risk characteristic analysis results; and applying the natural language generation technology preset in the target model to convert the risk characteristic analysis results into text descriptions to obtain data analysis results in natural language.

[0012] Further, the step of filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report for the target financial business includes: for structured data in the data analysis results, filling the structured data into a first area to be filled in the preset report template; for unstructured data in the data analysis results, formatting the unstructured data according to the text display requirements of the preset report template, and filling the formatted data into a second area to be filled; after all the content in the data analysis results is integrated, the business risk analysis report for the target financial business is obtained.

[0013] Furthermore, after obtaining the business risk analysis report for the target financial business, the method further includes: sending the business risk analysis report to an interactive interface; collecting user feedback on the business risk analysis report through the interactive interface; and updating the business risk analysis report based on the feedback information.

[0014] To achieve the above objectives, according to another aspect of this application, a business risk analysis apparatus is also provided. The apparatus includes: an acquisition unit for acquiring target multi-source data related to a target financial business from N financial data sources, where N is a positive integer; an analysis unit for inputting the target multi-source data into a preset target model and outputting data analysis results in natural language form, wherein the preset target model is a large model pre-adjusted according to the financial field, used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form; and a filling unit for filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report for the target financial business.

[0015] Furthermore, the target multi-source data includes structured data and unstructured data, and the business risk analysis device further includes: a first preprocessing module, used to perform a first preprocessing on the structured data before inputting the target multi-source data into a preset target model, wherein the first preprocessing includes: data cleaning processing and data structure standardization processing; and a second preprocessing module, used to perform a second preprocessing on the unstructured data, wherein the second preprocessing includes: natural language processing and entity recognition processing.

[0016] Furthermore, the financial data source includes: an internal database of the financial system, a data access interface within the permissions of the financial system, and an external data platform pre-authorized to the financial system. The acquisition unit includes: an extraction module, used to extract first relevant data related to the target financial business from the internal database; an access module, used to access the data access interface within the permissions of the financial system and request second relevant data related to the target financial business within the latest update cycle; a receiving module, used to receive data push messages pre-authorized to the external data platform of the financial system and obtain third relevant data related to the target financial business; and a first integration module, used to integrate the first relevant data, the second relevant data, and the third relevant data to obtain target multi-source data involved in the target financial business.

[0017] Furthermore, the business risk analysis device further includes: an acquisition module for acquiring a natural language processing framework as the base model and a financial domain corpus pre-provided by the financial system; a selection module for selecting a model optimization algorithm and a dynamic learning rate adjustment strategy to obtain a model iteration adjustment strategy; a training module for training the base model based on the financial domain corpus and the model iteration adjustment strategy, and adjusting the model parameters during the iteration process; and a stopping module for stopping iterative training until the accuracy of the base model in identifying the risk correlation of historical business data in the financial domain corpus exceeds a preset threshold, thereby obtaining the target model.

[0018] Furthermore, the analysis unit includes: an analysis module, used to analyze the risk correlation patterns of structured data in the target multi-source data using a multimodal data fusion algorithm pre-set in the target model, and to decode the potential risk characteristics of unstructured data in the target multi-source data to obtain risk characteristic analysis results; and a conversion module, used to apply the natural language generation technology pre-set in the target model to convert the risk characteristic analysis results into text descriptions to obtain data analysis results in natural language form.

[0019] Further, the filling unit includes: a first filling module, used to fill the structured data in the data analysis results into a first area to be filled in the preset report template; a second filling module, used to format the unstructured data in the data analysis results according to the text display requirements of the preset report template, and fill the formatted data into a second area to be filled; and a second integration module, used to obtain the business risk analysis report of the target financial business after all the content in the data analysis results has been integrated.

[0020] Furthermore, the business risk analysis device further includes: a sending module, used to send the business risk analysis report to the interactive interface after obtaining the business risk analysis report of the target financial business; a collection module, used to collect user feedback information on the business risk analysis report through the interactive interface; and an updating module, used to update the business risk analysis report according to the feedback information.

[0021] To achieve the above objectives, according to another aspect of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the business risk analysis method described in any one of the above claims.

[0022] To achieve the above objectives, according to another aspect of this application, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the business risk analysis method described in any of the above claims.

[0023] To achieve the above objectives, according to another aspect of this application, a computer program product is also provided, including computer instructions, wherein when the computer instructions are executed by a processor, they implement the steps of the business risk analysis method described in any one of the above claims.

[0024] This invention proposes a business risk analysis method. First, target multi-source data related to the target financial business are obtained from N financial data sources, where N is a positive integer. Then, the target multi-source data is input into a preset target model, and the data analysis results are output in natural language form. The preset target model is a large model that has been pre-adjusted according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form. Finally, the data analysis results in natural language form are used to fill a preset report template to obtain a business risk analysis report of the target financial business.

[0025] This invention employs a multi-source data fusion approach and deep learning-driven large-scale model analysis to achieve intelligent and comprehensive risk assessment. This enables the automated generation of in-depth insight reports, thus solving the technical problem of insufficient data integration and analysis in related technologies, which leads to low accuracy in business risk reports. Specifically, this invention breaks through the limitations of traditional single data sources, integrating key information from diverse financial data channels. It utilizes a pre-trained and finely tuned deep language model for the financial field to deeply mine the inherent relationships between different data types. Through the natural language processing capabilities of the large model, complex analysis results are transformed into easily understandable and disseminated natural language descriptions, greatly improving the readability and practicality of the reports. This overcomes the bottlenecks of existing technologies, such as limited data processing capabilities, narrow analytical dimensions, and heavy reliance on manual intervention. It not only effectively improves the efficiency of business risk report generation from hours to minutes but also significantly enhances the depth and breadth of the reports, enabling the identification and description of hidden risk factors that are difficult for traditional rule engines to capture. Attached Figure Description

[0026] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0027] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a business risk analysis method is shown.

[0028] Figure 2 This is a flowchart of an optional business risk analysis method according to an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of an optional business risk analysis device according to an embodiment of the present invention;

[0030] Figure 4 This is a structural block diagram of an electronic device for analyzing business risks according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] To facilitate understanding of the present invention by those skilled in the art, some terms or nouns involved in the various embodiments of the present invention are explained below:

[0034] CRM, or Customer Relationship Management, is a system used to manage customer information, track sales leads, record customer interactions, and provide customer service. It helps companies analyze customer behavior, optimize marketing strategies, and improve sales efficiency. In this invention, the CRM system serves as one of the data sources, providing basic customer information and transaction records related to the target financial business for risk analysis and report generation.

[0035] NLP, or Natural Language Processing, is a technology used to parse, understand, and generate human language, playing a crucial role in applications such as search engines, speech recognition, sentiment analysis, and machine translation. In this invention, it is used to process unstructured data.

[0036] NER, or Named Entity Recognition, is a subfield of Natural Language Processing (NLP) used to automatically identify entity names (such as person names, place names, and organization names) in text. It is widely applied in information extraction, knowledge graph construction, and question-answering systems. This invention is used in unstructured data processing to identify key entities (such as specific customers or industries) in text, providing clues for risk association analysis.

[0037] AdamW, or Adaptive Moment Estimation with Weight Decay, is a deep learning optimization algorithm used for neural network training. In this invention, the AdamW optimization algorithm is applied to the training of large models. By dynamically adjusting the learning rate and weight decay, it ensures the model's training performance on a financial corpus, avoids overfitting, and improves the model's generalization ability.

[0038] Cosine annealing, or cosine annealing learning rate strategy, is a learning rate adjustment strategy that simulates the annealing process in physics by gradually reducing the learning rate to help the model converge to the global optimum. It is commonly used in the later stages of deep learning training. This invention employs a cosine annealing strategy to dynamically adjust the learning rate of the AdamW optimizer, ensuring the stability and effectiveness of model training, especially contributing to improved model performance during the fine-tuning phase.

[0039] Principal Component Analysis (PCA) is a statistical method used to identify the main directions of change in data, reducing data dimensionality while preserving most of the information. It is widely used in data visualization, feature reduction, and other fields. In this invention, it is used as part of feature engineering to extract key features from structured data, reducing data dimensionality and thus improving model training efficiency and analytical accuracy.

[0040] Factor analysis (FA) is a statistical technique used to identify a set of underlying factors behind multiple variables to simplify complex datasets. In this invention, factor analysis is used in structured data preprocessing to identify key factors influencing business risk and to help the model understand the intrinsic relationships between data.

[0041] It should be noted that the business risk analysis method and apparatus in this application can be used in the field of artificial intelligence technology for intelligent risk assessment and report generation of financial business data, and can also be used in any field other than artificial intelligence technology for intelligent risk assessment and report generation of financial business data. This application does not limit the application field of the business risk analysis method and apparatus.

[0042] It should be noted that all relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this application are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, processing, transmission, provision, disclosure, use, and handling of such data comply with the laws, regulations, and standards of the relevant regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse access. For example, this system has interfaces with relevant users or organizations. Before obtaining relevant information, a request to obtain the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained only after receiving consent from the aforementioned user or organization.

[0043] The information collection (e.g., user voice, video, and text collection) and analysis operations involved in this application have provided users with corresponding operation entry points during execution, allowing users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0044] The following embodiments of the present invention can be applied to various systems / applications / devices that require financial business risk monitoring and intelligent report generation, enabling a solution for dynamic risk assessment and automatic natural language report generation based on a large model. This invention uses a deep learning large model to perform intelligent correlation analysis and learning on multi-source heterogeneous data, then generates risk insights and suggestions in natural language form. This better captures complex correlations between data and dynamic market trends, while simultaneously verifying the rationality of potential risk signals using domain knowledge graphs.

[0045] This invention also utilizes a dynamic template engine to precisely adapt to different business scenarios and data types, intelligently adjusting report templates. The testing process is automated and easily customizable, generating reports quickly and with rich, accurate content. Furthermore, a feedback optimization mechanism collects user suggestions for revising automatically generated reports, continuously fine-tuning model parameters to ensure the accuracy and timeliness of analysis results, further improving the efficiency and quality of business decision-making.

[0046] The present invention will now be described in detail with reference to various embodiments.

[0047] Example 1

[0048] According to an embodiment of the present invention, an embodiment of a business risk analysis method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0049] The business risk analysis method provided in Embodiment 1 of the present invention can be executed on a mobile terminal, computer terminal or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a business risk analysis method is shown. Figure 1As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0050] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0051] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the business risk analysis method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned business risk analysis method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0052] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0053] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0054] Under the above operating environment, the present invention provides, as follows: Figure 2 The business risk analysis method shown can be implemented by a large-scale inclusive finance post-loan risk report generation assistant system, which combines deep learning and natural language processing technologies for post-loan risk management scenarios of financial institutions, especially for dynamic risk trend analysis and intelligent report generation. Through multimodal data fusion and prompt-driven natural language generation, the method involves data collection and preprocessing, large-scale intelligent analysis, report generation and output, and feedback optimization, to automatically generate comprehensive, accurate and insightful post-loan risk reports to support efficient business decision-making.

[0055] The embodiments of the present invention will now be described in detail with reference to each specific step.

[0056] Figure 2 This is a flowchart of an optional business risk analysis method according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes the following steps:

[0057] Step S201: Obtain target multi-source data related to the target financial business from N financial data sources, where N is a positive integer.

[0058] In this embodiment, N financial data sources refer to multiple sources of financial information from different channels. These sources may include internal databases (such as CRM systems and transaction history records), external data services (such as data provided by credit rating agencies, macroeconomic indicators, industry reports, social media sentiment, etc.), and data feedback from real-time transaction systems. It should be noted that a single data source often cannot capture all possible risk signals. Using multiple financial data sources ensures comprehensive risk assessment through data diversity. For example, customer repayment history recorded by an internal CRM system provides micro-level information, while macroeconomic data can reveal the potential impact of the macroeconomic environment on credit risk.

[0059] In this embodiment of the invention, the target financial business refers to specific business types conducted by financial institutions such as banks, insurance companies, and securities firms, including personal loans, corporate credit, credit card services, and investment and wealth management products. Different financial businesses may involve different types of data sources and analytical focuses. For example, personal loan business may focus more on an individual's credit history and income stability, while corporate credit may focus on the company's financial condition, industry position, and market dynamics.

[0060] Targeted multi-source data refers to financial data obtained from N financial data sources for a specific financial business, including structured data (such as numbers, dates, and fixed-format spreadsheet data) and unstructured data (such as text, PDF reports, audio, and video). It's important to note that structured data facilitates quantitative analysis, such as overdue days and repayment frequency; while unstructured data provides richer contextual understanding, such as verbal intent in customer communication records and changes in business rules in industry news. Analyzing and correlating this unstructured data helps identify potential risks.

[0061] Optionally, in the business risk analysis method provided in this embodiment of the invention, the financial data source includes: an internal database of the financial system, a data call interface within the permissions of the financial system, and an external data platform pre-authorized to the financial system. The step of obtaining target multi-source data related to the target financial business from N financial data sources includes: extracting first relevant data related to the target financial business from the internal database; accessing the data call interface within the permissions of the financial system to request second relevant data related to the target financial business in the latest update cycle; receiving a data push message from the external data platform pre-authorized to the financial system to obtain third relevant data related to the target financial business; and integrating the first relevant data, the second relevant data, and the third relevant data to obtain target multi-source data related to the target financial business.

[0062] It's important to note that internal databases are the storage locations for financial institutions' own data, containing structured information such as customer information, transaction history, account status, and overdue records. In practice, the primary relevant data may include, but is not limited to, customer credit scores, repayment history, account balance trends, and existing loan and credit card records. The purpose of this step is to mine the historical data accumulated within the financial institution to provide a foundation for risk analysis.

[0063] Furthermore, the data access interface within the financial system's permissions allows for the retrieval of data from external systems that are updated in real time, such as interest rate changes stipulated by upstream institutions, the latest credit ratings from third-party credit rating agencies, and shared data from other financial institutions. The second set of relevant data emphasizes real-time data and an external perspective, such as changes in market interest rates, industry development trends, and updates to financial regulations, used to reflect the impact of the current market environment and external conditions on financial business risks.

[0064] Furthermore, pre-authorization to external data platforms within the financial system signifies prior access rights granted to the financial system. These platforms, such as social media, news websites, and academic research databases, can provide unstructured information about customer behavior, market sentiment, and economic research data. Thirdly, relevant data can encompass customer social media activity, discussions on public forums, macroeconomic research reports, and industry analyst commentary. This data helps assess risk from a broader perspective, evaluating the risk impact of indirect financial indicators.

[0065] The steps outlined above construct a multi-source data-driven intelligent risk assessment framework. By collecting primary, secondary, and tertiary relevant data from internal databases, real-time data interfaces, and authorized external data platforms, and utilizing a large-scale model fine-tuned for the financial sector for comprehensive analysis, a comprehensive and insightful business risk analysis report is generated. This not only improves the speed and efficiency of analysis but also enhances the accuracy and depth of the report, providing financial institutions with a powerful risk management tool. By integrating information from different data sources, financial institutions can better monitor market dynamics, customer behavior, and internal conditions, thereby anticipating and responding to potential risks in advance.

[0066] Step S202: Input the target multi-source data into the preset target model and output the data analysis results in natural language form. The preset target model is a large model that has been adjusted in advance according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form.

[0067] It should be noted that the preset target model is a deep learning model specifically fine-tuned for the financial field. It undergoes secondary training based on a general-purpose model to enhance its performance in financial data understanding and risk assessment. The fine-tuning process may involve using financial-related corpora (such as regulatory documents, financial reports, industry news, etc.) to ensure the model can accurately understand financial terminology and concepts, as well as complex data association patterns. For example, the model needs to be able to identify potential credit risk factors from multi-source data such as customer repayment records, industry dynamics, and economic indicators.

[0068] Natural language processing (NLP) data analysis results refer to presenting the data analysis process and conclusions in human-readable text, including but not limited to trend predictions, anomaly detection, risk assessments, and corresponding management recommendations. The purpose of NLP generation is to enable business personnel without a technical background to understand complex analytical results. For example, a generated report might include phrases like, "Recent rule adjustments in a certain industry have been observed to potentially impact loan repayment capacity in specific regions; it is recommended to increase the frequency of loan monitoring in these regions." Such descriptions are not only intuitive but also provide concrete action recommendations.

[0069] Risk data correlation analysis, targeting multi-source data, refers to discovering the mutual influence and potential correlations between different datasets across multiple sources, particularly in identifying potential business risks. It is accomplished by examining the relationships between structured data (such as delinquency rates and market interest rates) and unstructured data (such as customer communication texts and social media sentiment). For example, the model can discover a correlation between certain keywords in customer feedback (such as "job instability" and "declining income") and rising loan default rates, thus providing financial institutions with early warning signals to adjust their business execution recommendations and mitigate business risks for their clients.

[0070] Optionally, in the business risk analysis method provided in this embodiment of the invention, the target multi-source data includes structured data and unstructured data. Before inputting the target multi-source data into a preset target model, the method further includes: performing a first preprocessing on the structured data, wherein the first preprocessing includes: data cleaning processing and data structure standardization processing; and performing a second preprocessing on the unstructured data, wherein the second preprocessing includes: natural language processing and entity recognition processing.

[0071] It's important to note that the goal of the first preprocessing step is to structure the data. This involves eliminating noise, inconsistencies, and missing values ​​to ensure the data quality and structure meet the model's input requirements. Specific operations include, but are not limited to, removing duplicate records, filling in or deleting missing values, converting data formats (e.g., standardizing dates), and data standardization (e.g., normalizing numerical values ​​to the same range). In practice, data science tools (such as Python's Pandas library and SQL queries) can be used for data cleaning and preprocessing to ensure that each record is clean and uniformly formatted, facilitating model understanding and analysis.

[0072] The second preprocessing step targets unstructured data, transforming unstructured data such as text, images, and audio into a computer-understandable format to extract key information. Specific operations include using Natural Language Processing (NLP) techniques for text cleaning (removing stop words and punctuation), word segmentation, and stemming / lexical reconstruction; and using Natural Entity Recognition (NER) techniques to identify entities in the text (such as names of people, places, organizations, dates, etc.). In practice, deep learning frameworks and NLP libraries can be used to transform unstructured data into vectors or other structured forms so that large models can perform effective analysis.

[0073] Preprocessed data is cleaner and more structured. For structured data, every piece of information is standardized and can be directly input into the model. Unstructured data is converted into vectors or labels after NLP and entity recognition processing, making it easier for the model to understand and analyze.

[0074] Optionally, in the business risk analysis method provided in this embodiment of the invention, the preset target model is constructed through the following steps: obtaining a natural language processing framework as the basic model and obtaining a financial domain corpus provided in advance by the financial system; selecting a model optimization algorithm and a dynamic learning rate adjustment strategy to obtain a model iteration adjustment strategy; training the basic model based on the financial domain corpus and the model iteration adjustment strategy, and adjusting the model parameters during the iteration process; stopping iterative training and obtaining the target model when the basic model's accuracy in identifying the risk correlation of historical business data in the financial domain corpus exceeds a preset threshold.

[0075] The base model in this embodiment of the invention can be a pre-trained Natural Language Processing (NLP) framework. This model has been pre-trained on a large amount of general text data and possesses adequate text understanding and generation capabilities. The financial corpus contains a large amount of financial-related text data, used to fine-tune the base model, making it more adaptable to the data characteristics and risk analysis needs of the financial field. The financial corpus can include documents from the State Financial Regulatory Commission, corporate financial reports, anonymized credit records, industry reports, financial news, new business regulations, etc. By training on the financial corpus, the target model can learn knowledge in areas such as financial terminology, business processes, and market trends, thereby more accurately identifying risk correlations when processing specific business risk data.

[0076] Furthermore, model optimization algorithms are used to update model parameters to minimize the loss function. For example, AdamW is a widely used optimization algorithm that effectively handles sparse gradients and noisy data. The learning rate determines the step size for updating model parameters. Dynamic adjustment strategies (such as cosine annealing) can gradually reduce the learning rate during training, helping the model converge to the optimal solution more stably.

[0077] Furthermore, the base model is trained on a financial corpus to improve its accuracy in identifying risk correlations, meaning the model can accurately identify which factors are related to risk from historical business data. Iterative adjustments to model parameters include the weight matrix, bias term, and learning rate. By adjusting these parameters, the model can better fit the data and improve its predictive and analytical capabilities. Performance-based adjustments can be used, such as reducing the learning rate, increasing regularization terms, or adjusting the network architecture when the model's validation set performance reaches a bottleneck.

[0078] Another point to note is that the preset threshold is the standard for stopping model training. For example, training stops when the recognition accuracy exceeds 85%. The threshold setting is based on a balance between business needs, model performance, and training resources. Typically, the threshold setting will refer to the model's performance on the validation set to ensure that the model can provide sufficient risk insights in practical applications, while avoiding the waste of computational resources and the risk of overfitting caused by overtraining.

[0079] Optionally, in the business risk analysis method provided in this embodiment of the invention, the step of inputting target multi-source data into a preset target model and outputting data analysis results in natural language form includes: analyzing the risk correlation patterns of structured data in the target multi-source data using a multimodal data fusion algorithm preset in the target model, and decoding the potential risk characteristics of unstructured data in the target multi-source data to obtain risk characteristic analysis results; and applying the natural language generation technology preset in the target model to convert the risk characteristic analysis results into text descriptions to obtain data analysis results in natural language form.

[0080] It's important to note that multimodal data fusion algorithms are a machine learning technique designed to integrate and analyze data from different modalities (such as text, images, and sound) to uncover deeper and more comprehensive understandings or predictions. In this specific application scenario, it involves the fusion of structured and unstructured data to improve the accuracy and comprehensiveness of risk feature analysis. Structured data refers to data that can be organized according to fixed formats and rules, such as tabular customer information, transaction records, and financial statements; unstructured data includes text, emails, social media posts, and contract content. This type of data lacks a fixed format and requires special processing before it can be used for analysis.

[0081] In a specific embodiment for analyzing risk correlation patterns in structured data, the structured data is first cleaned and standardized to ensure consistency and accuracy, including steps such as filling missing values, removing outliers, and data type conversion. Then, statistical methods (such as PCA and factor analysis) and machine learning techniques (such as neural networks) are used to extract risk-related features, for example, extracting features such as the number of overdue payments and the number of overdue days from repayment records. Finally, the extracted structured features are combined with features from unstructured data using a multimodal data fusion algorithm, which can employ techniques such as attention mechanisms and joint encoders to capture potential connections between different data sources.

[0082] In a specific embodiment of decoding the potential risk characteristics of unstructured data, firstly, NLP techniques (such as word embedding, syntactic analysis, and sentiment analysis) can be used to preprocess the text data, transforming it into a vector representation that machines can understand. Then, entity recognition (NER) can be used to identify key entities such as people, places, and institutions in the text, and relation extraction can be used to identify the dynamic relationships between these entities, such as the connection between a customer and the economic situation of a certain region. Finally, special attention is paid to risk-related terms in the financial field, such as "default," "overdue," and "bankruptcy," and NLP techniques are used to identify and label the occurrence of these terms in the text, providing clues for risk analysis.

[0083] After completing the above two steps, the risk characteristics of both structured and unstructured data are combined to predict potential risk levels or trends using a model. For example, "We have observed an increase in talk of economic downturn in a certain region, and customer loan delinquency rates are also rising, indicating potential regional credit risk."

[0084] The specific operation of multimodal data fusion algorithms includes the following steps: 1. Modality matching: Identifying common themes or entities between structured and unstructured data, such as matching customer IDs in text with IDs in a structured database to achieve data correlation. 2. Feature encoding: Encoding the features of structured data and NLP-processed unstructured data into a unified vector space representation, facilitating model analysis within a unified framework. 3. Joint modeling: In an ensemble model, encoded structured and unstructured data are input simultaneously, and algorithms such as attention mechanisms are used to discover the mutual influence between different modalities. 4. Risk prediction: Based on the fused features, the model predicts the risk level or trend of a specific business, outputting risk feature analysis results.

[0085] Step S203: Use the data analysis results in natural language to populate the preset report template to obtain the business risk analysis report of the target financial business.

[0086] A pre-prepared report template is a standardized document framework designed to organize and present the risk analysis results of a specific financial transaction. The template may contain multiple chapters or sections, such as "Business Overview," "Risk Point Overview," "Detailed Risk Analysis," and "Management Recommendations." It provides a structured layout and format, ensuring the generated report is both professional and easy to understand. The template design must consider the regulatory requirements and internal standards of the financial industry to ensure the report's standardization and consistency. For example, the report must include quantitative indicators for risk assessment, as well as qualitative analysis based on these indicators and recommended action plans.

[0087] The business risk analysis report is the final output document, integrating the analysis of structured data and the interpretation of unstructured data, presenting the overall situation and in-depth insights into business risks in natural language. A business risk analysis report is more than just a list of data; more importantly, it provides explanations of risk factors and predictions of potential consequences, helping decision-makers understand the nature of the risk and its potential impact. For example, the report might state: "In the past quarter, due to the sharp rise in raw material prices, the loan default rate of small and medium-sized manufacturing enterprises increased by 10%." Such a description not only conveys a risk signal but also identifies the root cause of the risk, providing a basis for subsequent risk management measures.

[0088] Optionally, in the business risk analysis method provided in this embodiment of the invention, the step of filling a preset report template with data analysis results in natural language form to obtain a business risk analysis report for the target financial business includes: for structured data in the data analysis results, filling the structured data into a first area to be filled in the preset report template; for unstructured data in the data analysis results, formatting the unstructured data according to the text display requirements of the preset report template, and filling the formatted data into a second area to be filled; after all the content in the data analysis results has been integrated, a business risk analysis report for the target financial business is obtained.

[0089] It's important to note that structured data refers to quantitative data, such as loan delinquency rates, credit scores, and market interest rates, which are directly filled into fixed locations in the report template, such as the "Repayment Overview" and "Risk Indicators" areas. Unstructured data mainly includes qualitative information, such as customer complaints in emails, negative comments on social media, and industry news reports. This data needs to be processed using NLP (Natural Language Processing) to format key information summaries or analytical conclusions before being filled into sections such as "Market Dynamics" and "Customer Feedback Analysis" in the report.

[0090] The above steps in this embodiment of the invention achieve a high degree of automation and intelligence in business risk analysis through multimodal data fusion and natural language generation technology, filling the gap in existing technologies when processing complex, multi-source data, and greatly improving the risk management efficiency and decision-making accuracy of financial institutions.

[0091] Optionally, in the business risk analysis method provided in this embodiment of the invention, after obtaining the business risk analysis report of the target financial business, the method further includes: sending the business risk analysis report to an interactive interface; collecting user feedback information on the business risk analysis report through the interactive interface; and updating the business risk analysis report based on the feedback information.

[0092] In the business risk analysis method of this invention, the generated business risk analysis report is not only a one-time output but also includes continuous optimization steps to ensure the report's accuracy and usability. After the report is generated, it can be sent in a readable format (such as PDF or HTML) to an interactive interface designed for displaying and editing the report, ensuring that the report can be directly accessed by target users (such as risk managers and senior decision-makers). Users can receive the business risk analysis report generated from the large model through an intuitive interface, which includes the analysis results of structured and unstructured data, as well as intelligent suggestions based on these results.

[0093] The interactive interface is capable of collecting user feedback, including suggestions for revising report content, questioning or supplementing specific analysis results, and providing feedback on report format improvements. Users can annotate, mark, or suggest modifications to any part of the report through the interface; this feedback is recorded and processed by the implementation system. The implementation system can automatically or semi-automatically update report content based on user feedback, involving adjustments to model parameters, reassessment of data sources, or optimization of report templates. Based on collected user feedback, the implementation system automatically adjusts certain analytical conclusions; for example, it may reassess a financial institution's risk level based on specific unstructured data (such as the latest market news) indicated by users and update this information in the report.

[0094] Through the above steps S201 to S204, target multi-source data related to the target financial business can be obtained from N financial data sources, where N is a positive integer. Then, the target multi-source data is input into a preset target model, and the data analysis results in natural language form are output. The preset target model is a large model that has been adjusted in advance according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form. Finally, the data analysis results in natural language form are used to fill the preset report template to obtain the business risk analysis report of the target financial business.

[0095] In this embodiment of the invention, a multi-source data fusion approach is adopted, employing deep learning-driven large-scale model analysis to achieve intelligent and comprehensive risk assessment. This enables the automated generation of in-depth insight reports, thereby solving the technical problem of insufficient data integration and analysis in related technologies, which leads to low accuracy in business risk reports. Specifically, this invention breaks through the limitations of traditional single data sources, integrating key information from diversified financial data channels. It utilizes a pre-trained and finely tuned deep language model for the financial field to deeply mine the inherent relationships between different data types. Through the natural language processing capabilities of the large model, complex analysis results are transformed into easily understandable and disseminated natural language descriptions, greatly improving the readability and practicality of the report. This overcomes the bottlenecks of existing technologies, such as single data processing capabilities, narrow analysis dimensions, and heavy reliance on manual intervention. It not only effectively improves the generation efficiency of business risk reports, reducing the time from hours to minutes, but also significantly enhances the depth and breadth of the reports, enabling the identification and description of hidden risk factors that are difficult for traditional rule engines to capture.

[0096] The invention will now be described in conjunction with another alternative embodiment.

[0097] Example 2

[0098] This invention also provides a business risk analysis device. It should be noted that the business risk analysis device of this invention includes multiple implementation units, which can be used to execute the business risk analysis method provided in the first embodiment above. Each implementation unit corresponds to each implementation step in the first embodiment above.

[0099] Figure 3 This is a schematic diagram of an optional business risk analysis device according to an embodiment of the present invention, such as... Figure 3 As shown, the device may include: an acquisition unit 31, an analysis unit 32, and a filling unit 33.

[0100] Among them, the acquisition unit 31 is used to acquire target multi-source data related to the target financial business from N financial data sources, where N is a positive integer.

[0101] Analysis unit 32 is used to input target multi-source data into a preset target model and output data analysis results in natural language form. The preset target model is a large model that has been adjusted in advance according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form.

[0102] Filling unit 33 is used to fill a preset report template with data analysis results in natural language form to obtain a business risk analysis report for the target financial business.

[0103] The aforementioned business risk analysis device can first acquire target multi-source data related to the target financial business from N financial data sources through the acquisition unit 31, where N is a positive integer. Then, the analysis unit 32 inputs the target multi-source data into a preset target model and outputs data analysis results in natural language form. The preset target model is a large model that has been pre-adjusted according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and expresses the risk data correlation in natural language form. Finally, the filling unit 33 fills the preset report template with the data analysis results in natural language form to obtain the business risk analysis report of the target financial business.

[0104] In this embodiment of the invention, a multi-source data fusion approach is adopted, employing deep learning-driven large-scale model analysis to achieve intelligent and comprehensive risk assessment. This enables the automated generation of in-depth insight reports, thereby solving the technical problem of insufficient data integration and analysis in related technologies, which leads to low accuracy in business risk reports. Specifically, this invention breaks through the limitations of traditional single data sources, integrating key information from diversified financial data channels. It utilizes a pre-trained and finely tuned deep language model for the financial field to deeply mine the inherent relationships between different data types. Through the natural language processing capabilities of the large model, complex analysis results are transformed into easily understandable and disseminated natural language descriptions, greatly improving the readability and practicality of the report. This overcomes the bottlenecks of existing technologies, such as single data processing capabilities, narrow analysis dimensions, and heavy reliance on manual intervention. It not only effectively improves the generation efficiency of business risk reports, reducing the time from hours to minutes, but also significantly enhances the depth and breadth of the reports, enabling the identification and description of hidden risk factors that are difficult for traditional rule engines to capture.

[0105] Furthermore, the target multi-source data includes structured data and unstructured data. The business risk analysis device also includes: a first preprocessing module, used to perform a first preprocessing on the structured data before inputting the target multi-source data into the preset target model, wherein the first preprocessing includes: data cleaning processing and data structure standardization processing; and a second preprocessing module, used to perform a second preprocessing on the unstructured data, wherein the second preprocessing includes: natural language processing and entity recognition processing.

[0106] Furthermore, the financial data sources include: the internal database of the financial system, data access interfaces within the permissions of the financial system, and external data platforms pre-authorized to the financial system. The acquisition unit includes: an extraction module, used to extract first relevant data related to the target financial business from the internal database; an access module, used to access the data access interfaces within the permissions of the financial system and request second relevant data related to the target financial business within the latest update cycle; a receiving module, used to receive data push messages from the external data platform pre-authorized to the financial system and obtain third relevant data related to the target financial business; and a first integration module, used to integrate the first, second, and third relevant data to obtain the target multi-source data involved in the target financial business.

[0107] Furthermore, the business risk analysis device also includes: an acquisition module for acquiring the natural language processing framework used as the base model and a financial domain corpus pre-provided by the financial system; a selection module for selecting the model optimization algorithm and dynamic learning rate adjustment strategy to obtain the model iteration adjustment strategy; a training module for training the base model based on the financial domain corpus and the model iteration adjustment strategy, and adjusting the model parameters during the iteration process; and a stopping module for stopping iterative training and obtaining the target model when the accuracy of the base model in identifying the risk correlation of historical business data in the financial domain corpus exceeds a preset threshold.

[0108] Furthermore, the analysis unit includes: an analysis module, used to analyze the risk correlation patterns of structured data in the target multi-source data through a multi-modal data fusion algorithm pre-built in the target model, and to decode the potential risk characteristics of unstructured data in the target multi-source data to obtain risk characteristic analysis results; and a conversion module, used to apply the natural language generation technology pre-built in the target model to convert the risk characteristic analysis results into text descriptions to obtain data analysis results in natural language form.

[0109] Furthermore, the filling unit includes: a first filling module, used to fill the structured data in the data analysis results into the first area to be filled in the preset report template; a second filling module, used to format the unstructured data in the data analysis results according to the text display requirements of the preset report template, and fill the formatted data into the second area to be filled; and a second integration module, used to obtain a business risk analysis report of the target financial business after all the content in the data analysis results has been integrated.

[0110] Furthermore, the business risk analysis device also includes: a sending module, used to send the business risk analysis report to the interactive interface after receiving the business risk analysis report of the target financial business; a collection module, used to collect user feedback information on the business risk analysis report through the interactive interface; and an updating module, used to update the business risk analysis report based on the feedback information.

[0111] It should be noted that the acquisition unit 31, analysis unit 32, and filling unit 33 mentioned above correspond to steps S201 to S204 in Embodiment 1. The instances and application scenarios implemented by the above units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules or units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0112] The invention will now be described in conjunction with another alternative embodiment.

[0113] Example 3

[0114] The present invention can also provide an electronic device. Figure 4 This is a structural block diagram of an electronic device for analyzing business risks according to an embodiment of the present invention, such as... Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) processor 702, memory 704, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0115] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the business risk analysis method and apparatus in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned business risk analysis method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0116] The processor can access information and applications stored in memory via a transmission device to perform the following steps: acquiring target multi-source data related to the target financial business from N financial data sources, where N is a positive integer; inputting the target multi-source data into a preset target model and outputting data analysis results in natural language form, where the preset target model is a large model pre-adjusted according to the financial field, used to analyze the risk data correlation of the target multi-source data and expressing the risk data correlation in natural language form; filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report for the target financial business.

[0117] The processor can also invoke information and applications stored in the memory via a transmission device to perform the following steps: performing a first preprocessing on structured data, wherein the first preprocessing includes: data cleaning and data structure standardization; and performing a second preprocessing on unstructured data, wherein the second preprocessing includes: natural language processing and entity recognition.

[0118] The processor can also invoke information and applications stored in the memory via the transmission device to perform the following steps: extracting first relevant data related to the target financial business from the internal database; accessing the data call interface within the financial system's permissions to request second relevant data related to the target financial business within the latest update cycle; receiving data push messages pre-authorized to the external data platform of the financial system to obtain third relevant data related to the target financial business; and integrating the first, second, and third relevant data to obtain target multi-source data involved in the target financial business.

[0119] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: acquire the natural language processing framework as the base model and acquire the financial domain corpus provided in advance by the financial system; select the model optimization algorithm and the dynamic learning rate adjustment strategy to obtain the model iterative adjustment strategy; train the base model based on the financial domain corpus and the model iterative adjustment strategy, and adjust the model parameters during the iteration process; stop iterative training and obtain the target model when the accuracy of the base model in identifying the risk correlation of historical business data in the financial domain corpus exceeds a preset threshold.

[0120] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: using the multimodal data fusion algorithm pre-set in the target model to analyze the risk correlation patterns of structured data in the target multi-source data, and decode the potential risk characteristics of unstructured data in the target multi-source data to obtain risk characteristic analysis results; applying the natural language generation technology pre-set in the target model to convert the risk characteristic analysis results into text descriptions to obtain data analysis results in natural language form.

[0121] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: for structured data in the data analysis results, fill the structured data into the first area to be filled in the preset report template; for unstructured data in the data analysis results, format the unstructured data according to the text display requirements of the preset report template, and fill the formatted data into the second area to be filled; after all the content in the data analysis results has been integrated, a business risk analysis report of the target financial business is obtained.

[0122] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: sending the business risk analysis report to the interactive interface; collecting user feedback on the business risk analysis report through the interactive interface; and updating the business risk analysis report based on the feedback.

[0123] This invention provides a business risk analysis solution. By employing multi-source data fusion and deep learning-driven large-scale model analysis, it achieves intelligent and comprehensive risk assessment, thereby enabling the automated generation of in-depth insight reports. This solves the technical problem of insufficient data integration and analysis in related technologies, leading to low accuracy in business risk reports. Specifically, this invention breaks through the limitations of traditional single data sources, integrating key information from diverse financial data channels. It utilizes a pre-trained and fine-tuned deep language model for the financial field to deeply mine the inherent relationships between different data types. Through the natural language processing capabilities of the large model, complex analysis results are transformed into easily understandable and disseminated natural language descriptions, greatly improving the readability and practicality of the report. This overcomes the bottlenecks of existing technologies, such as single data processing capabilities, narrow analytical dimensions, and heavy reliance on manual intervention. It not only effectively improves the efficiency of business risk report generation from hours to minutes but also significantly enhances the depth and breadth of the report, enabling it to identify and describe hidden risk factors that are difficult for traditional rule engines to capture.

[0124] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0125] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0126] The invention will now be described in conjunction with another alternative embodiment.

[0127] Example 4

[0128] This invention also provides a computer-readable storage medium. Optionally, in this invention, the computer-readable storage medium can be used to store the program code executed by the business risk analysis method provided in Embodiment 1.

[0129] Optionally, in this embodiment of the invention, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0130] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of a business risk analysis method: obtaining target multi-source data related to the target financial business from N financial data sources, where N is a positive integer; inputting the target multi-source data into a preset target model, and outputting data analysis results in natural language form, wherein the preset target model is a large model pre-adjusted according to the direction of the financial field, used to analyze the risk data correlation of the target multi-source data, and expressing the risk data correlation in natural language form; filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report of the target financial business.

[0131] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0132] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0134] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0135] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0136] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0137] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for analyzing business risks, characterized in that, include: Obtain target multi-source data related to the target financial business from N financial data sources, where N is a positive integer; The target multi-source data is input into a preset target model, and the data analysis results are output in natural language form. The preset target model is a large model that has been adjusted in advance according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and to describe the risk data correlation in natural language form. The data analysis results, expressed in natural language, are used to populate a preset report template to obtain a business risk analysis report for the target financial business.

2. The analytical method according to claim 1, characterized in that, The target multi-source data includes structured and unstructured data. Before inputting the target multi-source data into the preset target model, the method further includes: The structured data undergoes a first preprocessing step, which includes data cleaning and data structure standardization. The unstructured data undergoes a second preprocessing step, which includes natural language processing and entity recognition processing.

3. The analytical method according to claim 1, characterized in that, The financial data sources include: the internal database of the financial system, data access interfaces within the permissions of the financial system, and external data platforms pre-authorized to the financial system. The steps for obtaining target multi-source data related to the target financial business from N financial data sources include: Extract first relevant data related to the target financial business from the internal database; Access the data call interface within the permissions of the financial system to request second relevant data related to the target financial business in the latest update cycle; Receive a data push message pre-authorized to the external data platform of the financial system, and obtain third-party related data related to the target financial business; By integrating the first relevant data, the second relevant data, and the third relevant data, the target multi-source data involved in the target financial business is obtained.

4. The analytical method according to claim 1, characterized in that, The preset target model is constructed through the following steps: Obtain the natural language processing framework as the base model, and obtain the financial domain corpus pre-provided by the financial system; By selecting the model optimization algorithm and the dynamic learning rate adjustment strategy, the model iterative adjustment strategy is obtained; Based on the financial corpus and the model iteration adjustment strategy, the basic model is trained, and the model parameters are adjusted during the iteration process. The iterative training stops when the accuracy of the basic model in identifying the risk correlation of historical business data in the financial corpus exceeds a preset threshold, thus obtaining the target model.

5. The analytical method according to claim 1, characterized in that, The steps of inputting the target multi-source data into a preset target model and outputting data analysis results in natural language form include: By using the multimodal data fusion algorithm pre-set in the target model, the risk correlation patterns of structured data in the target multi-source data are analyzed, and the potential risk characteristics of unstructured data in the target multi-source data are decoded to obtain risk characteristic analysis results; By applying the natural language generation technology pre-built into the target model, the risk feature analysis results are converted into text descriptions to obtain the data analysis results in natural language form.

6. The analytical method according to claim 1, characterized in that, The steps for filling a preset report template with the data analysis results in natural language form to obtain a business risk analysis report for the target financial business include: For the structured data in the data analysis results, the structured data is filled into the first area to be filled in the preset report template; For the unstructured data in the data analysis results, the unstructured data is formatted according to the text display requirements of the preset report template, and the formatted data is filled into the second area to be filled. After all the data analysis results are integrated, the business risk analysis report of the target financial business is obtained.

7. The analytical method according to claim 1, characterized in that, After obtaining the business risk analysis report for the target financial business, the method further includes: Send the business risk analysis report to the interactive interface; The user feedback information regarding the business risk analysis report is collected through the interactive interface; The business risk analysis report is updated based on the feedback information.

8. A business risk analysis device, characterized in that, include: The acquisition unit is used to acquire target multi-source data related to the target financial business from N financial data sources, where N is a positive integer; The analysis unit is used to input the target multi-source data into a preset target model and output data analysis results in natural language form. The preset target model is a large model that has been pre-adjusted according to the direction of the financial field. It is used to analyze the risk data correlation of the target multi-source data and express the risk data correlation in natural language form. A fill unit is used to fill a preset report template with the data analysis results in the form of natural language to obtain a business risk analysis report of the target financial business.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the business risk analysis method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the business risk analysis method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, It includes computer instructions, wherein when executed by a processor, the computer instructions implement the steps of the business risk analysis method according to any one of claims 1 to 7.

Images