Business object analysis method, device and equipment

Abstract

The invention provides a business object analysis method, device and equipment. According to the scheme, the method comprises the steps that first information data related to a to-be-analyzed business object is acquired, and the duration from the available moment of the first information data to the current moment is not larger than a preset duration threshold value; then, a pre-trained large language model is adopted, service behavior information is extracted from the first intelligence data, the service behavior information is used for describing service behaviors occurring between service elements related to the service object, and the service elements comprise service entities or service events; furthermore, element relation data related to the business object is inquired from a structure database, the structure database comprises information determined based on historical business information extracted from the second intelligence data, and the available moment of the second intelligence data is earlier than the available moment of the first intelligence data; and then, based on the business behavior information and the element relation data, determining a prediction result of the characteristic parameters of the business object by adopting a pre-trained trend prediction model.

Description

Technical Field

[0001] The present application relates to the technical field of big data analysis, and in particular to a business object analysis method, device and equipment. Background Art

[0002] In the current era of information explosion, all walks of life are generating unprecedented amounts of data. Big data analysis can be applied to almost all industries, such as retail, manufacturing, healthcare, transportation, education, telecommunications, entertainment and media, energy and utilities, agriculture, and finance, all of which need to be analyzed based on these data to help people better understand their business environment and make more informed decisions. Whether for enterprises, organizations or individuals, how to effectively extract valuable information and insights from these massive amounts of data has become a problem that needs to be solved urgently.

[0003] The scale of data, the speed at which data is generated, and the complexity of data make it difficult for traditional data processing methods to cope with it. The diverse data includes not only structured data, but also a large amount of semi-structured and unstructured data, such as text, images, videos, and other forms of data, which increases the complexity of data processing. In addition, since the data is updated and iterated very quickly, in this context, how to ensure the effectiveness of data analysis results has also become a challenge.

[0004] In view of this, it is necessary to provide a method that can analyze business objects based on a large amount of complex and rapidly updated data. Summary of the invention

[0005] In view of this, embodiments of the present application provide a business object analysis method, apparatus, and device for improving the timeliness of business object analysis.

[0006] According to a first aspect of an embodiment of the present application, a business object analysis method is provided, including:

[0007] Acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold;

[0008] Using a pre-trained large language model, extracting business behavior information from the first intelligence data; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events;

[0009] Querying element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data;

[0010] Based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine the prediction result of the characteristic parameter of the business object.

[0011] According to a second aspect of an embodiment of the present application, a business object analysis device is provided, including:

[0012] An intelligence data acquisition module, used to acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold;

[0013] An information extraction module, configured to extract business behavior information from the first intelligence data using a pre-trained large language model; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events;

[0014] a structure data acquisition module, configured to query element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data;

[0015] The analysis module is used to determine the prediction results of the characteristic parameters of the business object based on the business behavior information and the element relationship data using a pre-trained trend prediction model.

[0016] According to a third aspect of an embodiment of the present application, a computing device is provided, comprising a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein the processor implements the steps of the business object analysis method when executing the computer instructions.

[0017] One embodiment of the present specification can achieve at least the following beneficial effects: on the one hand, by utilizing a large language model to extract structured business behavior information from current intelligence data associated with business objects, and on the other hand, by obtaining element relationship data determined based on historical intelligence data from a structural database to reflect the association relationship between business elements related to business objects, and then analyzing and predicting the characteristic parameters of the business objects, thereby enabling rapid and in-depth analysis of large amounts of intelligence data and improving the timeliness of the analysis results of business objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of this specification or related technologies, the drawings required for use in the embodiments or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 A schematic diagram of an application scenario of a business object analysis method provided in an embodiment of this specification;

[0020] Figure 2 A flowchart of a business object analysis method provided in an embodiment of this specification;

[0021] Figure 3 The embodiments of this specification provide corresponding to Figure 2 A schematic diagram of the structure of a business object analysis device;

[0022] Figure 4 The embodiments of this specification provide corresponding to Figure 2 A structural schematic diagram of a business object analysis device. DETAILED DESCRIPTION

[0023] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0024] The terms used in one or more embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present application. The singular forms of "a", "said" and "the" used in one or more embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of the present application refers to and includes any or all possible combinations of one or more associated listed items.

[0025] It should be understood that, although the terms first, second, etc. may be used to describe various information in one or more embodiments of the present application, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0026] It should be noted that the user 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, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards in the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0027] Data analysis plays a vital role in today's society. It is an indispensable part in various fields such as business decision-making, scientific research, and industrial operations. By collecting and analyzing a large amount of data, valuable information can be extracted from it to provide strong support for predicting trends and formulating strategies. For example, in the field of medical health, hospitals can use patients' historical medical records, symptom descriptions and other information, combined with dynamic unstructured data such as environmental factors that may be associated with the disease, combined with big data analysis technology, to help doctors diagnose diseases and predict possible epidemic trends. For example, in the field of risk prevention and control, financial institutions can effectively identify potential risk points and prevent fraud by analyzing customers' credit history, transaction patterns and other data, and combining customers' related operations, income and expenditure and other dynamic information.

[0028] However, with the explosive growth of information (especially unstructured data), combined with the frequent changes of this information, traditional data analysis methods face huge challenges. Some data used for analysis may come from multiple channels such as online platforms and social media. Its volume is huge and constantly updated, making it extremely difficult to process it using traditional means. This leads to some problems. For example, the analysis process takes a long time, which affects the speed of decision-making, and the analysis results often cannot reflect the latest status of the analysis object, and the timeliness of the analysis results is poor; and due to the limitation of analysis capabilities, it is often impossible to obtain comprehensive information from all parties, affecting the accuracy of the analysis results.

[0029] In view of this, in an embodiment of the present specification, a business object analysis tool that combines a large language model and a knowledge graph is proposed. This automated, data-driven business object analysis tool can not only accelerate the entire analysis process and ensure the timeliness and comprehensiveness of the analysis, but also increase the depth of the analysis and improve the accuracy of the analysis results.

[0030] First, the terms involved in one or more embodiments of the present application are explained.

[0031] Natural Language Processing (NLP): Natural language processing is a branch of computer science, artificial intelligence, and linguistics that focuses on the interaction between computers and human language. The goal is to enable computers to understand and generate human language.

[0032] Large Language Model: A large language model is a computer program trained using deep learning techniques that can understand and generate human language text. Large language models have outstanding performance in natural language processing, especially in understanding and generating text. These models are usually trained based on large-scale data sets and can capture complex patterns and contextual relationships of language. Typical examples of large language models include models such as GPT-3 based on the Transformer architecture.

[0033] Transformer architecture: Transformer architecture is a model architecture in deep learning that relies on the self-attention mechanism to process sequence data and is particularly suitable for processing large amounts of text information. Compared to previous recurrent neural network models (RNNs) and long short-term memory network models (LSTMs), the Transformer model has significant advantages in processing long-distance dependencies and parallel computing.

[0034] Knowledge Graph: Knowledge graph is a method of representing knowledge in a graphical way. It structurally represents entities in the real world and various relationships between entities. In a knowledge graph, nodes usually represent entities and edges represent relationships between entities. Knowledge graphs are often used to enhance information retrieval, recommendation systems, natural language understanding and other fields.

[0035] Data Analysis: Data analysis refers to the process of examining, cleaning, transforming and modeling data sets using logical techniques and analytical algorithms in order to discover useful information, draw conclusions and support decision making.

[0036] Structured Data: Structured data refers to data organized in a fixed format or model, usually stored in a relational database, for example, data in a table, list, or CSV file.

[0037] Unstructured Data: Unstructured data refers to data that has no predefined data model or organization. Common unstructured data includes text files, emails, social media posts, videos, images, etc.

[0038] Figure 1 A schematic diagram of an application scenario of a business object analysis method provided in an embodiment of this specification.

[0039] like Figure 1 As shown, a business object analysis system 120 for analyzing business objects is provided. In actual application, the business object analysis system 120 can collect data from the data source device 110, and then perform data analysis to obtain data analysis results, generate an evaluation report, and then send the evaluation report to the user terminal device 130.

[0040] The data source device 110 may be a server, platform, terminal, etc. for providing business intelligence data. The type and number of the data source device 110 are not specifically limited. In actual application, each data source in the data source device 110 may correspond to a variety of data providers. Taking the business object to be analyzed as a fund as an example, different data providers may include but are not limited to financial service institutions, news agencies, social media platforms, etc., and the data information obtained from different data providers may include but are not limited to public fund reports, stock exchange announcements, financial news or comments, etc.

[0041] The business object analysis system 120 may include a data collection module 121 , a data preprocessing module 122 , an information extraction module 123 , a structure database 124 , an analysis and prediction module 125 , and a report generation and distribution module 126 .

[0042] like Figure 1 As shown in , the data collection module 121 can be used to continuously collect information from the data source device 110 .

[0043] In actual application, the intelligence data collected by the data acquisition module 121 may not only include structured data, but also generally include a large amount of unstructured data. In actual application, unstructured data may include text data, audio data, image data, video data, etc. In the embodiments of this specification, in order to facilitate unified operation, various forms of unstructured data may be converted into data in text form, and then subsequently processed. In the following, the analysis process is described by taking the unstructured data of the data in this article as an example.

[0044] In the embodiments of this specification, in order to efficiently and comprehensively extract effective information from these unstructured data, an information extraction module based on a large language model can be used. The large language model can capture accurate information points through a deep understanding of the context, and has the ability to predict sentences and generate coherent explanatory text when necessary.

[0045] like Figure 1 As shown in , the data preprocessing module 122 can be used to preprocess the unstructured data collected by the data acquisition module 121. In practical applications, the data preprocessing module 122 can be set as a separate module independent of the data acquisition module 121 and the information extraction module 123, or can be included in the data acquisition module 121, or can be included in the information extraction module 123.

[0046] Furthermore, the data preprocessing module 122 can be used to perform formal preprocessing on the collected data. Specifically, the collected data can be subjected to standardization, noise removal, and redundancy removal, etc., to facilitate subsequent information extraction and analysis.

[0047] Furthermore, the data preprocessing module 122 can be used to preprocess the collected data for feature extraction. Specifically, first, the data preprocessing module 122 can be used to use natural language processing (NLP) technology to perform preliminary analysis on unstructured data. Taking unstructured text data as an example, the tokenization algorithm can be used to segment the text into meaningful words or phrases; then, the named entity recognition (NER) technology can be used to identify and mark the named entities in the text, such as company names, market terms, etc. In actual application, the output of the data preprocessing module 122 can be processed text data, which contains structured entity information, providing a clear input format for the next step of information extraction.

[0048] like Figure 1As shown in , for the unstructured data in the information collected by the data collection module 121, after the text data is pre-processed, the information extraction module 123 based on the large language model can be used to perform in-depth analysis and information extraction on the text data. In actual application, the large language model in the information extraction module 123 can be a specific neural network model (for example, a transformer-based model) obtained by pre-training, which can understand the patterns and context in the text data and capture accurate information points through a deep understanding of the context. For example, the large language model can process and parse the huge unstructured text data sets in the financial field, such as fund reports, market news, analytical comments, etc., and automatically extract key business behavior information therefrom.

[0049] In actual application, the input to the information extraction module 123 may be the standardized text data containing entity identification information provided by the data preprocessing module 122. The output of the information extraction module 123 may include a series of structured business behavior information, which may be used to be input into the structure database 124 for structured storage, and may be used to be provided to the analysis and prediction module 125 for performing analysis and prediction on the business object to be analyzed.

[0050] like Figure 1 As shown in , in the structure database 124, the association relationship information between various business elements can be stored in the form of a data graph. By storing relationship information in a structured form such as a knowledge graph, a huge relationship network is constructed and maintained, which can be convenient for quick query and use in subsequent analysis. For example, in the financial field, a relationship network composed of business elements in the financial field (such as companies, commodities, industries and other entities, market changes, technological mutations and other events) and various relationship data between them can be constructed. By providing a clear relationship structure, it is possible to provide information such as how the various business elements are connected to each other and the possible impact of these connections on business objects or market dynamics, so as to be used in analyzing business objects. In an embodiment of the present specification, the data contained in the knowledge graph can include not only operation information extracted from current intelligence data, but also operation information extracted from historical intelligence data.

[0051] In the embodiments of the present specification, the knowledge graph of the structure database 124 can be implemented based on most open source knowledge graph engines, such as Apache Jena, OpenSPG, etc. The present application does not limit the specific implementation of the knowledge graph.

[0052] like Figure 1As shown in , the analysis and prediction module 125 can obtain the current business behavior information from the information extraction module 123, and obtain the element relationship data that can reflect the historical business behavior information from the structure database 124, so as to analyze the business object. In actual application, traditional machine learning technology can be used to implement the analysis and prediction of this module.

[0053] Afterwards, if Figure 1 As shown in , the report generation and distribution module 126 based on the large language model can process the analysis and prediction results given by the analysis and prediction module 125, integrate various analysis and prediction results into a human-readable report format, and distribute the generated report to the user terminal device 130. Specifically, the report can be generated using preset templates and customized strategies to put scattered data and analysis results into a coherent and clear document.

[0054] like Figure 1 The modules in the business object analysis system 120 shown in FIG. 1 are divided according to their functions. In actual application, from the perspective of hardware, at least some of the modules may have no obvious boundaries. For example, the data preprocessing module 122 may be integrated into the data acquisition module 121, or may be integrated into the information extraction module 123. For another example, the information extraction module 123 based on the large language model and the report generation and distribution module 126 based on the large language model may be integrated into one module, which may be based on the same large language model.

[0055] like Figure 1 Each module in the business object analysis system 120 shown in FIG. 1 may correspond to different hardware devices, or multiple modules may correspond to the same hardware device. Figure 1 The devices of each module of the business object analysis system 120 shown in FIG. 1 may include but are not limited to any device, equipment, platform, device cluster, etc. with computing and processing capabilities. Figure 1 The user terminal device 130 may include but is not limited to a smart phone, a tablet computer, a laptop computer, a PDA, a personal computer, a smart home device, a vehicle-mounted device, etc.

[0056] In such Figure 1 In the application scenario shown, the business object analysis system 120 and the data source device 110, and the business object analysis system 120 and the user terminal device 130 can be connected via a local area network, a wide area network, the Internet, or other types of data networks.

[0057] like Figure 1In the scheme, a business object analysis tool combining a large language model and a knowledge graph is proposed. First, considering that large language models have outstanding performance in natural language processing, especially in understanding and generating text, in the scheme of the embodiment of this specification, the large language model is applied to the analysis of unstructured text data, which can effectively extract information from complex reports and news, and understand and analyze the meaning and connections behind this information through machine learning technology. Second, as a data structure that can express elements and their relationships, the knowledge graph provides structured contextual information for model analysis. In the scheme of the embodiment of this specification, by utilizing the knowledge graph, the extracted information can be combined with the existing knowledge system to provide a deeper level of analysis, such as the relationship between entities, the potential impact of environmental information on specific entities or industries, etc.

[0058] In short, by using the latest artificial intelligence technology, combined with the advantages of the text processing capabilities of large language models and the structured analysis capabilities of knowledge graphs, the business object analysis tool provided by the embodiments of this specification has significant advantages in processing large-scale unstructured data. Through automated data processing and deep learning technology, the efficiency and coverage of analysis can be greatly improved, while reducing the impact of human errors and subjectivity. The dynamically updated knowledge graph can reflect the changes in intelligence data in real time, ensuring the timeliness and accuracy of the analysis results. In addition, the deep learning model can also reveal complex relationships between data that are difficult to discover with traditional methods, providing deeper insights. As a result, it will be possible to provide technical support for people to make wise decisions in the context of fast-paced updates of intelligence data and complex and intensive information.

[0059] In the present application, a business object analysis method is provided. The present application also relates to a business object analysis apparatus and a computing device, which are described in detail one by one in the following embodiments.

[0060] Figure 2 A flowchart of a business object analysis method provided in an embodiment of this specification.

[0061] From the program perspective, the execution subject of the process can be a program installed on an application server or an application terminal. It can be understood that the method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities.

[0062] like Figure 2 As shown, the process may include the following steps:

[0063] Step 202: Acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold.

[0064] Intelligence data usually refers to a part of valuable data selected from a wider range of data, which can be relevant information, information or data obtained through various channels. Specifically, intelligence data may include knowledge or data with practical application value for specific purposes or problems after collection, analysis and processing. In the embodiments of this specification, intelligence data may include data that is valuable for analyzing specified business objects. In actual applications, intelligence data may include data obtained from various public or authorized channels.

[0065] The first intelligence data may include data related to the business object to be analyzed, specifically, data used to describe, evaluate, etc. the business object, its associated entities, its associated business objects, or its associated entities.

[0066] In different business fields, there are different business objects to be analyzed. For example, taking the financial field as an example, the business object may include financial investment tools or financial products, and more specifically, the business object may include stocks, bonds, funds, futures, options, foreign exchange, etc. For another example, taking the medical field as an example, the business object may include a certain disease or a certain symptom, and more specifically, the business object may include a certain type of infectious disease, etc. The types of business objects are not limited to this.

[0067] As an example, if the business object is a target fund, then the associated entities of the business object may include a financial service institution that manages the target fund, etc., the associated business objects of the business object may include investment targets such as stocks and bonds held by the fund, and the associated entities of the associated business objects may include associated companies, institutions, organizations, etc. of the investment targets held by the fund. On this basis, the first intelligence data may include, for example, fund holding reports, market news, analyst opinions, financial statements, etc., and the types of intelligence data are not limited thereto. The data source for obtaining the first intelligence data may include fund reports, announcements of stock exchanges, financial news websites, social media platforms, and other financial data services, etc., and the types of data sources for obtaining intelligence data are not limited thereto.

[0068] In actual application, in order to ensure the timeliness of the analysis results for the business object, only the intelligence data released recently can be obtained, that is, only the intelligence data whose time from the available time to the current time is not greater than the preset time threshold can be obtained. For example, the preset time threshold can be set to 24 hours, 48 ​​hours, 3 days, etc. Among them, the available time of intelligence data can be the time when the intelligence data can be obtained, which can usually refer to the release time of the intelligence data.

[0069] From another perspective, in order to ensure the timeliness of the analysis results for the business object, it is usually necessary to obtain intelligence data regularly, for example, to obtain intelligence data from each data source at fixed intervals. In practical applications, the fixed interval between two acquisitions of intelligence data can be the preset time threshold. For example, the fixed interval can be set to 24 hours, 48 ​​hours, 3 days, etc.

[0070] From another perspective, in order to ensure the timeliness of the analysis results of business objects, the timeliness of data released by multiple data sources can be evaluated, and only the intelligence data released by data sources with higher timeliness can be obtained. In this way, the timeliness of intelligence data can be ensured from the perspective of the data source, thereby providing a guarantee for the timeliness of the analysis results.

[0071] Of course, in actual applications, in order to ensure the accuracy and objectivity of the analysis results, the accuracy, objectivity, authority and other aspects of the information released by the data source can also be evaluated, and the data source for obtaining intelligence data in the embodiments of this specification can be determined based on the comprehensive evaluation results.

[0072] In practical applications, the first intelligence data may include structured data or unstructured data. In an embodiment of the present specification, in order to efficiently and accurately extract effective information from a large amount of unstructured data, for the unstructured data in the intelligence data, the text analysis and insight capabilities of the large language model can be utilized, and a pre-trained large language model can be used to extract key information from the intelligence data, such as step 204.

[0073] Step 204: Use a pre-trained large language model to extract business behavior information from the first intelligence data; the business behavior information is used to describe the business behavior occurring between business elements related to the business object; the business elements include business entities or business events.

[0074] In the embodiments of the present specification, faced with the massive amount of unstructured text data in intelligence data, the outstanding performance of the large language model in natural language processing can be utilized to utilize the large language model to perform deep information extraction from complex reports, news and other intelligence data, which can greatly improve the efficiency and coverage of the analysis and ensure the timeliness and accuracy of the analysis results.

[0075] Among them, the pre-trained large language model can be pre-trained by learning a large amount of intelligence data. In practical applications, the large language model can include models such as GPT-3 based on the Transformer architecture. Specifically, the large language model can be implemented based on an open source large language model, such as Qwen-14B, Claude3Opus, etc. This application does not limit the architecture of the large language model.

[0076] The pre-trained large language model in the embodiments of this specification can be used to identify specific patterns in intelligence data.

[0077] Specifically, the pre-trained large language model can identify business elements and business behaviors of specific patterns. Still taking the analysis of financial objects as an example, for example, it can identify business elements of entity types such as "Alibaba". For another example, it can identify business elements of event types such as "oversupply of crude oil". For another example, it can identify business behaviors such as "acquisition", "increase holdings", and "forecasting".

[0078] Furthermore, the pre-trained large language model can identify specific structural patterns such as "subject business element-business behavior-object business element". In actual application, the pre-trained large language model can identify element information combinations consisting of "subject business element", "object business element" and "business behavior" applicable to the subject business element and the object business element. For example, business behavior information such as "Alibaba → Acquisition → Artificial Intelligence Startup Company" can be identified.

[0079] In step 204, the business behavior information may include information such as “subject business element-business behavior-object business element.” Thus, through step 204, structured business behavior information can be extracted from unstructured intelligence data to facilitate subsequent analysis.

[0080] Optionally, the business behavior information may include direct indicator information, and the direct indicator information may reflect the situation of a business event executed or triggered by a business entity.

[0081] In direct indicator information, the subject business element may include a business entity, and the object business element may include a business entity or a business event. For example, when the business object to be analyzed is a financial object, the subject business element may include an enterprise, a research institution, a fund service company, etc., and the object business element may include an enterprise, a project, a stock, a bond, an asset, a market trend, etc. As an example, the direct indicator information may be as follows, "Berkshire Hathaway → Buy → Coca-Cola Company Stock", "JPMorgan Chase → Issue → Corporate Bonds", "Alibaba → Acquire → Artificial Intelligence Startup Company", "Von Peterson Institute for Economic Research → Forecast → Global Economic Growth Slowdown", "Blackstone Group → Investment → New Energy Vehicle Project", "Goldman Sachs → Sell → European Real Estate Assets", etc.

[0082] Optionally, the business behavior information may further include environmental indicator information, and the environmental indicator information may reflect a situation in which one business event triggers another business event.

[0083] In the environmental indicator information, the subject business element may include a business event, and the object business element may include another business event. For example, when the business object to be analyzed is a financial object, the subject business element may include market trends, economic data releases, technological progress, etc., and the object business element may include the impact on a specific market sector, the impact on the value of a certain currency, etc. As an example, the environmental indicator information may be as follows, "The release of US non-farm payroll data → promotes → a stronger US dollar", "Clean energy technology breakthroughs → boost → renewable energy stocks", "Oversupply of crude oil → leads to → a decline in energy company stocks", etc.

[0084] In actual application, the business event can reflect the business environment in which the business object is located. In other words, the business event can carry business environment information reflecting the business environment in which the business object is located. In this case, by extracting business behavior information from intelligence data, and the business behavior information contains business events reflecting the business environment information, information reflecting the impact of the business environment information on the business object can be obtained by extracting the business behavior information.

[0085] Continuing with the example that the business objects to be analyzed include financial objects, the business environment information can reflect macroeconomic indicator information. The macroeconomic indicator information can be used to reflect the macro-financial environment in which the business object is located, or in other words, the global financial environment. For example, the macroeconomic indicator information can include gross domestic product, inflation and deflation, investment indicators, consumption indicators (including total retail sales of consumer goods, urban and rural residents' savings deposit balances, etc.), financial indicators (including interest rates, exchange rates, money supply, financial institutions' deposit and loan balances, total financial assets, etc.), fiscal indicators, etc., without limitation thereto. Further, the business environment information can also reflect market feature information related to the business object. The market feature information can be used to reflect a local financial environment that has a direct or indirect association with the business object. For example, the market feature information can include price volatility, trading volume, corporate financial reports or announcements, etc., of the business object or an associated business object of the business object, without limitation thereto.

[0086] In step 204, by extracting business behavior information from intelligence data, not only can the event information directly affecting the business object to be analyzed be obtained, but also the impact and trend of different industries, markets and social events can be obtained. Therefore, the method implemented in this specification based on this operation can obtain the potential impact of various internal and external factors on the business object, so as to provide more accurate analysis results.

[0087] Step 206: Query element relationship data related to the business object from a structural database; the structural database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data.

[0088] The structure database may specifically include a graph database, in which information may be stored in the form of a knowledge graph. In an embodiment of the present specification, a knowledge graph may be constructed to store element relationship data between business elements in the business domain to which the business object belongs in a structured form. In actual application, the element relationship data may be constructed based on business intelligence data in the business domain to which the business object belongs.

[0089] As an example, the business field to which the business object belongs may specifically include the financial field. More specifically, the business field to which the business object belongs may include a subdivided financial field corresponding to the business object. For example, the subdivided financial field may include a fund field, a stock field, a bond field, etc. In fact, there may be overlaps between different subdivided financial fields.

[0090] In the embodiments of this specification, Figure 2 Before executing the method in , a structured knowledge graph has been constructed based on the unstructured historical business intelligence data. Figure 2 When the method shown in is used, on the one hand, business behavior information can be extracted based on the current business intelligence data (such as steps 202 and 204); on the other hand, historical element relationship data related to the business object (corresponding to historical business behavior information) can be queried from the pre-built knowledge graph data. In this way, business behavior information originally hidden in a large amount of unstructured text can be easily obtained and stored in the form of structured data, thereby facilitating efficient and accurate analysis of business objects.

[0091] Knowledge graph data may be data used to describe a knowledge graph. In an abstract knowledge graph, nodes and edges connecting the nodes are included. Accordingly, in the knowledge graph data, node data used to describe the nodes and relationship data used to describe the edges may be included. In an embodiment of the present specification, the node data in the knowledge graph data may be used to represent business elements, and the edge data in the knowledge graph data may be used to represent business behaviors occurring between business elements.

[0092] The business field to which the business object belongs can specifically include the financial field as an example. The node data in the knowledge graph data may include enterprises, research institutions, fund service companies, projects, stocks, bonds, assets, market trends, economic data releases, technological progress, etc.; edge data may include operations (for example, increase holdings, decrease holdings), influences, etc.

[0093] Step 208: Based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine a prediction result of the characteristic parameter of the business object.

[0094] In actual application, based on the direct indicator information and environmental indicator information in the business behavior information, the direct factors and external factors that affect the value change trend of the characteristic parameters of the business object can be obtained; on this basis, combined with the element relationship data, the correlation relationship information between the known influencing factors can be obtained, so as to conduct in-depth mining based on various influencing factors; thereby, the prediction results of the characteristic parameters of the business object can be predicted more accurately and efficiently.

[0095] In actual application, before executing step 208, a trend prediction model that can identify the correlation between the values ​​of the characteristic parameters of the business object and various influencing factors can be pre-trained. Then, when executing step 208, both the business behavior information and the element relationship data can be input into the trend prediction model, so that the trend prediction model can predict the value change trend of the characteristic parameters of the business object based on this information.

[0096] Specifically, the predicted values ​​of the characteristic parameters of the business object at one or more future time points may be determined in step 208. When the predicted values ​​at multiple time points are predicted, a predicted value sequence may be obtained, that is, the changing trend of the characteristic parameters of the business object over a period of time in the future may be predicted.

[0097] In step 208, taking the business object including a fund as an example, the characteristic parameter may include at least one of a fund holding characteristic and a fund value characteristic. Among them, the fund holding refers to the various assets held by an investment fund at a certain point in time, including different types of investment tools such as stocks, bonds, and cash. The analysis of fund holdings helps to understand the investment strategy and market performance of the fund. Specifically, the fund holding characteristics may include information such as the fund holding ratio, the industry distribution ratio of the fund holdings, the asset distribution ratio of the fund holdings, and the fund holding concentration. The fund value characteristics may include information such as the market price of the fund.

[0098] It should be understood that in the methods described in one or more embodiments of this specification, the order of some steps can be adjusted according to actual needs, or some steps can be omitted.

[0099] Figure 2 The method in the paper extracts structured business behavior information from the current intelligence data associated with the business object by using a large language model on the one hand, and obtains element relationship data based on historical intelligence data from the knowledge graph on the other hand, which is used to reflect the association relationship between business elements related to the business object, and then analyzes and predicts the characteristic parameters of the business object. This can achieve rapid and in-depth analysis of a large amount of intelligence data and improve the timeliness of the analysis results of the business object.

[0100] based on Figure 2 The method, the embodiments of this specification also provide some improved or specific implementation methods of the method, which are described below.

[0101] In one or more embodiments of the present specification, crawler technology may be used to obtain the first intelligence data from various data sources. Alternatively, the first intelligence data may be obtained by calling APIs provided by various data sources. In actual applications, the means of obtaining intelligence data is not limited thereto.

[0102] In one or more embodiments of the present specification, the first intelligence data includes unstructured data. For the unstructured intelligence data, the intelligence data may be preprocessed before being input into the large language model to extract the business behavior information.

[0103] On the one hand, the acquired intelligence data may be preprocessed in a formal manner. Specifically, in order to facilitate subsequent information extraction and analysis, the acquired intelligence data may be preprocessed by such preprocessing operations as standardization, noise removal, and redundancy removal. Furthermore, a preset preprocessing script may be used to perform various preprocessing operations on the first intelligence data.

[0104] Furthermore, the standardization processing scheme may specifically include: using a preset standardization processing script to perform format standardization processing on the first intelligence data. Specifically, the data format of the intelligence data may be converted into a unified preset data format that is more convenient for subsequent data processing. Among them, the preset data format may include a CSV format or a database table format, etc. Usually, the initial data format of the first intelligence data may include a JSON format, an XML format, etc. Through standardization processing, data in the form of JSON format, XML format, etc. may be converted into data in the form of CSV format or a database table, etc.

[0105] Furthermore, the noise removal scheme may specifically include: using a preset noise processing script to perform noise processing on the first intelligence data. Specifically, information that is not closely related to the business object to be analyzed in the intelligence data may be removed to reduce the workload of subsequent data processing and analysis and improve analysis efficiency and accuracy.

[0106] Furthermore, the de-redundancy processing scheme may specifically include: using a preset redundancy processing script to perform redundancy processing on the first intelligence data. Specifically, duplicate information of the intelligence data may be identified, and the redundant parts of the duplicate information may be deleted. Thus, the workload of subsequent data processing and analysis may be reduced, and the analysis efficiency may be improved.

[0107] On the other hand, the first intelligence data may be preprocessed for feature extraction. Specifically, natural language processing technology may be used to perform word segmentation on the intelligence data, and then the named entities therein may be identified and marked, and then provided to the large language model to improve the efficiency and accuracy of information extraction of the large language model.

[0108] More specifically, before the use of the pre-trained large language model to extract the business behavior information from the first intelligence data (step 204), it may also include: first, the first intelligence data may be segmented to obtain segmented words after the text segmentation; then, named entity recognition may be performed on the segmented words to determine the named entities in the first intelligence data; thereafter, the named entities in the first intelligence data may be labeled to obtain the labeled first intelligence data. Thus, the use of the pre-trained large language model to extract the business behavior information from the first intelligence data (step 204) may specifically include: using the pre-trained large language model to extract the business behavior information from the labeled first intelligence data.

[0109] In practical applications, considering that the elements in intelligence data usually involve one or more named entities, by labeling the named entities in the intelligence data, the efficiency and accuracy of the pre-trained large model in extracting business behavior information from the intelligence data can be improved, and the probability of information extraction omissions can be reduced to a certain extent.

[0110] To facilitate understanding, the following is an example of the process of extracting business behavior information from the first intelligence data.

[0111] Suppose that a piece of intelligence data is obtained, specifically a news article about a fund's latest holdings report, and the article contains the following information, "According to the latest financial report, XYZ Fund increased its holdings in Company A's shares in the last quarter, while reducing its holdings in Company B."

[0112] Then, according to the embodiments of this specification, the data can be preprocessed for feature extraction by first performing word segmentation and then identifying named entities from the word segmentation, identifying "XYZ Fund", "Company A", "Company B" as key entities and "increase holdings" and "reduce holdings" as key operations.

[0113] Afterwards, step 204 may be performed to extract business behavior information from the first intelligence data using a pre-trained large language model. Specifically, the intelligence data annotated with the aforementioned key entities and key operations may be input into the large language model.

[0114] For business behavior information, following the example above, the large language model can identify the business behaviors of "increasing holdings" and "reducing holdings" and link these business behaviors with the corresponding business entities. The business behavior information output by the large language model can be in the following form, "XYZ Fund→Increasing holdings→Company A", "XYZ Fund→Reducing holdings→Company B". In this example, the structured representation of "entity-operation-entity" output by the large language model can be used to represent the fund's holdings for a specific company.

[0115] As a result, the large language model can extract key business behavior information intelligence from the input intelligence data and convert it into concise, reader-friendly text output to downstream analysis and prediction modules.

[0116] In one or more embodiments of the present specification, after extracting the current business behavior information from the current intelligence data, the current business behavior information can also be combined with the historical business information to construct a knowledge graph, that is, the results of understanding by the large language model are stored in the knowledge graph.

[0117] Optionally, if a knowledge graph has not yet been established, a knowledge graph may be established.

[0118] Specifically, after extracting the business behavior information from the first intelligence data using a pre-trained large language model (step 204), the method may also include: combining the business behavior information with historical business information related to the business object to determine the knowledge graph data related to the business object; wherein the historical business information may be extracted from second intelligence data related to the business object, and the second intelligence data may be available earlier than the first intelligence data used to extract the current business behavior information.

[0119] More specifically, the historical business information corresponding to the business object may include indicator information extracted from the second intelligence data using a pre-trained large language model. Furthermore, the second intelligence data may be information related to the business object, and the time from the available moment of the first historical reference information to the current moment may be greater than the preset time threshold. In other words, if a knowledge graph has not yet been established, all business behavior information (including current business behavior information and previous historical business information) can be extracted from all intelligence data related to the business object (including the currently acquired first intelligence data and the previously acquired second intelligence data) to construct a knowledge graph.

[0120] Optionally, if a knowledge graph has been currently established, the knowledge graph can be updated.

[0121] Specifically, after extracting business behavior information from the first intelligence data using a pre-trained large language model (step 204), the process may also include: combining the business behavior information with the current knowledge graph data corresponding to the business object to obtain updated knowledge graph data related to the business object.

[0122] More specifically, before updating the knowledge graph, the method may further include: pre-determining the knowledge graph data corresponding to the business field to which the business object belongs, and pre-building the knowledge graph. The pre-built knowledge graph data may include the association relationship between the business elements related to each business object in the business field.

[0123] Furthermore, the method of pre-constructing the knowledge graph may include: obtaining second intelligence data related to each business object in the corresponding business field, wherein the second intelligence data may be obtainable earlier than the first intelligence data; using a pre-trained large language model to extract historical business information from the second intelligence data, wherein the historical business information is used to describe business behaviors between business elements related to the business object; and determining the knowledge graph data related to the business object based on the historical business information.

[0124] Based on the embodiments of this specification, the extracted indicator information can be stored in a structured database in the form of a knowledge graph. By structuring the indicator information, it is convenient to obtain this information when analyzing any business object in the business field later.

[0125] In practical applications, the knowledge graph data may be data used to describe an abstract knowledge graph. In an abstract knowledge graph, nodes and edges connecting the nodes are included. Accordingly, in the knowledge graph data, node data used to describe the nodes and relationship data used to describe the edges may be included. In an embodiment of the present specification, the node data in the knowledge graph data may be used to represent business elements, and the edge data in the knowledge graph data may be used to represent business behaviors occurring between business elements.

[0126] Based on the embodiments of this specification, the structured business behavior information output by the information extraction module based on the large language model can be combined with the historical business behavior information, the current intelligence data can be linked with the historical intelligence data, and a relationship map between business elements can be constructed.

[0127] Furthermore, the knowledge graph data may carry time attribute information to clearly distinguish intelligence data at different times, thereby facilitating subsequent time-series-based analysis of business objects. The time attribute information may be used to reflect the timeliness of the indicator information stored in the knowledge graph data. More specifically, the time attribute information may be used to indicate the timeliness of at least one of the corresponding node data and edge data.

[0128] Optionally, the time attribute information may include timestamp information, which may include the time of acquiring the intelligence data used to generate the indicator information, or may include the time of extracting the indicator information from the intelligence data, or may include the time of storing the indicator information in the knowledge graph.

[0129] Optionally, the time attribute information may include validity period information. The validity period information may be specifically embodied as a validity period, which is used to indicate that the information represented by the corresponding node or edge is valid within the validity period. The validity period information may be specifically embodied as a valid expiration time, which is used to indicate that the information represented by the corresponding node or edge is valid before the valid expiration time.

[0130] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: setting validity period information for the node data and edge data in the knowledge graph; the validity period information is used to indicate the validity period of the corresponding node data or edge data.

[0131] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: for business elements that change over time in the knowledge graph, multiple node versions corresponding to the changed business elements are set; each node version carries a timestamp; different node versions among the multiple node versions reflect the status of the changed business elements at different times.

[0132] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: setting a specific type of edge in the knowledge graph to represent state changes, and the specific type of edge is used to connect different states of the same business element; different states are separate nodes and contain corresponding timestamps.

[0133] Continuing with the above example, in step 204, the business behavior information output by the information extraction module based on the large language model may include "XYZ Fund → Increase Holdings → Company A", "XYZ Fund → Reduce Holdings → Company B". And in step 206, element relationship data such as the historical holding trend of Company A's stock, the comparison of the operations of XYZ Fund and other funds on Company A's stock, etc. can be extracted from the knowledge graph data. In fact, the information stored in the knowledge graph data can not only reveal the single-point operations of XYZ Fund on Company A and Company B, but also display the complex relationship network from industry trends, market events to the operating modes of other funds.

[0134] Based on the solution of the embodiments of this specification, the business behavior information extracted from unstructured intelligence data using a large language model is structured and stored in the form of a knowledge graph. The knowledge graph not only contains explicit business operation information in the intelligence data, but also can deeply reflect the complex relationships and trends behind each business element through the associations between each other, thereby constructing a detailed knowledge graph covering a wide range of business entities and operations. Both this form and meaning can provide a solid and convenient data foundation for subsequent data analysis.

[0135] In one or more embodiments of the present specification, a specific implementation scheme for analyzing the business object based on the business behavior information and the element relationship data is also provided.

[0136] In one or more embodiments of this specification, when executing Figure 2 Before the prediction scheme of the method shown in the figure is implemented, a deep analysis of the characteristic parameters of the business object can be performed in advance based on the known characteristic parameters of the business object and the historical intelligence data to obtain the correlation between the value of the characteristic parameters of the business object and the intelligence data. Figure 2 The method shown can predict the values ​​of characteristic parameters of business objects based on known association relationships, current intelligence data, and element relationship data determined based on historical intelligence data.

[0137] In actual application, the predetermined association relationship between the value of the characteristic parameter of the business object and the intelligence data can be specifically embodied as a trend prediction model that has learned the association relationship between the value of the characteristic parameter of the business object and the intelligence data. The process of determining the association relationship between the value of the characteristic parameter of the business object and the intelligence data can specifically be a process of training a trend prediction model based on the historical values ​​of the characteristic parameters of the business object and the corresponding historical intelligence data.

[0138] Furthermore, the trend prediction model may be based on a machine learning model, and further, in order to obtain a trend prediction model with better prediction effect, time series analysis technology may be used to perform feature processing.

[0139] Specifically, time series analysis technology can be used to extract business object feature sequences from historical values ​​of feature parameters of multiple business objects in the business field to which the business object to be analyzed belongs at multiple times; then, a machine learning model can be trained based on the business object feature sequence and intelligence data related to the multiple business objects.

[0140] The historical intelligence data corresponds to the historical values ​​of the characteristic parameters of the multiple business objects. Specifically, they may correspond to a consistent historical time period. For example, the time period to which the historical intelligence data belongs may include the multiple moments of the historical values ​​of the characteristic parameters of the multiple business objects. For another example, the first time period to which the historical intelligence data belongs and the second time period to which the multiple moments of the historical values ​​of the characteristic parameters of the multiple business objects belong may at least partially overlap, so as to ensure the correlation between the historical intelligence data and the historical values ​​of the characteristic parameters of the multiple business objects.

[0141] In addition, in an embodiment of the present specification, the time period to which the historical intelligence data belongs may be the same as the time period to which the second intelligence data belongs. In this case, the historical intelligence data may include the second intelligence data.

[0142] Taking the business object including funds as an example, the characteristic parameters may include at least one of the multiple dimensions such as fund holding ratio, fund holding industry distribution, fund market value, etc. In actual application, the industry distribution, asset allocation ratio, holding trend and other dimensions of the fund can be analyzed based on historical data; the holding data at multiple time points can also be compared to find the patterns of long-term holding and short-term trading; and the correlation between each holding asset and external market factors (such as industry dynamics, macroeconomic events, policy changes, etc.) can also be considered.

[0143] Furthermore, the aforementioned correlation information can be learned by a pre-trained trend prediction model, and then used for subsequent predictions. Specifically, the pre-trained trend prediction model can identify the relationship between the value of the characteristic parameter of the business object and the intelligence data, and then be used to predict the change trend of the characteristic parameter of the business object based on the business behavior information extracted from the current intelligence data.

[0144] Furthermore, in the embodiments of this specification, when training a model, the data in the knowledge graph can also be used as context information of the input features to improve the prediction accuracy of the trained model. When training a model and using the model for prediction, the data in the knowledge graph can also be used as a reference to improve the accuracy and efficiency of the model prediction.

[0145] Specifically, based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine the prediction results of the characteristic parameters of the business object (step 208), which may specifically include: first, extracting features from the business behavior information and the element relationship data to obtain feature information for input into the trend prediction model; then, inputting the feature information into the pre-trained trend prediction model, the trend prediction model is trained based on intelligence data related to multiple business objects in the business field to which the business object belongs and historical values ​​of the characteristic parameters of the multiple business objects at multiple times; thereafter, the prediction results of the characteristic parameters of the business object output by the trend prediction model can be obtained.

[0146] In practical applications, in the process of determining the characteristic information for input into the trend prediction model, the data to be used can be standardized and converted into a format suitable for time series analysis and machine learning models.

[0147] In actual application, the data used to train the trend prediction model and the data input when using the trained trend prediction model are generally consistent to ensure the applicability of the model. In the embodiments of this specification, in the process of training the model, intelligence data related to each business object within a period of time and the historical values ​​of the characteristic parameters of each business object within the period of time are used; and in the process of using the model, intelligence data related to the current business object at the current time is used, and on this basis, element relationship data related to the intelligence data previously obtained related to the current business object is also used to predict the prediction results of the characteristic parameters of the current business object; thereby, the applicability of the model can be ensured.

[0148] Furthermore, in order to improve the prediction accuracy of the trend prediction model, in actual application, the trend prediction model may include not only the prediction results of the machine learning model, but also the prediction results of the time series model. Specifically, the trained time series model and the machine learning model may be used to predict the characteristic parameters of the business object in the same time period.

[0149] More specifically, the trend prediction model includes an integrated model obtained by integrating a time series model and a machine learning model; obtaining the prediction result of the characteristic parameters of the business object output by the trend prediction model may specifically include: obtaining a first prediction result output by the time series model; and obtaining a second prediction result output by the machine learning model; and then, using a preset result integration rule to integrate the first prediction result and the second prediction result to obtain the trend prediction result.

[0150] The result integration rules may include, but are not limited to, the average method, the weighted average method, the voting method, etc. In practical applications, the integration method of the output results of the two models can be determined based on a comprehensive consideration of multiple factors such as the prediction performance of the two models and the stability of the prediction results. For example, if the prediction performance of the two models is comparable, the average method or the weighted average method can be used for integration; if the prediction performance of a model is significantly better than that of another model, the model with better performance can be given a greater weight.

[0151] Optionally, the time series model may include an LSTM (Long Short-Term Memory) neural network, an ARIMA (Autoregressive Integrated Moving Average) model, etc., but is not limited thereto.

[0152] Optionally, the machine learning model may include a random forest model, a gradient boosted tree (GBDT) model, etc., but is not limited thereto. In practical applications, the machine learning model may include a classification model or a regression model. Taking the business object including a fund as an example, for the classification model, the model classification result can be used to predict whether the fund will increase or decrease the holdings of a specific stock; for the regression model, the model output result can be used to predict the change trend or amplitude of the model's holding ratio for a specific stock.

[0153] In one or more embodiments of the present specification, not only can the characteristic parameters of the business object (for example, the fund's position data, market value, etc.) be predicted, but also the risks that the business object may currently face can be predicted based on the prediction results of the characteristic parameters of the business object.

[0154] Specifically, after determining the prediction result of the characteristic parameter of the business object by using the pre-trained trend prediction model based on the business behavior information and the element relationship data, the method may further include: determining the prediction result of the risk characteristic of the business object based on the prediction result of the characteristic parameter of the business object and the element relationship data. More specifically, a risk assessment model may be customized in advance, and the risk assessment model may be used to predict the risk characteristics of the business object.

[0155] Take the business object including funds as an example. For example, it is possible to combine risk management tools such as VaR (Value at Risk) model, Monte Carlo simulation, and customize risk assessment models based on historical volatility and correlation analysis; then, the predicted position changes, market characteristics and macroeconomic indicators can be used as inputs to use the risk assessment model to assess the maximum potential losses and risk sources that the fund may face; then, the risk assessment results can be generated, including the main risks faced by the fund, the expected maximum potential losses (for example, the VaR value at a 95% confidence level), and the individual positions and market factors that have the greatest impact on the fund.

[0156] In one or more embodiments of the present specification, after obtaining the prediction results and risk assessment results, an evaluation report on the business object to be analyzed may be generated and sent to the terminal user.

[0157] Specifically, after determining the prediction result of the characteristic parameters of the business object by using the pre-trained trend prediction model based on the business behavior information and the element relationship data, the method may further include: using a large language model to generate an evaluation report based on the prediction result of the characteristic parameters of the business object; and sending the evaluation report to a user terminal logged in to a user account related to the business object. Furthermore, the evaluation report may also include a risk prediction result.

[0158] Optionally, taking the example of a business object including a fund, the user accounts related to the fund may include user accounts that hold the fund, and may also include user accounts that do not hold the fund but hold asset objects in the fund. Optionally, the user accounts related to the fund may also include user accounts that have accessed information about the fund, or may also include user accounts that hold other funds similar to the fund. In practice, the user accounts that receive the evaluation report may be determined based on actual business needs.

[0159] In practical applications, all analysis results can be integrated into an easy-to-understand report. And, since the report is generated by a large language model, the report can contain not only data and charts, but also explanatory text generated by the large language model to facilitate users to understand the deep meaning of the analysis.

[0160] As described above, based on one or more embodiments of the present specification, a business object analysis system is provided, and the business object analysis system is used to perform data collection and preprocessing, information extraction and relationship mining, in-depth position analysis, trend forecasting and risk assessment, report generation and strategy recommendations, and other stages.

[0161] The following is a detailed introduction taking the business objects including funds as an example.

[0162] Financial analysis is an important application scenario of big data analysis technology. In the financial field, financial analysis is one of the important tools for investors to make wise investment decisions. Before artificial intelligence technology was widely used in financial analysis, it mainly relied on the experience and knowledge of professional analysts. Taking fund holdings analysis as an example, the process of fund holdings analysis includes experts reading and parsing the fund's quarterly reports, annual reports, market news, management team interviews and other relevant financial and economic data to conduct analysis. Based on this information, experts can evaluate the fund's holdings structure, investment strategy, market performance and other aspects, and give professional opinions on future trends and potential investment risks.

[0163] However, with the expansion of financial markets and the dramatic increase in data volume, the model of manual analysis by experts faces huge challenges in processing large-scale data. Traditional expert analysis methods are time-consuming and require a lot of manpower to collect, organize and analyze data. Due to human resource limitations, it may take weeks or even months from data acquisition to the output of analysis results, resulting in poor timeliness of the obtained analysis results and difficulty in quickly responding to market changes. In view of this, it is necessary to provide a more timely analysis method for financial objects such as funds.

[0164] With the continuous advancement of artificial intelligence technology, it has become particularly important to use advanced data analysis tools to extract and analyze information from numerous data sources. This is because, first of all, the analysis of large-scale text data has become a new challenge. Fund holdings reports, market news, analyst opinions, financial statements and other rich and large text data contain rich information. How to accurately and quickly extract key information that affects fund holdings from these unstructured data is a major challenge in the current financial technology field. In addition, the complexity of the financial market requires that analysis tools should not only be able to understand individual entities, but also be able to grasp the relationship between them. The analysis of a single stock or bond can no longer meet the needs of investors. Analysis tools must be able to grasp the dynamics of the entire market, including macroeconomic trends, industry development, and competitive and cooperative relationships between enterprises. The analysis method for business objects containing financial objects such as funds based on a large language model and a structured database provided in the embodiments of this specification can at least partially solve the above problems.

[0165] The following is an example of a specific application of the solution of the embodiment of this specification in the financial field.

[0166] Specifically, in the data collection and preprocessing stage: First, an efficient and scalable data collection system is built, which can continuously collect information from multiple data sources, including but not limited to public fund reports, stock exchange announcements, financial news websites, social media platforms and other financial data services. Then, the collected data is standardized, de-noised and de-redundant to facilitate subsequent information extraction and analysis. Next, the data preprocessing module can use natural language processing (NLP) technology to perform preliminary analysis on text data, such as using tokenization to segment text into meaningful words or phrases, and using named entity recognition (NER) to identify key entities in the text, such as company names, market terms, etc.

[0167] Information extraction and relationship mining stage: After preprocessing, on the one hand, large language models can be used to conduct in-depth analysis and information extraction of text content. Specifically, specific neural network models can be trained to understand the patterns and context in text data. The trained models can automatically extract key financial indicators and opinions from financial reports, news articles, and market analysis. On the other hand, the system can also use knowledge graphs to discover the relationship connections between text information, such as financial connections between companies, industry classifications, and the association of market events. The knowledge graph can store these connections in the form of a graphical database, so that these relationships can be quickly queried and analyzed.

[0168] In-depth position analysis stage: The system can use the information extracted from the text and the structured data in the knowledge graph to comprehensively evaluate the fund's position structure. The algorithm can analyze the fund's industry distribution, asset allocation ratio, position change trend and other dimensions. The system can also compare the position data at multiple time points to discover the patterns of long-term holding and short-term trading. For each position asset, the system can also consider external market factors, such as the correlation with industry dynamics, macroeconomic events, policy changes, etc.

[0169] Trend prediction and risk assessment stage: Using machine learning and time series analysis techniques, the system can predict future changes in fund holdings and conduct risk assessments. By analyzing past and current data, the model can identify market trends and predict their possible impact on fund holdings. The system can also conduct risk assessments on the fund's investment portfolio based on a variety of financial risk models and generate risk-adjusted return expectations.

[0170] Report generation and strategy recommendation stage: The system can finally integrate all analysis results into an easy-to-understand report. The report can contain not only data and charts, but also explanatory text generated by a large language model to facilitate users to understand the deeper meaning of the analysis. Based on the analysis results, the system can also make strategic recommendations, such as possible buy / sell signals and portfolio adjustment recommendations.

[0171] The various technical features in the above embodiments can be combined arbitrarily as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they are not described one by one. Therefore, any combination of the various technical features in the above embodiments also falls within the scope of this specification.

[0172] Based on one or more embodiments of this specification, a large pre-trained language model is integrated for in-depth text understanding, and combined with a dynamically updated knowledge graph, detailed position analysis and prediction are provided. At the same time, personalized report generation and interactive user interface can be used, which greatly improves analysis efficiency, reduces subjective bias, and improves the ability to process unstructured data.

[0173] Based on one or more embodiments of the present specification, the following technical effects can be at least partially achieved.

[0174] First, it can provide more efficient and wider analysis results.

[0175] The system can simultaneously collect and process data from different sources, at different times and in different regions. The efficient data collection system can monitor market changes and important news in real time, greatly increasing the coverage and timeliness of analytical data, and ensuring that the analysis results reflect the latest market conditions as comprehensively and quickly as possible.

[0176] The system can seamlessly process large-scale data sets, and large language models can quickly perform text summarization, information extraction, and trend analysis without requiring a lot of time for reading and manual screening. The automated data collection, processing, and analysis processes greatly improve the speed of analysis and greatly reduce the burden of manual analysis. It can process and analyze large amounts of text in a relatively short period of time. Work that may take weeks for an analysis team to complete can be completed in a short period of time based on the system provided in the embodiments of the present application.

[0177] In addition, the knowledge graph supports fast query and relationship mining, further shortening the time to generate results.

[0178] Second, it can provide more in-depth, accurate and reliable analysis results.

[0179] Specifically, by understanding the context and making inferences, the system can provide accurate analysis in complex business environments. Considering the complexity of business decisions such as finance, the system can understand abstract concepts and consider the interactions between different positions in the analysis. Moreover, the technology combined with knowledge graphs can provide deeper insights than traditional analysis.

[0180] The system provides more objective data-driven analysis and reduces human bias. The machine learning model makes predictions based on historical data and statistical laws, eliminating the influence of expert subjective opinions. Automation reduces common mistakes in human analysis, reduces errors, and improves the overall reliability of analysis.

[0181] Third, it can process large amounts of unstructured data.

[0182] The system utilizes a large language model, which has learned to recognize complex patterns and implicit meanings of language through training on large-scale data sets, and can understand natural language and extract and transform useful, structured information from it. In addition, combined with the capabilities of knowledge graphs, the system can also understand and characterize entities and their relationships, which makes unstructured data analysis not only possible, but also a strength of the system provided in the embodiments of this specification.

[0183] Fourth, the analysis system provided is capable of dynamic learning and updating.

[0184] Model-based systems can learn dynamically and evolve over time. The system's algorithms can adjust themselves based on the latest information without human intervention, and continuously improve their predictive capabilities through continuous learning, accurately reflecting the latest market trends. Through continuous learning, the model is constantly optimizing its analysis and decision-making process to achieve more fair and consistent analysis results.

[0185] Fifth, it can provide personalized strategy recommendations based on the analysis results.

[0186] The system can not only provide users with standardized position analysis, but also give personalized investment advice based on users' specific needs and preferences. And with the use of large language models, users can interact with the system, ask questions and get instant, detailed feedback and analysis, further enhancing the user experience.

[0187] Based on the same idea, the embodiments of this specification also provide a device corresponding to the above method.

[0188] Figure 3 The embodiments of this specification provide corresponding to Figure 2 A structural schematic diagram of a business object analysis device.

[0189] like Figure 3 As shown, the device may include:

[0190] The intelligence data acquisition module 302 is used to acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold;

[0191] An information extraction module 304 is used to extract business behavior information from the first intelligence data using a pre-trained large language model; the business behavior information is used to describe the business behavior occurring between business elements related to the business object, and the business elements include business entities or business events;

[0192] A structure data acquisition module 306, configured to query element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data;

[0193] The analysis module 308 is used to determine the prediction result of the characteristic parameter of the business object based on the business behavior information and the element relationship data using a pre-trained trend prediction model.

[0194] based on Figure 3 The present specification also provides some specific implementation schemes of the method, which are described below.

[0195] Optionally, the business object analysis device also includes a preprocessing module, which is used to: perform text segmentation on the first intelligence data to obtain word segments after text segmentation; perform named entity recognition on the word segments to determine the named entities in the first intelligence data; and label the named entities in the first intelligence data to obtain the labeled first intelligence data.

[0196] Optionally, the business object analysis device also includes a graph construction module, which is used to: combine the business behavior information with historical business information related to the business object to determine the knowledge graph data related to the business object; the historical business information is extracted from the second intelligence data related to the business object.

[0197] Optionally, the business object analysis device also includes a graph update module, which is used to: combine the business behavior information with current knowledge graph data corresponding to the business object to obtain updated knowledge graph data related to the business object.

[0198] Optionally, the structural database specifically includes a graph database, in which the knowledge graph data of the graph database, the node data is used to represent the business elements, and the edge data is used to represent the business behaviors occurring between the business elements.

[0199] Optionally, the knowledge graph data carries time attribute information; the time attribute information is used to indicate the timeliness of at least one of the node data and the edge data.

[0200] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: setting validity period information for the node data and edge data in the knowledge graph; the validity period information is used to indicate the validity period of the corresponding node data or edge data.

[0201] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: for business elements that change over time in the knowledge graph, multiple node versions corresponding to the changed business elements are set; each node version carries a timestamp; different node versions among the multiple node versions reflect the status of the changed business elements at different times.

[0202] Optionally, the specific form of carrying the time attribute information in the knowledge graph data may include: setting a specific type of edge in the knowledge graph to represent state changes, and the specific type of edge is used to connect different states of the same business element; different states are separate nodes and contain corresponding timestamps.

[0203] Optionally, the analysis module 308 can be specifically used to: perform feature extraction on the business behavior information and the element relationship data to obtain feature information for input into a trend prediction model; input the feature information into a pre-trained trend prediction model; the trend prediction model is trained based on intelligence data related to multiple business objects in the business field to which the business object belongs and historical values ​​of feature parameters of the multiple business objects at multiple moments; and obtain a prediction result of the feature parameters of the business object output by the trend prediction model.

[0204] Optionally, the trend prediction model includes an integrated model obtained by integrating a time series model and a machine learning model; obtaining the prediction result of the characteristic parameters of the business object output by the trend prediction model may specifically include: obtaining a first prediction result output by the time series model; obtaining a second prediction result output by the machine learning model; and integrating the first prediction result and the second prediction result using a preset result integration rule to obtain the trend prediction result.

[0205] Optionally, before inputting the feature information into the pre-trained trend prediction model, the process may also include training the trend prediction model, specifically, including: using time series analysis technology to extract a business object feature sequence from the historical values ​​of feature parameters of multiple business objects in the business field to which the business object belongs at multiple moments; and training a machine learning model based on the business object feature sequence and intelligence data related to the multiple business objects.

[0206] Optionally, the business object analysis device may also include a report generation and push module, which is used to: after determining the prediction results of the characteristic parameters of the business object based on the business behavior information and the element relationship data using a pre-trained trend prediction model, generate an evaluation report based on the prediction results of the prediction results of the characteristic parameters of the business object using a large language model; and send the evaluation report to a user terminal that is logged in to a user account related to the business object.

[0207] Optionally, the business object includes a fund; and the characteristic parameter includes at least one of a fund holding characteristic and a fund value characteristic.

[0208] It is understood that the above modules refer to computer programs or program segments for executing one or more specific functions. In addition, the distinction between the above modules does not mean that the actual program codes must also be separated.

[0209] The above is a schematic scheme of a business object analysis device of this embodiment. It should be noted that the technical scheme of the business object analysis device and the technical scheme of the above-mentioned business object analysis method belong to the same concept, and the details not described in detail in the technical scheme of the business object analysis device can be found in the description of the technical scheme of the above-mentioned business object analysis method.

[0210] Based on the same idea, the embodiments of this specification also provide a device corresponding to the above method.

[0211] Figure 4 The embodiments of this specification provide corresponding to Figure 2 A structural schematic diagram of a business object analysis device.

[0212] like Figure 4 As shown, the device 400 may include:

[0213] at least one processor 410; and,

[0214] A memory 430 is communicatively connected to the at least one processor; wherein,

[0215] The memory 430 stores instructions 420 executable by the at least one processor 410. The instructions are executed by the at least one processor 410 to enable the at least one processor 410 to:

[0216] Acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold;

[0217] Using a pre-trained large language model, extracting business behavior information from the first intelligence data; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events;

[0218] Querying element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data;

[0219] Based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine the prediction result of the characteristic parameter of the business object.

[0220] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned business object analysis method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above-mentioned business object analysis method.

[0221] An embodiment of the present application further provides a computer-readable storage medium storing computer instructions, which implement the steps of the business object analysis method as described above when executed by a processor.

[0222] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above-mentioned business object analysis method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above-mentioned business object analysis method.

[0223] An embodiment of the present specification also provides a computer program product, including a computer program / instruction, which implements the steps of the above-mentioned business object analysis method when executed by a processor.

[0224] The above is a schematic scheme of a computer program product of this embodiment. It should be noted that the technical scheme of the computer program product and the technical scheme of the above-mentioned business object analysis method belong to the same concept, and the details not described in detail in the technical scheme of the computer program product can be referred to the description of the technical scheme of the above-mentioned business object analysis method.

[0225] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, computer program product or medium embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device, equipment, computer program product or medium provided in the embodiment of this specification corresponds to the method, so the device, equipment, computer program product or medium also has beneficial technical effects similar to the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding device, equipment, computer program product or medium will not be repeated here.

[0226] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0227] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0228] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0229] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0230] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0231] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0232] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0233] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0234] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0235] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0236] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0237] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0238] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0239] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0240] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A business object analysis method, comprising: Acquire first intelligence data related to the business object to be analyzed; The time from the available time of the first intelligence data to the current time is not greater than a preset time threshold; Using a pre-trained large language model, extracting business behavior information from the first intelligence data; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events; Querying element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from the second intelligence data; The time at which the second intelligence data is available is earlier than the time at which the first intelligence data is available; Based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine the prediction result of the characteristic parameter of the business object.

2. The method according to claim 1, wherein the first intelligence data comprises unstructured data; and before extracting the business behavior information from the first intelligence data using a pre-trained large language model, the method further comprises: Performing text segmentation on the first intelligence data to obtain segmented words after the text segmentation; performing named entity recognition on the word segmentation to determine named entities in the first intelligence data; The named entities in the first intelligence data are labeled to obtain labeled first intelligence data.

3. The method according to claim 1, after extracting the business behavior information from the first intelligence data using a pre-trained large language model, further comprises: Combining the business behavior information with historical business information related to the business object to determine knowledge graph data related to the business object; or, The business behavior information is combined with the current knowledge graph data corresponding to the business object to obtain updated knowledge graph data related to the business object.

4. The method of claim 1, wherein: The structural database specifically includes a graph database. In the knowledge graph data of the graph database, node data is used to represent business elements, and edge data is used to represent business behaviors occurring between business elements.

5. The method of claim 4, wherein: The knowledge graph data carries time attribute information; the time attribute information is used to indicate the timeliness of at least one of the node data and the edge data.

6. The method of claim 5, wherein: The specific forms of carrying the time attribute information in the knowledge graph data include: Setting validity period information for the node data and edge data in the knowledge graph; the validity period information is used to indicate the validity period of the corresponding node data or edge data; or, For the business elements in the knowledge graph that change over time, multiple node versions corresponding to the changed business elements are set; each node version carries a timestamp; different node versions in the multiple node versions reflect the status of the changed business elements at different times; or, A specific type of edge is set in the knowledge graph to represent state changes, and the specific type of edge is used to connect different states of the same business element; different states are treated as separate nodes and contain corresponding timestamps.

7. The method according to claim 1, wherein the prediction result of the characteristic parameter of the business object is determined based on the business behavior information and the element relationship data using a pre-trained trend prediction model, specifically comprising: Extracting features from the business behavior information and the element relationship data to obtain feature information for input into a trend prediction model; Inputting the characteristic information into a pre-trained trend prediction model; the trend prediction model is trained based on intelligence data related to multiple business objects in the business field to which the business object belongs and historical values ​​of characteristic parameters of the multiple business objects at multiple times; Obtain the prediction result of the characteristic parameters of the business object output by the trend prediction model.

8. The method of claim 7, wherein: The trend prediction model includes an integrated model obtained by integrating a time series model and a machine learning model; the prediction result of the characteristic parameters of the business object output by the trend prediction model specifically includes: Obtaining a first prediction result output by the time series model; Obtaining a second prediction result output by the machine learning model; The first prediction result and the second prediction result are integrated using a preset result integration rule to obtain the trend prediction result.

9. The method according to claim 8, before inputting the feature information into the pre-trained trend prediction model, further comprising: Using time series analysis technology, extracting a business object feature sequence from historical values ​​of feature parameters of multiple business objects in the business field to which the business object belongs at multiple times; A machine learning model is trained based on the business object feature sequence and intelligence data related to the multiple business objects.

10. The method according to claim 1, after determining the prediction result of the characteristic parameter of the business object based on the business behavior information and the element relationship data using a pre-trained trend prediction model, further comprising: Using a large language model, generating an evaluation report based on the prediction results of the characteristic parameters of the business object; The evaluation report is sent to a user terminal logged in to a user account related to the business object.

11. The method according to any one of claims 1 to 10, wherein: The business object includes a fund; and the characteristic parameter includes at least one of a fund holding characteristic and a fund value characteristic.

12. A business object analysis device, comprising: An intelligence data acquisition module, used to acquire first intelligence data related to the business object to be analyzed; The time from the available time of the first intelligence data to the current time is not greater than a preset time threshold; An information extraction module, configured to extract business behavior information from the first intelligence data using a pre-trained large language model; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events; a structure data acquisition module, used to query element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from the second intelligence data; The time at which the second intelligence data is available is earlier than the time at which the first intelligence data is available; The analysis module is used to determine the prediction results of the characteristic parameters of the business object based on the business behavior information and the element relationship data using a pre-trained trend prediction model.

13. A business object analysis device, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire first intelligence data related to the business object to be analyzed; the time from the available time of the first intelligence data to the current time is not greater than a preset time threshold; Using a pre-trained large language model, extracting business behavior information from the first intelligence data; the business behavior information is used to describe business behaviors occurring between business elements related to the business object, the business elements including business entities or business events; Querying element relationship data related to the business object from a structure database; the structure database includes information determined based on historical business information extracted from second intelligence data; the second intelligence data is available earlier than the first intelligence data; Based on the business behavior information and the element relationship data, a pre-trained trend prediction model is used to determine the prediction result of the characteristic parameter of the business object.

Images