Big model-based risk early warning method and device, electronic equipment and storage medium

Abstract

The application relates to the computer technical field and discloses a risk early warning method and device based on a large model, an electronic device and a storage medium. The method comprises the following steps: acquiring to-be-processed data; performing text processing on the to-be-processed data to obtain to-be-early-warned data; acquiring early warning field information and early warning rules corresponding to the early warning field; screening the to-be-early-warned data to determine to-be-recognized data; generating early warning prompt words and inputting the early warning prompt words into an early warning model to obtain a risk early warning result of a first detection object. The application integrates data in multiple data sources into the form of to-be-processed data, can decouple the data collection and risk early warning process, and solves the problem that the system needs to be overall reconstructed when a new data source is adapted or a model version is upgraded. In addition, the corresponding early warning field information and early warning rules are input according to requirements, the early warning personnel operation can be simplified in the case that the early warning scene needs to be expanded, and the system does not need to be reconstructed.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to risk warning methods, devices, electronic devices, and storage media based on large models. Background Technology

[0002] In today's era of accelerated digital transformation and increasingly complex risk scenarios, early warning intelligent agents have become a core technological support for key areas such as financial risk control, disaster monitoring, and public safety. However, most early warning intelligent agents in related technologies adopt an "integrated customized development" model, with system components and core algorithms deeply coupled. This architecture is gradually revealing its limitations in the context of rapid technological iteration: when it is necessary to adapt to new data sources, upgrade model versions, or expand early warning scenarios, the entire system needs to be reconstructed, resulting in a long adaptation cycle and lagging behind the real-time needs of risk prevention and control. Summary of the Invention

[0003] This application provides a risk early warning method, device, electronic device, and storage medium based on a large model to solve the problem in related technologies that when adapting to new data sources, upgrading model versions, or expanding early warning scenarios, the entire system needs to be reconstructed, resulting in a long adaptation cycle and lagging behind the real-time needs of risk prevention and control.

[0004] Firstly, this application provides a risk warning method based on a large model, including: The data to be processed is obtained from the first detection object. The first detection object has multiple data sources, which are used to generate the data to be processed.

[0005] The data to be processed is converted into text to obtain the data to be warned. The data to be warned includes the corresponding first field and the value of the first field.

[0006] Obtain the warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object. The warning field information includes the warning field name.

[0007] Based on the warning field information and the first field, the data to be warned is filtered to determine the data to be identified.

[0008] Based on the warning field name, warning rules, and data to be identified, a warning prompt word is generated and input into the warning model to obtain the risk warning result for the first detection object.

[0009] The risk warning method based on a large model provided in this embodiment decouples the data collection and risk warning processes by integrating data from multiple data sources into data to be processed after the data is generated from the data source. This means that when a new data source needs to be imported, the risk warning process does not need to be modified; instead, the data from the new data source is directly added to the data to be processed, solving the problem of overall system reconstruction required to adapt to new data sources in related technologies. By text-based processing of the data to be processed, the system obtains the data to be warned, including the first field and its corresponding value. This allows the system to quickly filter out the field data related to the warning from massive amounts of data, forming the data to be identified. This reduces the reading of unnecessary data during warning analysis, thereby reducing system overhead and improving warning efficiency. By allowing risk warning personnel to input warning field information and warning rules, it is easier for them to control specific warning scenarios. When the warning scenario needs to be expanded, they can input the corresponding warning field information and warning rules as needed, simplifying operations without requiring system reconstruction. Pre-screening reduces unnecessary field data in the data to be warned, ensuring that only the data truly needed for warning (i.e., the data to be identified) is input into the warning model, reducing the resource consumption of the warning model and improving warning efficiency. By separating the warning field information, warning rules, and data to be identified from the warning model, the warning model can be directly replaced or updated when upgrading the model version without affecting the processing of the data to be identified. This eliminates the need to reconstruct the system and solves the problem of long model adaptation cycles in related technologies.

[0010] In some optional implementations, the data to be processed is converted into text to obtain data to be alerted, including: The data to be processed is converted into text, and the text-based data to be processed is determined.

[0011] Extract the text information from the text-based data to be processed to obtain the corresponding first field and its value, and generate the warning data corresponding to the data to be processed.

[0012] In some optional implementations, the data to be processed is text-based, and the text-based data to be processed is determined, including: The data to be processed is converted into text using a preset logical algorithm, and the text-converted data to be processed is determined.

[0013] or, The data to be processed is analyzed using a pre-defined large language model to extract textual information and determine the textualized data to be processed.

[0014] In some optional implementations, the data to be warned is filtered based on the warning field information and the first field to determine the data to be identified, including: A filter statement is generated based on the warning field information. The filter statement includes the warning field information.

[0015] The data to be identified is obtained by filtering the data to be warned using filtering statements.

[0016] In some optional implementations, after obtaining the risk warning result for the first detection target, the method further includes: Obtain feedback information on the risk warning results, update the risk warning results based on the feedback information, and determine the updated risk warning results.

[0017] The data to be identified and the updated risk warning results are used as warning samples, and the warning samples are added to the warning sample library. The warning samples in the warning sample library are used to assist the warning model in determining the risk warning results.

[0018] In some optional implementations, corresponding warning prompts are generated based on the warning field name, warning rules, and data to be identified, including: Generate corresponding initial warning prompts based on the warning field name, warning rules, and data to be identified.

[0019] Add at least one warning sample from the warning sample library to the initial warning prompt word to generate the corresponding warning prompt word.

[0020] In some alternative implementations, the risk warning method based on large models also includes: Get the number of warning samples in the warning sample library.

[0021] If the number of samples exceeds a preset threshold, add the warning samples from the warning sample library to the warning training sample library and delete the warning samples from the warning sample library.

[0022] The early warning model is fine-tuned using an early warning training sample library to obtain the fine-tuned early warning model, which is then used as the early warning model for risk warning.

[0023] The risk warning method based on a large model provided in this embodiment decouples the data collection and risk warning processes by integrating data from multiple data sources into a single data set to be processed after the data is generated from the data source. This means that when a new data source needs to be imported, the risk warning process does not need to be modified; instead, the data from the new data source is directly added to the data set to be processed, solving the problem of requiring a complete system reconstruction to adapt to new data sources in related technologies. By text-based processing of the data to be processed, the system obtains the data to be warned, including the first field and its corresponding value, making it easier to filter data requiring risk warning from massive amounts of data and improving warning efficiency. By allowing risk warning personnel to input warning field information and warning rules, they can more easily control specific warning scenarios. When the warning scenario needs to be expanded, they can input the corresponding warning field information and warning rules as needed, simplifying operations and eliminating the need for system reconstruction. Pre-screening reduces the amount of data to be warned, ensuring that only the data truly requiring warning (i.e., the data to be identified) is input into the warning model, reducing the resource consumption of the warning model and improving warning efficiency. By separating the warning field information, warning rules, and data to be identified from the warning model, the warning model can be directly replaced or updated when upgrading the model version without affecting the processing of the data to be identified, and without requiring system reconstruction. Continuous optimization through fine-tuning of the warning model makes it better suited to the warning needs of risk warning personnel. Subsequent warning analysis can then be based on the fine-tuned warning model, improving its learning ability and making risk warning results more accurate.

[0024] Secondly, this application provides a risk warning device based on a large model, comprising: The data acquisition module is used to acquire data to be processed. The data to be processed comes from the first detection object, which has multiple data sources. The data sources are used to generate the data to be processed.

[0025] The text processing module is used to process the data to be processed into text to obtain the data to be warned. The data to be warned includes the corresponding first field and the value of the first field.

[0026] The field determination module is used to obtain the warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object. The warning field information includes the warning field name.

[0027] The filtering module is used to filter the data to be warned based on the warning field information and the first field, and to determine the data to be identified.

[0028] The prompt word generation module is used to generate warning prompt words based on the warning field name, warning rules and the data to be identified, and input the warning prompt words into the warning model to obtain the risk warning result of the first detection object.

[0029] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the risk warning method based on a large model as described in the first aspect or any corresponding embodiment.

[0030] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the risk warning method based on a large model as described in the first aspect or any corresponding embodiment.

[0031] Fifthly, this application provides a computer program product, including computer instructions for causing a computer to execute the risk warning method based on a large model as described in the first aspect or any corresponding embodiment. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application; Figure 2 This is a schematic diagram of the first type of risk warning method based on a large model according to an embodiment of this application; Figure 3 This is a schematic diagram of a second process for a risk warning method based on a large model according to an embodiment of this application; Figure 4 This is a risk warning flowchart according to an embodiment of this application; Figure 5 This is a structural block diagram of a risk warning device based on a large model according to an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0036] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0037] As one optional application scenario in the embodiments of this application, such as Figure 1 As shown, application 101 is installed in terminal device 110, and user 130 can interact with application 101 through terminal device 110 and / or access device of terminal device 110.

[0038] For example, application 101 can be any application that provides risk warning services. Figure 1 In the application scenario shown, if application 101 is active, the terminal device 110 can display the interface 102 of application 101. The interface 102 may include various pages that application 101 can provide, such as interactive pages, settings pages, warning pages, etc.

[0039] In some embodiments, terminal device 110 is communicatively connected to server 120 to provide services to application 101. Terminal device 110 may be a mobile terminal, fixed terminal, or portable terminal, etc., including but not limited to mobile phones, desktop computers, laptop computers, multimedia tablets, e-book devices, gaming devices, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, terminal device 110 may also support any type of interface, and server 120 may be various types of computing systems or servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, computing devices in cloud environments, etc.

[0040] It should be noted that, Figure 1This is merely an example of an application scenario and does not limit the scope of protection of this application.

[0041] The embodiments of this application will be described below with reference to the accompanying drawings. It should be understood that the pages shown in the drawings are merely examples, and various page designs are possible in practice. The various graphic elements on the page may have different arrangements and different visual representations, one or more elements may be omitted or replaced, and one or more other elements may also be present; no limitations are imposed on the embodiments of this application. Furthermore, the embodiments are primarily described below with reference to terminal device 110. It should be understood that the actions described relative to terminal device 110 can be performed by application 101 on terminal device 110, or can be performed by application 101 in conjunction with its server (e.g., server 120).

[0042] This application provides a risk warning method based on a large model. By integrating data from multiple data sources into a single data set to be processed, the data collection and risk warning processes can be decoupled. When importing new data sources or upgrading the model version, the risk warning process does not need to be modified. Furthermore, by having risk warning personnel input warning field information and rules, they can more easily control specific warning scenarios and simplify operations when the warning scenario needs to be expanded, by inputting the corresponding warning field information and rules as required. Secondly, by text-processing the data to be processed to obtain the data to be warned, including the first field and its corresponding value, it is easier to filter the data that needs risk warning from massive amounts of data, improving warning efficiency.

[0043] According to an embodiment of this application, a risk warning method based on a large model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0044] This embodiment provides a risk warning method based on a large model, which can be used in the aforementioned terminal devices, such as desktop computers, mobile phones, and tablet computers. Figure 2 This is a flowchart of a risk warning method based on a large model according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain the data to be processed. The data to be processed comes from the first detection object. The first detection object has multiple data sources, which are used to generate the data to be processed.

[0045] During the risk warning process, data from various data sources of the first detection target can be collected to obtain data to be processed. For example, data from various data sources of the first detection target can be collected at preset intervals and used as data to be processed for subsequent risk warnings.

[0046] For example, the first detection target can be a monitored system or application. For instance, the first detection target could be a message board application, a forum application, or a real-time text messaging system. For the first detection target, the data sources that can generate data can be identified through the system architecture. Based on this, it can be determined in advance which data sources need to be acquired before issuing a risk warning for the first detection target. The determination of which data sources to acquire can be set according to actual needs, and no limitations are imposed here.

[0047] In some alternative implementations, the data source may include signaling data sources, communication data sources, etc. Accordingly, the collected data may include signaling data and communication data, etc. For example, when the first detection object is a forum application, it includes signaling data generated by user operations on the client and communication data between the client and the forum server.

[0048] The data to be processed can be used to represent the aggregate of data collected within the same risk warning period, meaning it includes data from multiple data sources. After the data sources produce their data, the data from these multiple sources is integrated into the form of the data to be processed. In other words, before issuing a risk warning, the data to be processed needs to be processed to obtain the data to be warned, and then the risk warning is issued using this data. Because the data to be processed is pre-processed, the data collection and risk warning processes can be decoupled.

[0049] For example, when a new data source needs to be imported, there is no need to modify the risk warning process. Instead, the data from the new data source is directly added to the data to be processed, which solves the problem that adapting to a new data source requires a complete system reconstruction in related technologies.

[0050] Step S202: The data to be processed is converted into text to obtain the data to be warned. The data to be warned includes the corresponding first field and the value of the first field.

[0051] Since the data to be processed may come from different data sources, the data may be presented in text form or other forms, such as voice or images. Based on this, by text-based processing, the data to be processed in different forms can be transformed into text form that is easily recognized by the early warning model. The transformed data to be processed is the data to be warned.

[0052] For the data to be processed, a first field and its corresponding value can be generated based on the content of the data. The first field can include multiple fields to identify the attributes of the data to be alerted. For example, the first field can include fields for the user who made the statement, such as name, age, and occupation, and fields for text information, such as whether there was risky speech and the frequency of risky speech. The first field value is the value corresponding to the first field in the data to be alerted. For example, when the first field is "whether there was risky speech," its corresponding first field value can be a Boolean value, i.e., 0 or 1; when the first field is "age," the first field value can be an integer value, such as 30.

[0053] By text-izing the data to be processed, we can obtain the data to be warned, including the first field and the corresponding value of the first field. This makes it easier to filter out the data that needs to be warned from massive amounts of data, thus improving the efficiency of warning.

[0054] Step S203: Obtain the warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object. The warning field information includes the warning field name.

[0055] The warning field is the field that needs to be judged during this risk warning process. That is, which fields in the data to be warned need to be judged and warned. The warning field information includes the name of the database entity where the warning field data source is located, the name of the table entity where it is located, the warning field name, and the warning field description. The warning rule is the rule for issuing a warning, that is, under what circumstances a warning is required. In this embodiment, the warning rule is matched with the warning field. For example, if the warning field is A, the warning rule can be A>10, etc. When issuing a risk warning, the system can receive the warning field information (which can come from one or more fields in one or more entity tables of one or more entity databases) and the warning rule corresponding to the warning field input by the risk warning staff. For example, if the risk warning staff needs to issue a warning for data that is "risky to speak" and has a "risky speaking frequency greater than 10", the user-input warning field information can be: [ { 'is_threaten':{ 'Table name':'_monitor_data', 'Field name':'is_threaten', 'Field Description': 'Whether there is any risky speech' }, }, { 'threaten_frequency':{ 'Table name':'_statistics_data', 'Field name':'threaten_frequency', 'Field Description': 'Frequency of Risky Speech' } ] ”; The 'field description' is used to address the issue of inaccurate data field matching when the large model matches user natural language input due to similar field names. Additionally, the warning rule corresponding to the warning field needs to be entered, specifying the conditions under which a warning should be issued. Taking the example from the previous section where risk warning staff need to issue warnings for data categorized as "risky speech" and "risky speech frequency greater than 10," the warning fields are "whether there is risky speech" and "risky speech frequency." In this case, the warning rule entered by the user could be: [ { 'is_threaten':{ 'Field name':'is_threaten', 'Field value': 'yes' }, }, { 'threaten_frequency':{ 'Field name':'threaten_frequency', 'Field value': '>10' } ] ”; In this embodiment, to enable risk warning staff to quickly set the association between warning field information and warning rules, the user-input data structure is designed in JSON format with field names as keys. The system finds the warning-related fields "is_threaten" and "threaten_frequency" in multiple entity tables "_monitor_data" and "_statistics_data" based on the user-input warning field information, and obtains the data to be identified through a filtering module. Next, the system determines the corresponding warning field names based on the user-input warning rules, namely "is_threaten" and "threaten_frequency" in the warning rules, and the field values ​​are "yes" and ">10" respectively, obtaining the warning result through a large model.

[0056] In this embodiment, designing the user-input field information and warning rules into a JSON format that is easy for the large model to process can improve the accuracy of the large model's data processing. This design can also be replaced with the simplest natural language input to the large model, provided that the expression is clear. For example, risk warning personnel can input "Issue a warning for data that has been threatened and has a risk speech frequency greater than 10. Here, 'whether there is a threat' requires checking if the value of the field 'is_threaten' in the '_monitor_data' table in the system is yes, and 'risk speech frequency greater than 10' requires checking if the value of the field 'threaten_frequency' in the '_statistics_data' table in the system is >10. If both are met, issue a warning." Regardless of the implementation method, after receiving the data source of the warning field information input by the user, this invention will call the filtering module to find the corresponding data as the data to be identified and hand it over to the big language model. The big language model will then combine the data to be identified with the warning rules input by the user for analysis.

[0057] By having risk warning staff input warning field information and warning rules, they can more easily control specific warning scenarios. When the warning scenarios need to be expanded, they can input the corresponding warning field information and warning rules as needed, simplifying the operation and eliminating the need to refactor the system.

[0058] Step S204: Based on the warning field information and the first field, filter the warning data to be identified.

[0059] After obtaining the warning field information, the data to be warned can be filtered based on the first field corresponding to the data to be warned. The data to be warned can come from multiple table entities and databases. Here, we take an example where only one field in the same table of the same database is a warning-related field. That is, from the data to be warned, all fields related to the warning that match the warning field information description in the first field are selected as the data to be identified. For example, if the warning field information is "Table name: '_monitor_data', field name: 'is_threaten' (i.e., whether there is a risk of speaking)," then the fields in the data to be warned that match the first field are selected as the data to be identified. When there are multiple warning fields, all fields in the first field that match the warning field information can be selected as the data to be identified. By pre-filtering, the number of data to be identified can be reduced, ensuring that only the data that truly needs to be warned (i.e., the data to be identified) is input into the warning model, reducing the resource consumption of the warning model and improving warning efficiency.

[0060] Step S205: Generate warning prompt words based on the warning field name, warning rules and data to be identified, and input the warning prompt words into the warning model to obtain the risk warning result of the first detection object.

[0061] After filtering out the data to be identified, the warning field information, warning rules, and the input to be identified can be filled into the warning prompt words to improve them. The warning prompt words can be a phrase with variables. By filling in the variables, warning prompt words can be generated for input into the warning model, allowing the model to issue risk warnings based on the prompt words and generate risk warning results. In this embodiment, the template for the warning prompt words can be: Role: You are an early warning assistant who needs to determine whether an early warning needs to be issued based on given data and judgment logic.

[0062] Alert field name: $Param1; Warning rule: $Param2; Warning sample: $Param3; Question: Should the "$ data to be identified" receive an alert? Please provide a reason.

[0063] In this template, `$Param1` is used to enter the 'alert field name', corresponding to the 'field name' in the alert field information. For example, it can be the field name entered by the user in step S203. For instance, `$Param1` could be `is_threaten`, `threaten_frequency`, etc., used to quickly associate the field data found in the numerous / single specified database / table of data to be alerted based on the user-input field information (such as the database / table / field information described in this case), thereby determining the matching object of the alert rule. `$Param2` is used to enter the alert rule in the template. For example, it can be the alert rule entered by the user in step S203. For instance, `$Param2` can be an alert rule in JSON (JavaScript Object Notation) format as described above, or it can be in natural language form, such as "is_threaten='yes', and threshold_frequency>10". `$Data to be identified` is used to enter the data to be identified in the template. $Param3 is the warning sample, used to provide warning samples for the large language model. This sample is the result of the large language model's analysis of all content of the PROMPT (warning prompt word) except $Param3, such as " [{ Warning field name: is_threaten Warning rule: is_threaten = yes question:"

[0064] "Is an early warning necessary? Give your reasons."

[0065] Final assessment result: An alert is needed because, according to the given data item, the value of the `is_threaten` field is "yes", which meets the alert rule `is_threaten = yes`, so an alert is required.

[0066] }]”; As the analysis accumulates, the data is recorded in the early warning sample database. Early warning samples can be obtained using a short SQL statement, as shown below.

[0067] “ SELECT FROM 'early warning sample' ORDER BY create_time DESC Limit 10 ”; When there is no data in the sample database, the statement returns no content, and the value of $Param3 can be [] (i.e., empty).

[0068] After constructing the warning prompt words, the warning prompt words can be input into the warning model. Based on the warning model, the data to be identified is judged to determine the risk warning result for the first detection object.

[0069] By separating the warning field information, warning rules, and data to be identified from the warning model, the warning model can be directly replaced or updated when upgrading the model version without affecting the processing of the data to be identified. This eliminates the need to reconstruct the system and solves the problem of long model adaptation cycles in related technologies.

[0070] The risk warning method based on a large model provided in this embodiment decouples the data collection and risk warning processes by integrating data from multiple data sources into a single data set to be processed after the data is generated from the data source. This means that when a new data source needs to be imported, the risk warning process does not need to be modified; instead, the data from the new data source is directly added to the data set to be processed, solving the problem of requiring a complete system reconstruction to adapt to new data sources in related technologies. By text-based processing of the data to be processed, obtaining the data to be warned, including the first field and its corresponding value, it is easier to filter data requiring risk warning from massive amounts of data, improving warning efficiency. The method allows risk warning staff to input warning field information and warning rules, making it easier for them to control specific warning scenarios. When the warning scenario needs to be expanded, they can input the corresponding warning field information and warning rules as needed, simplifying operations without requiring system reconstruction. Pre-screening reduces the amount of data to be warned, ensuring that only the data truly requiring warning (i.e., the data to be identified) is input into the warning model, reducing the resource consumption of the warning model and improving warning efficiency. By separating the warning field information, warning rules, and data to be identified from the warning model, the warning model can be directly replaced or updated when upgrading the model version without affecting the processing of the data to be identified. This eliminates the need to reconstruct the system and solves the problem of long model adaptation cycles in related technologies.

[0071] This embodiment provides a risk warning method based on a large model, which can be used in the aforementioned terminal devices, such as desktop computers, mobile phones, and tablet computers. Figure 3 This is a flowchart of a risk warning method based on a large model according to an embodiment of this application, such as... Figure 3 As shown, the process includes the following steps: Step S301: Obtain the data to be processed. The data to be processed comes from the first detection object. The first detection object has multiple data sources, which are used to generate the data to be processed.

[0072] Please see details Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0073] Step S302: Perform text processing on the data to be processed to obtain the data to be warned. The data to be warned includes the corresponding first field and the value of the first field.

[0074] Specifically, step S302, "to process the data to be processed into text to obtain the data to be warned", includes steps S3021 and S3022.

[0075] Step S3021: Perform text processing on the data to be processed and determine the text-processed data to be processed.

[0076] Step S3022: Extract the text information from the text-based data to be processed, obtain the corresponding first field and the first field value, and generate the warning data corresponding to the data to be processed.

[0077] Because the data to be processed may take various forms, such as voice or images, computers or early warning models may not be able to handle all these formats. Therefore, the data can be digitized into text, transforming it into a format easily recognizable by computers or early warning models. This allows for the determination of the first field and its value, and the generation of risk warnings. After converting the data into text, the textual information can be extracted to determine the corresponding first field and its value. The textual information is the content of the data, which can be extracted using text processing programs, such as large language models, to extract keywords. After keyword extraction, the first field and its value corresponding to the extracted keywords can be generated. Once the first field and its value are generated, the corresponding warning data can be determined.

[0078] In some optional implementations, step S3021, "to process the data to be processed into text and determine the text-processed data to be processed", includes step a1 or step a2.

[0079] Step a1: The data to be processed is processed into text using a preset logical algorithm to determine the text-processed data to be processed.

[0080] Step a2: Analyze the data to be processed using a pre-set large language model, extract the text information from the data to be processed, and determine the text-based data to be processed.

[0081] During text processing, pre-defined text processing algorithms can be used to convert data of different formats into text data. For example, Python batch processing can be used to convert key information in spectrophotometer data, signaling data, and other data to be processed into text information, i.e., to textify the data to be processed, thereby determining the text-based data to be processed.

[0082] In addition, data processing methods such as artificial intelligence can be used to process the data into text. For example, a large language model can be used to analyze the data and determine the text-based data to be processed.

[0083] Step S303: Obtain the warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object.

[0084] Please see details Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0085] Step S304: Based on the warning field information and the first field, filter the warning data to be identified.

[0086] Please see details Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0087] In some optional implementations, step S304, "filtering the data to be warned based on the warning field information and the first field, and determining the data to be identified," includes steps b1 and b2.

[0088] Step b1: Generate a filter statement based on the warning field information. The filter statement includes the warning field information.

[0089] Step b2: Filter the data to be warned using filtering statements to obtain the data to be identified.

[0090] After determining the warning field information, since the data to be warned may be stored in multiple databases corresponding to multiple data sources, a corresponding filtering statement can be generated based on the warning field information to filter the data from multiple data sources. This filtering statement can be a program or script used to filter the data corresponding to the warning field, thus obtaining the data to be identified. In this embodiment, the filtering statement can be generated based on the warning field information using SQL statements or other query languages, such as SQL (Structured Query Language), to select the warning field from numerous first fields in multiple databases and tables as the data to be identified. By using multi-data source filtering, multi-dimensional data can be used for comprehensive warnings, improving warning accuracy. Furthermore, it reduces the problem of large data redundancy caused by using all first fields as warning fields, reduces interference from irrelevant fields in warning judgment, and improves the accuracy of risk warnings.

[0091] Step S305: Generate warning prompt words based on the warning field name, warning rules and data to be identified, and input the warning prompt words into the warning model to obtain the risk warning result of the first detection object.

[0092] Please see details Figure 2 Step S205 of the illustrated embodiment will not be described again here.

[0093] In some optional implementations, after step S305, the risk warning method based on the large model further includes steps c1 and c2.

[0094] Step c1: Obtain feedback information regarding the risk warning result, update the risk warning result based on the feedback information, and determine the updated risk warning result.

[0095] Step c2: The data to be identified and the updated risk warning results are used as warning samples, and the warning samples are added to the warning sample library. The warning samples in the warning sample library are used to assist the warning model in determining the risk warning results.

[0096] After the early warning model generates a risk warning result for the first detected object, manual judgment can be made. This means that risk warning personnel can provide feedback on the risk warning result and determine the appropriate feedback information. The feedback information is the risk warning personnel's judgment of the result; for example, it could be "This risk warning result is correct" or "This risk warning result is incorrect." For correct large-scale model judgments, the large-scale model judgment is used as the final warning result. For partially correct large-scale model judgments, it is manually corrected and used as the final warning result. For incorrect large-scale model judgments, more information needs to be provided, directly providing the correct result and explanation according to the logic output by the large-scale model, replacing the risk warning result given by the large-scale model, to obtain the final risk warning result. After determining the feedback information for the risk warning result and obtaining the final warning result, a warning sample can be generated using the prompt words fed to the large-scale model (excluding the "warning sample" parameter content) and the final risk warning result, and then added to the warning sample library. The warning samples are used to assist the warning model in determining the risk warning result during the subsequent risk warning process. For example, when generating a risk warning result, the warning model can use one or more warning samples as examples to perform warning analysis on the prompt words to be analyzed (including warning fields, corresponding rules for warning fields, and data to be identified) excluding the content of the "warning sample". The warning sample library is used to store warning samples so that when the warning model needs warning samples for assistance, it can directly call the warning samples from the warning sample library.

[0097] In some optional implementations, when performing steps c1 and c2, step S305 "generates corresponding warning prompt words based on the warning field name, warning rules and data to be identified", including steps d1 and d2.

[0098] Step d1: Generate the corresponding initial warning prompt words based on the warning field name, warning rules, and data to be identified.

[0099] Step d2: Add at least one warning sample from the warning sample library to the initial warning prompt word to generate the corresponding warning prompt word.

[0100] The process of generating initial warning prompts based on the warning field name, warning rules, and data to be identified is similar to step S205, and will not be repeated here. At this point, the prompts generated based on the warning field name, warning rules, and data to be identified can be used as initial warning prompts, and warning samples can be added as aids to help the warning model generate more accurate risk warning results. For example, the initial warning prompts could be: Role: You are an early warning assistant who needs to determine whether an early warning needs to be issued based on given data and judgment logic.

[0101] Alert field name: $Param1 Warning rule: $Param2 Warning sample: [] Question: Should the "$ data to be identified" receive an alert? Please provide a reason.

[0102] In this template, $Param1 is the location where the alert field name is entered. $Param2 is the location where the alert rule is entered. $Data to be identified is the location where the data to be identified is entered.

[0103] Based on the initial warning message, when the warning sample library is not empty, the obtained sample data will also no longer be empty. At this time, the warning message can be: Role: You are an early warning assistant who needs to determine whether an early warning needs to be issued based on given data and judgment logic.

[0104] Alert field name: $Param1; Warning rule: $Param2; Warning sample: [ { Warning field name: is_threaten Warning rule: is_threaten = yes question:"

[0105] "Is an early warning necessary? Give your reasons."

[0106] Final assessment result: An alert is needed because, according to the given data item, the value of the `is_threaten` field is "yes", which meets the alert rule `is_threaten = yes`, so an alert is required.

[0107] } ] Question: Should the "$ data to be identified" receive an alert? Please provide a reason.

[0108] At this point, the statement is obtained through the warning sample, such as "SELECT The command `FROM 'Warning Sample Library' ORDER BY create_time DESC Limit 10` selects 10 warning samples from the warning sample library in chronological order. Only one sample is displayed here, indicating that only one data point is currently available in the warning sample library. It's understandable that if there are no warning samples in the library at this time, the data item extracted by this statement will be empty, meaning that the warning sample can be empty and will not affect the warning model's generation of the corresponding risk warning result.

[0109] In some alternative implementations, the risk warning method based on large models also includes steps e1 to e3.

[0110] Step e1: Obtain the number of warning samples in the warning sample library.

[0111] Step e2: If the number of samples exceeds a preset threshold, add the warning samples from the warning sample library to the warning training sample library and delete the warning samples from the warning sample library.

[0112] Step e2: Fine-tune the early warning model using the early warning training sample library to obtain the fine-tuned early warning model, and use the fine-tuned early warning model as the early warning model for risk warning.

[0113] Each time an early warning sample is added to the early warning sample library, the number of early warning samples in the library can be determined. If the number of early warning samples exceeds a preset threshold, all early warning samples in the early warning sample library can be added to the early warning training sample library. The preset threshold can be a value pre-entered by risk warning personnel; for example, a preset threshold of 2000. When the current number of early warning samples exceeds 2000, the early warning samples in the early warning sample library can be transferred to the early warning training sample library. The early warning training sample library is then used to fine-tune the early warning model. For example, the early warning model can learn from the early warning samples in the early warning training sample library, so that the fine-tuned early warning model can generate risk warning results more accurately. Since the early warning samples have been transferred from the early warning sample library to the early warning training sample library, and the early warning model has been fine-tuned based on the early warning training sample library, the auxiliary functions of the early warning sample library for the early warning model are weakened. At this point, early warning samples can be deleted from the early warning sample library. By continuously fine-tuning the early warning model, it can be made more suitable for the early warning needs of risk warning personnel. Subsequent early warning analysis can then be based on the fine-tuned model, improving its learning ability and making the risk warning results more accurate.

[0114] Figure 4 This is a risk warning flowchart according to an embodiment of this application. Figure 4 As shown, the system first processes the data, including a data acquisition module to collect data from multiple data sources, and a preprocessing module to convert the data into textual data, resulting in structured data for early warning. This data includes a corresponding first field and its value. Users can customize warning field information and warning rules. Based on the warning field information, the system selects data to be identified, generating data to be recognized. Warning prompts are generated based on the field names, warning rules, data to be recognized, and warning samples in the warning field information and input into the warning model. After the warning model outputs the risk warning result, risk warning personnel can provide feedback to obtain the final warning result. Corresponding warning samples are then generated and added to the warning sample library. If the number of warning samples in the warning sample library exceeds 2000, the warning samples are transferred to the warning training sample library, and the warning model is fine-tuned based on the training sample library.

[0115] The risk warning method based on a large model provided in this embodiment decouples the data collection and risk warning processes by integrating data from multiple data sources into a single data set to be processed after the data is generated from the data source. This means that when a new data source needs to be imported, the risk warning process does not need to be modified; instead, the data from the new data source is directly added to the data set to be processed, solving the problem of requiring a complete system reconstruction to adapt to new data sources in related technologies. By text-based processing of the data to be processed, the system obtains warning data including the first field and its corresponding value, making it easier to filter data fields related to risk warning from massive amounts of data, thus improving warning efficiency. By allowing risk warning staff to input warning field information and warning rules, they can more easily control specific warning scenarios. When the warning scenario needs to be expanded, they can input the corresponding warning field information and warning rules as needed, simplifying operations and eliminating the need for system reconstruction. Pre-screening reduces the amount of data to be read, ensuring that only the data fields that truly require warning (i.e., the data to be identified) are input into the warning model, reducing the resource consumption of the warning model and improving warning efficiency. By separating the warning field information, warning rules, and data to be identified from the warning model, the warning model can be directly replaced or updated when upgrading the model version without affecting the processing of the data to be identified, and without requiring system reconstruction. Continuous optimization through fine-tuning of the warning model makes it better suited to the warning needs of risk warning personnel. Subsequent warning analysis can then be based on the fine-tuned warning model, improving its learning ability and making risk warning results more accurate.

[0116] This embodiment also provides a risk warning device based on a large model, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0117] This embodiment provides a risk warning device based on a large model, such as Figure 5 As shown, it includes: The data acquisition module 501 is used to acquire data to be processed. The data to be processed comes from the first detection object. The first detection object has multiple data sources, which are used to generate the data to be processed.

[0118] The text processing module 502 is used to perform text processing on the data to be processed to obtain the data to be warned. The data to be warned includes the corresponding first field and the value of the first field.

[0119] The field determination module 503 is used to obtain the warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object. The warning field information includes the warning field name.

[0120] The filtering module 504 is used to filter the data to be warned based on the warning field information and the first field to determine the data to be identified.

[0121] The prompt word generation module 505 is used to generate warning prompt words based on the warning field name, warning rules and data to be identified, and input the warning prompt words into the warning model to obtain the risk warning result of the first detection object.

[0122] In some alternative implementations, the text processing module 502 includes: The textification submodule is used to process the data to be processed into text and to determine the textified data to be processed.

[0123] The information extraction submodule is used to extract text information from the text-based data to be processed, obtain the corresponding first field and the first field value, and generate the warning data corresponding to the data to be processed.

[0124] In some alternative implementations, the textification submodule includes: The logic algorithm unit is used to process the data to be processed into text using a preset logic algorithm, and to determine the text-processed data to be processed.

[0125] or, The language model unit is used to analyze the data to be processed using a pre-defined large language model, extract textual information from the data to be processed, and determine the textualized data to be processed.

[0126] In some alternative implementations, the filtering module 504 includes: The filter statement submodule is used to generate filter statements based on the warning field information. The filter statements include the warning field information.

[0127] The data filtering submodule is used to filter the data to be warned using filtering statements to obtain the data to be identified.

[0128] In some alternative implementations, the risk warning device based on the large model also includes: The feedback acquisition module is used to acquire feedback information on the risk warning results, update the risk warning results based on the feedback information, and determine the updated risk warning results.

[0129] The early warning sample module is used to take the data to be identified and the updated risk warning results as early warning samples, and add the early warning samples to the early warning sample library. The early warning samples in the early warning sample library are used to assist the early warning model in determining the risk warning results.

[0130] In some alternative implementations, the prompt word generation module 505 includes: The initial alert word submodule is used to generate corresponding initial alert alert words based on the alert field name, alert rules, and data to be identified.

[0131] The prompt word improvement submodule is used to add at least one warning sample from the warning sample library to the initial warning prompt word to generate the corresponding warning prompt word.

[0132] In some alternative implementations, the risk warning device based on the large model also includes: The sample quantity module is used to obtain the sample quantity of warning samples in the warning sample library.

[0133] The sample transfer module is used to add warning samples from the warning sample library to the warning training sample library and delete warning samples from the warning sample library when the number of samples exceeds a preset threshold.

[0134] The model fine-tuning module is used to fine-tune the early warning model using the early warning training sample library, obtain the fine-tuned early warning model, and use the fine-tuned early warning model as the early warning model when issuing risk warnings.

[0135] The risk warning device based on a large model provided in this application can execute the risk warning method based on a large model provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0136] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0137] The following is a detailed reference. Figure 6 This diagram illustrates a suitable structural schematic for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 601, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 602 or a program loaded from memory 608 into random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the electronic device. The processor 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0138] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0139] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a memory 608, or installed from a ROM 602. When the computer program is executed by the processor 601, it performs the functions defined in the risk warning method based on a large model according to embodiments of this application.

[0140] Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0141] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the risk warning method based on a large model shown in the above embodiments is implemented.

[0142] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0143] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A risk warning method based on a large model, characterized in that, The method includes: Acquire data to be processed, the data to be processed comes from a first detection object, the first detection object has multiple data sources, and the data sources are used to generate the data to be processed; The data to be processed is converted into text to obtain the data to be warned, which includes a corresponding first field and the value of the first field. Obtain warning field information and the warning rule corresponding to the warning field. The warning field information and the warning rule correspond to the first detection object. The warning field information includes the warning field name. Based on the warning field information and the first field, the data to be warned is filtered to determine the data to be identified; Based on the warning field name, the warning rule, and the data to be identified, a warning prompt word is generated, and the warning prompt word is input into the warning model to obtain the risk warning result of the first detection object.

2. The method according to claim 1, characterized in that, The step of text-based processing of the data to be processed to obtain the data to be warned includes: The data to be processed is converted into text to determine the text-converted data to be processed. Extract the text information from the text-based data to be processed to obtain the corresponding first field and the first field value, and generate the warning data corresponding to the data to be processed.

3. The method according to claim 2, characterized in that, The step of text-ifying the data to be processed and determining the text-ified data to be processed includes: The data to be processed is processed into text using a preset logical algorithm to determine the text-processed data to be processed. or, The data to be processed is analyzed by a pre-set large language model, the text information in the data to be processed is extracted, and the text-based data to be processed is determined.

4. The method according to claim 1, characterized in that, The step of filtering the data to be warned based on the warning field information and the first field to determine the data to be identified includes: A filtering statement is generated based on the warning field information; the filtering statement includes the warning field information. The data to be warned is filtered using the filtering statement to obtain the data to be identified.

5. The method according to claim 1, characterized in that, After obtaining the risk warning result for the first detected object, the method further includes: Obtain feedback information regarding the risk warning result, update the risk warning result based on the feedback information, and determine the updated risk warning result; The data to be identified and the updated risk warning results are used as warning samples, and the warning samples are added to the warning sample library. The warning samples in the warning sample library are used to assist the warning model in determining the risk warning results.

6. The method according to claim 5, characterized in that, The step of generating warning prompt words based on the warning field name, the warning rule, and the data to be identified includes: Generate corresponding initial warning prompt words based on the warning field name, the warning rule, and the data to be identified; At least one warning sample from the warning sample library is added to the initial warning prompt word to generate the corresponding warning prompt word.

7. The method according to claim 5, characterized in that, The method further includes: Obtain the number of early warning samples in the early warning sample library; If the number of samples exceeds a preset threshold, add the warning samples from the warning sample library to the warning training sample library and delete the warning samples from the warning sample library. The early warning model is fine-tuned using the early warning training sample library to obtain the fine-tuned early warning model, which is then used as the early warning model for risk warning.

8. A risk warning device based on a large model, characterized in that, The device includes: The data acquisition module is used to acquire data to be processed, which comes from a first detection object. The first detection object has multiple data sources, and the data sources are used to generate the data to be processed. The text processing module is used to perform text processing on the data to be processed to obtain the data to be warned, wherein the data to be warned includes a corresponding first field and the value of the first field; The field determination module is used to obtain warning field information and warning rules corresponding to the warning fields. The warning field information and the warning rules correspond to the first detection object. The warning field information includes the warning field name. The filtering module is used to filter the data to be warned based on the warning field information and the first field to determine the data to be identified; The prompt word generation module is used to generate a warning prompt word based on the warning field name, the warning rule and the data to be identified, and input the warning prompt word into the warning model to obtain the risk warning result of the first detection object.

9. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the risk warning method based on a large model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the risk warning method based on a large model as described in any one of claims 1 to 7.

Images