Service early warning signal generation method and device based on large model, equipment and medium

Abstract

The invention relates to the technical field of data analysis, and discloses a service early warning signal generation method and device based on a large model, equipment and a medium, and the method comprises the steps: generating a target service early warning rule through a service early warning model and a standard heterogeneous data set; and generating a target service early warning signal according to the service early warning rule and the standard heterogeneous data set. By means of the mode, the multi-source heterogeneous data are integrated and preprocessed, a traditional analysis mode of isolated data sources is changed, all the data are converted into the high-quality and standardized data set, and the breadth and depth of data utilization are improved. And performing attribution analysis by using the logical reasoning ability of the large model, identifying a complex feature combination and a risk mode, and generating a service early warning signal in time. The method can be applied to the business fields of financial science and technology, medical health, old-age care and the like, and the reliability of risk identification system risk identification and timely early warning response making is improved.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular to a method, apparatus, device and medium for generating service early warning signals based on a large model. Background Technology

[0002] With the deepening of digital transformation, big data and artificial intelligence technologies are being applied more and more widely in key service areas such as fintech, healthcare, and elderly care. These areas share an urgent need to shift from "passive response" to "proactive early warning" and "intelligent intervention." However, existing technological solutions have significant shortcomings in multi-source data fusion, intelligent analysis and decision-making, and real-time response efficiency, which seriously restricts the improvement of service quality.

[0003] In the fintech sector, particularly in commercial property and casualty insurance and auto insurance services, risk management and customer service are paramount. Currently, most business scenarios still heavily rely on manual screening based on fixed rules or simple statistical models to identify problems. For example, the identification of customer churn risk and claims fraud risk is often based on offline reports and preset static threshold rules. This approach is not only time-consuming and labor-intensive, but also lacks rigid rules, making it difficult to track and respond to market changes and new fraudster tactics in real time. Existing systems lack the ability to deeply understand and correlate multi-source heterogeneous data, failing to effectively uncover potential risk clues, leading to delayed or missed warnings, causing companies to miss the best intervention opportunity, increasing operational risks and customer churn rates.

[0004] In the fields of healthcare and elderly care, there is a significant demand for health risk warnings, abnormal event predictions, and the rational allocation of elderly care resources for patients with chronic diseases. Existing systems largely rely on isolated health monitoring devices and regular medical examination reports. Data analysis is often limited to single dimensions or simple threshold alarms, failing to provide a comprehensive and continuous assessment of the health status of the elderly. For example, the system cannot identify early signs of health deterioration from multiple sources of information such as customer service call records, daily behavior data, and medication adherence. This type of warning has low accuracy and a high false alarm rate, leading to "alarm fatigue." Therefore, in the fields of fintech, healthcare, and elderly care, improving the reliability of risk identification systems in identifying risks and providing timely warning responses has become a pressing technical problem that needs to be solved. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for generating service early warning signals based on a large model, so as to improve the reliability of risk identification systems in identifying risks and making timely early warning responses.

[0006] Firstly, this application provides a method for generating service early warning signals based on a large model, the method comprising: Acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; Target service early warning rules are generated using a pre-trained service early warning model and the aforementioned standard heterogeneous dataset. Based on the service warning rules and the standard heterogeneous dataset, a target service warning signal is generated.

[0007] Secondly, this application also provides a service early warning signal generation device based on a large model, the device comprising: A standard heterogeneous dataset generation module is used to acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset. The target service early warning rule generation module is used to generate target service early warning rules through a pre-trained service early warning model and the standard heterogeneous dataset. The target service early warning signal generation module is used to generate target service early warning signals based on the service early warning rules and the standard heterogeneous dataset.

[0008] Thirdly, this application also provides a computer device, the computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement the service early warning signal generation method based on the large model as described above.

[0009] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the service warning signal generation method based on a large model as described above.

[0010] This application discloses a method, apparatus, device, and medium for generating service early warning signals based on a large model. The method includes acquiring initial multi-source heterogeneous data, preprocessing the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; generating target service early warning rules using a pre-trained service early warning model and the standard heterogeneous dataset; and generating target service early warning signals based on the service early warning rules and the standard heterogeneous dataset. Through this approach, this application, by integrating and preprocessing multi-source heterogeneous data, changes the traditional analysis model that relies on a single, isolated data source, transforming various types of structured, unstructured, and real-time streaming data into high-quality, standardized datasets, thereby improving the breadth and depth of data utilization. Utilizing the logical reasoning capabilities of a large model for attribution analysis, it identifies complex feature combinations and risk patterns and generates service early warning signals in a timely manner. In business areas such as fintech, healthcare, and elderly care, this improves the reliability of risk identification systems in identifying risks and issuing timely early warning responses. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments 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 based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of a service early warning signal generation method based on a large model provided in an embodiment of this application; Figure 2 A schematic block diagram of a service early warning signal generation device based on a large model, provided for embodiments of this application; Figure 3 A schematic block diagram of the structure of a computer device provided for an embodiment of this application. Detailed Implementation

[0013] 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.

[0014] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0015] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0016] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0017] This application provides a method, apparatus, device, and medium for generating service early warning signals based on a large model. The method can be applied to risk identification systems. By integrating and preprocessing multi-source heterogeneous data, it changes the traditional analysis model that relies on single, isolated data sources, transforming various structured, unstructured, and real-time streaming data into high-quality, standardized datasets, thus improving the breadth and depth of data utilization. Utilizing the logical reasoning capabilities of large models for attribution analysis, it identifies complex feature combinations and risk patterns and generates timely service early warning signals. In business areas such as fintech, healthcare, and elderly care, this improves the reliability of risk identification systems in identifying risks and issuing timely early warning responses.

[0018] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0019] Please see Figure 1 , Figure 1 This is a schematic flowchart illustrating a service early warning signal generation method based on a large model, provided in an embodiment of this application. This method can be applied to risk identification systems to improve the reliability of risk identification systems in identifying risks and issuing timely early warning responses.

[0020] like Figure 1 As shown, the service early warning signal generation method based on the large model specifically includes steps S10 to S30.

[0021] Step S10: Obtain initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; Specifically, in the fintech business field, taking the auto insurance scenario as an example: Structured data such as insurance records and claims status are extracted from the core business database via application programming interfaces (APIs); external structured data such as vehicle credit information and traffic violations are obtained by connecting to third-party data platforms through software development kits (SDKs); unstructured text data such as customer service call transcripts and social media comments are collected in real time through log collectors; and time-series data streams such as vehicle operating status uploaded by in-vehicle devices are received through IoT protocols.

[0022] An isolated forest algorithm was used to detect abnormal quotes in insurance policies, and a BERT model based on an attention mechanism was used to context-awarely fill in the missing customer occupation field. In the data augmentation stage, composite features were constructed based on domain knowledge, including calculating risk indicators, and oversampling was performed on imbalanced claims data. Through entity alignment and feature normalization, a high-quality standard heterogeneous dataset was generated.

[0023] In the field of medical and health care and elderly care business, initial multi-source heterogeneous data, including electronic medical records, real-time physiological parameters, medical images, nursing record texts and insurance settlement data, are collected through multiple channels such as medical IoT devices, hospital information systems, wearable monitoring devices and elderly care platforms.

[0024] Clustering anomaly detection algorithms were employed to identify and clean anomalous values ​​(such as sudden changes in blood oxygen saturation) and contradictory information (such as mismatches between medication records and diagnostic results) in the monitoring data. Simultaneously, the BERT model was used to fill in missing patient basic information with context awareness. A medical entity recognition model was used to extract key entities such as diseases, symptoms, and medications from the text. Sentiment analysis techniques were used to analyze the emotional tendencies of patients' complaints, and a topic extraction model was used to mine high-frequency health risk topics from long-term care records. All processed multimodal data were then integrated into a standard heterogeneous dataset.

[0025] Step S20: Generate target service early warning rules using the pre-trained service early warning model and the standard heterogeneous dataset; Specifically, a standard heterogeneous dataset is input into a service early warning model that has been pre-trained twice using a corpus from the auto insurance domain. The model is then combined with an LSTM time-series analysis module to output the customer churn probability. Its logical reasoning ability is used to perform attribution analysis on historical early warning cases, automatically generating initial early warning rules such as "IF (average premium reduction in the customer's last 3 quotes > 15%) and (number of complaints ≥ 2) → high-risk churned customer". Subsequently, a genetic algorithm is used to iteratively optimize the rule set, and the rule threshold is dynamically adjusted through quantile regression forest to generate the final target service early warning rule.

[0026] Step S30: Generate a target service early warning signal based on the service early warning rules and the standard heterogeneous dataset.

[0027] Specifically, various real-time data sources (such as user online inquiry behavior, APP click stream, policy change events, etc.) are continuously written to a preset message queue. The stream processing engine consumes these data streams from the preset message queue and extracts key features (such as customer ID, event type, timestamp, etc.). At the same time, the stream processing engine task queries the customer static profile and historical features (such as the number of visits in the past 7 days, the number of historical claims) stored in the remote dictionary service cluster in real time, and associates the real-time events with the background profile at the millisecond level to form a complete real-time feature vector.

[0028] Based on the subject ID (such as customer ID) in the data, the corresponding historical profile features are read from the remote dictionary service cache in milliseconds, and concatenated with the real-time data to form a complete feature vector. This feature vector is then quickly matched with all early warning rules in memory, and a target service early warning signal is generated based on the feature vector that matches the early warning rule.

[0029] This embodiment discloses a service early warning signal generation method based on a large model. The method includes acquiring initial multi-source heterogeneous data, preprocessing the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; generating target service early warning rules using a pre-trained service early warning model and the standard heterogeneous dataset; and generating target service early warning signals based on the service early warning rules and the standard heterogeneous dataset. Through this approach, this application, by integrating and preprocessing multi-source heterogeneous data, changes the traditional analysis model that relies on a single, isolated data source, transforming various types of structured, unstructured, and real-time streaming data into high-quality, standardized datasets, thereby improving the breadth and depth of data utilization. Utilizing the logical reasoning capabilities of the large model for attribution analysis, it identifies complex feature combinations and risk patterns and generates service early warning signals in a timely manner. In business areas such as fintech, healthcare, and elderly care, this improves the reliability of risk identification systems in identifying risks and issuing timely early warning responses.

[0030] based on Figure 1 In the illustrated embodiment, step S20 includes: Obtain historical warning cases whose similarity to the standard heterogeneous dataset is less than or equal to a preset similarity threshold; The service early warning model is used to perform attribution analysis on the historical early warning cases to generate an initial early warning rule template. The initial warning rule template is iteratively optimized based on a preset genetic algorithm to generate an initial warning rule set. The time-series features and business indicator features in the standard heterogeneous dataset are extracted by a quantile regression forest model, and the threshold values ​​of each early warning indicator in the initial early warning rule set are adjusted according to the time-series features and the business indicator features to generate the target service early warning rules.

[0031] Specifically, based on the current input standard heterogeneous dataset, the cosine similarity between its feature vector and cases in the historical case library is calculated. Historical warning cases with similarity less than or equal to a preset threshold (e.g., 0.3) are automatically selected as analysis samples. These highly relevant cases are then input into the service warning model that has been fine-tuned in the domain. The model uses chain reasoning technology to conduct in-depth attribution analysis on the reasons for the success or failure of the cases and automatically outputs the initial warning rule template.

[0032] The preset genetic algorithm optimization engine is activated, the condition items in the rule template are encoded as gene fragments, the rule set is used as the fitness function, and the optimal initial warning rule set is generated through iterative operations such as selection, crossover, and mutation. The time-series features and key business indicator features in the standard heterogeneous dataset are analyzed by quantile regression forest model, and the latest thresholds of each warning indicator in the initial rule set are dynamically calculated and adjusted, so as to output the final target service warning rule with both accuracy and environmental adaptability.

[0033] In a specific embodiment, the service early warning model is used to perform attribution analysis on the historical early warning cases to generate an initial early warning rule template, including: The key warning factors and warning decision logic of the historical warning cases are extracted using the natural language processing algorithm built into the service warning model. The service early warning model maps the key early warning factors into rule condition variables, and the service early warning model maps the early warning decision logic into a rule logic structure. The initial warning rule template is generated based on the rule condition variables and the rule logic structure.

[0034] Specifically, the service early warning model utilizes built-in natural language processing algorithms to deeply analyze historical early warning cases. For example, the semantic understanding module extracts key entities (such as "premium reduction" and "complaint frequency") and logical connectives (such as "due to" and "caused by") from the case text. A multi-head attention mechanism identifies the core factors and logical relationships influencing early warning decisions, constructing a structured knowledge graph of "key early warning factors - early warning decision logic." The identified key early warning factors (such as "increased frequency of recent premium inquiries by customers") are mapped to quantifiable rule condition variables (such as "number of price inquiries in the last 7 days") using named entity recognition and classification modules. Simultaneously, a sequence-to-sequence model transforms the decision logic described in the text (such as "marking a customer as a churn risk if they compare prices multiple times and have a history of complaints") into a standard "IF-THEN" rule logic structure. Based on a predefined rule template architecture, the standardized rule condition variables and rule logic structures are automatically combined to generate initial early warning rule templates containing configurable parameter placeholders. Each template is then assigned a confidence score and applicable scenario annotations, generating a rule template library.

[0035] based on Figure 1 In the illustrated embodiment, step S30 includes: The standard heterogeneous dataset and the service early warning rules are matched to generate events to be warned; The warning level of the event to be warned is determined according to the service warning rules, and the warning signal of the target service is generated according to the warning level.

[0036] Specifically, based on the real-time data stream from the heterogeneous dataset of the continuous consumption standard, the rule matching module matches each data record (such as real-time customer inquiry behavior and data uploaded by the vehicle self-diagnosis system) with a pre-loaded service warning rule library. When the data features meet the Boolean logic conditions set in the rules (such as "average premium reduction of the last 3 quotes > 15% AND number of complaints in the current month ≥ 2"), the system immediately generates a structured event to be warned, which includes metadata such as event identifier, trigger rule ID, associated customer ID, and trigger timestamp. Subsequently, the event to be warned is processed by the hierarchical decision module: based on the preset level threshold matrix in the trigger rules, combined with the severity of real-time calculated business indicators (such as customer value score and risk index), the warning level is automatically determined by a multi-level decision tree algorithm (Level 1 / Red, Level 2 / Orange, Level 3 / Yellow), and the corresponding signal generation template is called according to different levels. Finally, the target service warning signal containing complete information such as warning level, handling suggestions, and timeliness requirements is output.

[0037] For example, the original warning events can be classified according to the predefined severity of the triggering rules, typically using a three-level classification: Level 1 Warning (Red / Emergency): Corresponds to extremely high risk, such as suspected fraudulent transactions or impending loss of a key customer. The signal will be automatically encapsulated and a pre-defined application programming interface will be invoked, popping up a full-screen alert window in the relevant business personnel's customer relationship management system interface, interrupting their current operation and requiring immediate action.

[0038] Level 2 Warning (Orange / Important): This corresponds to risks that require priority attention, such as a decline in service experience or potential complaints. The signal will generate detailed alarm information and push it to the corresponding business team leader or specialist.

[0039] Level 3 Warning (Yellow / Alert): Corresponds to general risks or trends to be observed, such as a slight decline in customer satisfaction. The signal will automatically create a follow-up task and set a processing time limit (e.g., within 48 hours).

[0040] In a specific embodiment, the standard heterogeneous dataset and the service early warning rule are matched to generate events to be warned, including: Build a data processing engine and extract the current data stream from the standard heterogeneous dataset using the data processing engine; Match the current data stream with the service warning rules; The event corresponding to the current data stream that matches the service warning rule is identified as the event to be warned.

[0041] Specifically, a streaming data processing engine is built. By connecting a message queue and a feature database, the streaming data processing engine continuously extracts real-time updated current data streams from standard heterogeneous datasets. The data processing engine matches the incoming current data streams with service alert rules pre-loaded in memory in real time. The rule engine performs logical judgments on the feature variables in the data streams. When a data record simultaneously meets all the conditions set in the rule, the rule matching state is immediately triggered.

[0042] Mark the business event instance corresponding to the current data stream that successfully matches the rule as an event to be alerted, and generate a structured event object containing the event identifier, trigger time, associated rule number and complete context feature vector.

[0043] In a specific embodiment, step S30 is followed by: Based on the warning level, determine the target service warning signal distribution object, distribution channel, and distribution time interval; Based on the distribution channel and the distribution time interval, the target service warning signal is distributed to the distribution object corresponding to the target service warning signal.

[0044] Specifically, based on the determined warning level, the system automatically determines the distribution targets, channels, and time intervals for the warning signals to the target service by querying the pre-configured response strategy matrix. For red warnings, the distribution targets are set to business managers and system monitors, the distribution channels are designated as a dual channel of customer management system pop-ups and mobile SMS, and the distribution interval is set to send immediately and repeat the reminder every 10 minutes until confirmation. For orange warnings, the distribution targets are limited to the relevant customer service teams, the distribution channel uses group robots, and the distribution interval is set to send immediately once. For yellow warnings, the distribution targets are only the currently on-duty users, the distribution channel is pushed through the internal work order system, and the distribution interval is set to send immediately and escalate the reminder if it is not processed within 24 hours.

[0045] Based on the determined channel and interval strategies, the formatted warning signal content along with the processing link is automatically distributed to all specified objects by calling the API (Application Programming Interface) corresponding to each channel.

[0046] Based on any of the above embodiments, step S10 includes: Anomaly detection algorithm is used to identify anomalous data in the initial multi-source heterogeneous data, and the anomalous data is cleaned to generate valid multi-source heterogeneous data. The effective multi-source heterogeneous data is processed by entity recognition, sentiment analysis and topic extraction to generate the standard heterogeneous dataset.

[0047] Specifically, clustering algorithms are used to perform cluster analysis on numerical business data (such as policy amounts and vehicle ages). Sample points that fall outside all clusters and are more than three standard deviations away from the nearest cluster are identified as anomalous data. The Isolation Forest algorithm is used to detect anomalous patterns in text data (such as extreme emotional expressions in complaints). Identified anomalous data is processed differently based on its data type. For example, numerical outliers are replaced with the median of the same cluster, while textual outliers are corrected or removed based on contextual semantics, generating effective multi-source heterogeneous data labeled with data quality ratings.

[0048] This paper utilizes a BERT-based named entity recognition model to extract business entities such as insurance products and claims items from customer service recordings; a sentiment analysis model to calculate the sentiment polarity score of social media comments; and a topic model to extract topic distributions such as "service efficiency" and "fee disputes" from complaint content. The processed structured data is then aligned and fused with unstructured feature vectors, and a standard heterogeneous dataset is generated through feature scaling and normalization.

[0049] Please see Figure 2 , Figure 2 This application provides a schematic block diagram of a service warning signal generation device based on a large model, which is used to execute the aforementioned service warning signal generation method based on a large model. The service warning signal generation device based on a large model can be configured on a server.

[0050] like Figure 2 As shown, the service early warning signal generation device 400 based on a large model includes: The standard heterogeneous dataset generation module 410 is used to acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset. The target service early warning rule generation module 420 is used to generate target service early warning rules through a pre-trained service early warning model and the standard heterogeneous dataset. The target service early warning signal generation module 430 is used to generate a target service early warning signal based on the service early warning rules and the standard heterogeneous dataset.

[0051] Furthermore, the target service early warning rule generation module 420 includes: The historical early warning case acquisition unit is used to acquire historical early warning cases whose similarity to the standard heterogeneous dataset is less than or equal to a preset similarity threshold. The initial warning rule template generation unit is used to perform attribution analysis on the historical warning cases through the service warning model to generate an initial warning rule template. An initial warning rule set generation unit is used to iteratively optimize the initial warning rule template based on a preset genetic algorithm to generate an initial warning rule set. The target service early warning rule generation unit is used to extract time-series features and business indicator features from the standard heterogeneous dataset through a quantile regression forest model, and adjust the thresholds of each early warning indicator in the initial early warning rule set according to the time-series features and the business indicator features to generate the target service early warning rule.

[0052] Furthermore, the initial warning rule template generation unit includes: The natural language processing subunit is used to extract key warning factors and warning decision logic of the historical warning cases through the natural language processing algorithm built into the service warning model; The mapping subunit is used to map the key early warning factors into rule condition variables through the service early warning model, and to map the early warning decision logic into a rule logic structure through the service early warning model. The initial warning rule template generation subunit is used to generate the initial warning rule template based on the rule condition variables and the rule logic structure.

[0053] Furthermore, the target service early warning signal generation module 430 includes: The event to be warned generation unit is used to match the standard heterogeneous dataset with the service warning rules to generate events to be warned; The target service early warning signal generation unit is used to determine the early warning level of the event to be warned according to the service early warning rules, and generate the target service early warning signal according to the early warning level.

[0054] Furthermore, the event generation unit for the event to be warned includes: The current data stream extraction subunit is used to build a data processing engine and extract the current data stream from the standard heterogeneous dataset through the data processing engine. The matching subunit is used to match the current data stream with the service warning rules; The event to be warned subunit is used to determine the event corresponding to the current data stream that matches the service warning rule as the event to be warned.

[0055] Furthermore, the service early warning signal generation device 400 based on the large model includes: The early warning information determination module is used to determine the distribution target, distribution channel, and distribution time interval of the target service early warning signal based on the early warning level. The distribution module is used to distribute the target service warning signal to the distribution object corresponding to the target service warning signal according to the distribution channel and the distribution time interval.

[0056] Furthermore, the standard heterogeneous dataset generation module 410 includes: An effective multi-source heterogeneous data generation unit is used to identify abnormal data in the initial multi-source heterogeneous data through a preset clustering anomaly detection algorithm, and to perform data cleaning processing on the abnormal data to generate effective multi-source heterogeneous data. The standard heterogeneous dataset generation unit is used to perform entity recognition processing, sentiment analysis processing, and topic extraction processing on the effective multi-source heterogeneous data to generate the standard heterogeneous dataset.

[0057] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and modules can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0058] The aforementioned device can be implemented as a computer program, which can be used in, for example... Figure 3 It runs on the computer device shown.

[0059] Please see Figure 3 , Figure 3 This is a schematic block diagram illustrating the structure of a computer device according to an embodiment of this application. The computer device may be a server.

[0060] See Figure 3 The computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0061] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any service warning signal generation method based on a large model.

[0062] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0063] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When these computer programs are executed by the processor, the processor can execute any service warning signal generation method based on a large model.

[0064] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0065] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0066] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; Target service early warning rules are generated using a pre-trained service early warning model and the aforementioned standard heterogeneous dataset. Based on the service warning rules and the standard heterogeneous dataset, a target service warning signal is generated.

[0067] In one embodiment, target service early warning rules are generated using a pre-trained service early warning model and the standard heterogeneous dataset, for the following purposes: Obtain historical warning cases whose similarity to the standard heterogeneous dataset is less than or equal to a preset similarity threshold; The service early warning model is used to perform attribution analysis on the historical early warning cases to generate an initial early warning rule template. The initial warning rule template is iteratively optimized based on a preset genetic algorithm to generate an initial warning rule set. The time-series features and business indicator features in the standard heterogeneous dataset are extracted by a quantile regression forest model, and the threshold values ​​of each early warning indicator in the initial early warning rule set are adjusted according to the time-series features and the business indicator features to generate the target service early warning rules.

[0068] In one embodiment, the service alert model is used to perform attribution analysis on the historical alert cases to generate an initial alert rule template for the following purpose: The key warning factors and warning decision logic of the historical warning cases are extracted using the natural language processing algorithm built into the service warning model. The service early warning model maps the key early warning factors into rule condition variables, and the service early warning model maps the early warning decision logic into a rule logic structure. The initial warning rule template is generated based on the rule condition variables and the rule logic structure.

[0069] In one embodiment, a target service early warning signal is generated based on the service early warning rule and the standard heterogeneous dataset to achieve the following: The standard heterogeneous dataset and the service early warning rules are matched to generate events to be warned; The warning level of the event to be warned is determined according to the service warning rules, and the warning signal of the target service is generated according to the warning level.

[0070] In one embodiment, the standard heterogeneous dataset and the service alert rules are matched to generate alert events, which are used to achieve: Build a data processing engine and extract the current data stream from the standard heterogeneous dataset using the data processing engine; Match the current data stream with the service warning rules; The event corresponding to the current data stream that matches the service warning rule is identified as the event to be warned.

[0071] In one embodiment, after generating a target service warning signal based on the service warning rules and the standard heterogeneous dataset, it is used to achieve: Based on the warning level, determine the target service warning signal distribution object, distribution channel, and distribution time interval; Based on the distribution channel and the distribution time interval, the target service warning signal is distributed to the distribution object corresponding to the target service warning signal.

[0072] In one embodiment, the initial multi-source heterogeneous data is preprocessed to generate a standard heterogeneous dataset, which is used to achieve: Anomaly detection algorithm is used to identify anomalous data in the initial multi-source heterogeneous data, and the anomalous data is cleaned to generate valid multi-source heterogeneous data. The effective multi-source heterogeneous data is processed by entity recognition, sentiment analysis and topic extraction to generate the standard heterogeneous dataset.

[0073] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement any of the service early warning signal generation methods based on large models provided in the embodiments of this application.

[0074] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0075] The user personal information involved in this application embodiment is all authorized (with knowledge and consent) by the relevant parties or fully authorized by all parties, and the executing entity can obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with the relevant laws and regulations of the relevant countries and regions, and do not violate public order and good morals. It should be noted that if any software tools or components not belonging to this company appear in this application embodiment, they are merely illustrative examples and do not represent actual use.

[0076] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating service early warning signals based on a large model, characterized in that, include: Acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset; Target service early warning rules are generated using a pre-trained service early warning model and the aforementioned standard heterogeneous dataset. Based on the service warning rules and the standard heterogeneous dataset, a target service warning signal is generated.

2. The service early warning signal generation method based on a large model according to claim 1, characterized in that, The generation of target service early warning rules through the pre-trained service early warning model and the standard heterogeneous dataset includes: Obtain historical warning cases whose similarity to the standard heterogeneous dataset is less than or equal to a preset similarity threshold; The service early warning model is used to perform attribution analysis on the historical early warning cases to generate an initial early warning rule template. The initial warning rule template is iteratively optimized based on a preset genetic algorithm to generate an initial warning rule set. The time-series features and business indicator features in the standard heterogeneous dataset are extracted by a quantile regression forest model, and the threshold values ​​of each early warning indicator in the initial early warning rule set are adjusted according to the time-series features and the business indicator features to generate the target service early warning rules.

3. The service early warning signal generation method based on a large model according to claim 2, characterized in that, The step of performing attribution analysis on the historical early warning cases through the service early warning model to generate an initial early warning rule template includes: The key warning factors and warning decision logic of the historical warning cases are extracted using the natural language processing algorithm built into the service warning model. The service early warning model maps the key early warning factors into rule condition variables, and the service early warning model maps the early warning decision logic into a rule logic structure. The initial warning rule template is generated based on the rule condition variables and the rule logic structure.

4. The service early warning signal generation method based on a large model according to claim 1, characterized in that, The step of generating a target service early warning signal based on the service early warning rules and the standard heterogeneous dataset includes: The standard heterogeneous dataset and the service early warning rules are matched to generate events to be warned; The warning level of the event to be warned is determined according to the service warning rules, and the warning signal of the target service is generated according to the warning level.

5. The service early warning signal generation method based on a large model according to claim 4, characterized in that, The step of matching the standard heterogeneous dataset with the service early warning rules to generate early warning events includes: Build a data processing engine and extract the current data stream from the standard heterogeneous dataset using the data processing engine; Match the current data stream with the service warning rules; The event corresponding to the current data stream that matches the service warning rule is identified as the event to be warned.

6. The service early warning signal generation method based on a large model according to claim 4, characterized in that, After generating the target service early warning signal based on the service early warning rules and the standard heterogeneous dataset, the process includes: Based on the warning level, determine the target service warning signal distribution object, distribution channel, and distribution time interval; Based on the distribution channel and the distribution time interval, the target service warning signal is distributed to the distribution object corresponding to the target service warning signal.

7. The method for generating service early warning signals based on a large model according to any one of claims 1 to 6, characterized in that, The step of preprocessing the initial multi-source heterogeneous data to generate a standard heterogeneous dataset includes: Anomaly detection algorithm is used to identify anomalous data in the initial multi-source heterogeneous data, and the anomalous data is cleaned to generate valid multi-source heterogeneous data. The effective multi-source heterogeneous data is processed by entity recognition, sentiment analysis and topic extraction to generate the standard heterogeneous dataset.

8. A service early warning signal generation device based on a large model, characterized in that, include: A standard heterogeneous dataset generation module is used to acquire initial multi-source heterogeneous data and perform data preprocessing on the initial multi-source heterogeneous data to generate a standard heterogeneous dataset. The target service early warning rule generation module is used to generate target service early warning rules through a pre-trained service early warning model and the standard heterogeneous dataset. The target service early warning signal generation module is used to generate target service early warning signals based on the service early warning rules and the standard heterogeneous dataset.

9. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the service early warning signal generation 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 a computer program that, when executed by a processor, causes the processor to implement the service warning signal generation method based on a large model as described in any one of claims 1 to 7.

Images