Business system fault processing method and device based on large model, equipment and medium

Abstract

The invention relates to the technical field of intelligent decision making, can be applied to business system platforms of financial science and technology, medical treatment and health and the like, and discloses a business system fault processing method, device, equipment and medium based on a large model, and the method comprises the steps: obtaining fault data of a business system, and carrying out the content type extraction of the fault data to obtain a data content type; performing intention analysis on the fault data according to the data content type to obtain a fault data intention; dividing the fault data into consultation type faults and repair type faults according to the fault data intention; semantic matching retrieval is conducted on the consultation type faults through a preset knowledge base, and fault responses are obtained; performing plan library matching on the repair type fault to obtain a matching result, and determining a solution of the fault data according to the matching result; and collecting the fault response and the solution into a fault processing result. By performing intention analysis and matching on the fault data, automatic fault processing is realized, and the manual intervention cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, and in particular to a method, apparatus, device, and medium for handling business system faults based on a large model. Background Technology

[0002] With the rapid development of information technology, enterprises generally tend to develop dedicated business systems to support internal operations and business development. Currently, various industries have formed business system ecosystems adapted to their own scenarios. Taking the healthcare field as an example, electronic medical record systems integrate patient medical history, examination results and other data to support medical staff collaboration; hospital information systems (HIS) coordinate administrative and logistical businesses such as registration, payment, and drug management; and in the fintech field, core banking systems handle account management and transaction settlement, while risk control systems monitor abnormal transactions in real time and trigger early warnings, requiring high concurrency and low latency.

[0003] As some enterprises expand their businesses, the complexity of their business systems also increases, leading to a rise in system failures reported by users. Existing system failure handling methods mainly rely on manual analysis and processing, but this approach is inefficient, requires a lot of manual intervention, and is difficult to handle massive numbers of failures. At the same time, the processing flow is fragmented and lacks a unified automated solution. Summary of the Invention

[0004] This invention provides a business system fault handling method, apparatus, equipment, and medium based on a large model. By performing intent analysis and matching on fault data, it achieves automated fault handling and reduces the cost of manual intervention.

[0005] Firstly, a business system fault handling method based on a large model is provided, including: Obtain fault data from the business system, extract the content type from the fault data, and obtain the data content type; Based on the data content type, the intent of the fault data is obtained by performing intent analysis on the fault data. Based on the intent of the fault data, the fault data is divided into consultation-type faults and repair-type faults; The system uses a pre-defined knowledge base to perform semantic matching and retrieval on the consultation-type faults to obtain fault responses. The aforementioned repair-type faults are matched against a contingency plan database to obtain matching results, and solutions for the fault data are determined based on the matching results. The fault response and the solution are combined into a fault handling result.

[0006] Secondly, a fault handling device for business systems based on a large model is provided, including: The acquisition and extraction module is used to acquire fault data from the business system, extract the content type from the fault data, and obtain the data content type. The analysis module is used to perform intent analysis on the fault data according to the data content type to obtain the intent of the fault data; A classification module is used to classify the fault data into consultation-type faults and repair-type faults according to the fault data intent; The retrieval module is used to perform semantic matching retrieval of the consultation-type faults through a preset knowledge base to obtain fault responses; The matching and determination module is used to match the repair-type faults with the contingency plan library, obtain the matching results, and determine the solution for the fault data based on the matching results; The aggregation module is used to aggregate the fault response and the solution into a fault handling result.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for handling faults in a business system based on a large model.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned fault handling method for a business system based on a large model.

[0009] In the aforementioned solution implemented by a business system fault handling method, device, computer equipment, and storage medium based on a large model, the carrier form of fault data can be clearly identified by extracting the data content type, providing targeted methods for subsequent processing and avoiding low processing efficiency or errors caused by mixed data formats. The core intent of the user or system is extracted from the specific content, irrelevant information is removed, and the essence of the fault is focused on, reducing the blindness of subsequent classification and processing and ensuring that each type of fault can be accurately matched to the corresponding solution path. A binary search method simplifies the fault handling process: consultation-type faults can rely on the knowledge base for quick responses, while repair-type faults require the activation of contingency plans, avoiding resource waste caused by mixed processing of different types of faults and improving fault triage efficiency. Standardized answers accumulated in the knowledge base are used to quickly generate responses through semantic matching, achieving automated processing of consultation-type faults, reducing manual intervention costs, and ensuring the consistency and accuracy of answers. Repair-type faults usually correspond to known fault scenarios, and standardized repair processes stored in the contingency plan library can be quickly invoked through matching, shortening fault response time. Dispersed responses and solutions are integrated into a unified result, making it easier for users or processing personnel to obtain complete information at once, improving the convenience of fault resolution. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for a business system fault handling method based on a large model, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating a business system fault handling method based on a large model according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a business system fault handling device based on a large model according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to one embodiment of the present invention; Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] This invention provides a fault handling method for business systems based on a large model, which can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The server can obtain fault data from the business system, extract the content type of the fault data to obtain the data content type; perform intent analysis on the fault data based on the data content type to obtain the fault data intent; classify the fault data into consultation-type faults and repair-type faults based on the fault data intent; perform semantic matching retrieval on the consultation-type faults through a preset knowledge base to obtain fault responses; perform pre-plan matching on the repair-type faults to obtain matching results, and determine the solution for the fault data based on the matching results; aggregate the fault responses and solutions into a fault handling result, and feed the fault handling result back to the client. This invention provides a business system fault handling device based on a large model. For fault handling result business, by performing intent analysis and matching on fault data, it achieves automated fault handling and reduces the cost of manual intervention. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster composed of multiple servers. The invention will be described in detail below through specific embodiments.

[0014] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a business system fault handling method based on a large model, provided by an embodiment of the present invention, includes the following steps: S1. Obtain fault data from the business system, extract the content type from the fault data, and obtain the data content type.

[0015] In this embodiment of the invention, "acquisition" refers to collecting and gathering data related to system faults from the business system, and "content type extraction" refers to analyzing and judging the acquired fault data to determine its specific content category or data type.

[0016] Specifically, fault information reported by users is collected from business systems, such as error messages, consultation texts, and fault screenshots. By analyzing the format, carrier, or core characteristics of this data, it is classified into specific types, such as text, images, documents, and system logs, providing a basic classification basis for subsequent targeted processing.

[0017] In specific healthcare scenarios, fault data can be categorized into types such as abnormal electronic medical record viewing, errors in diagnosis and treatment system operations, and failures in uploading medical images. This helps to quickly match the corresponding processing paths, ensure the stable operation of the healthcare system, and reduce interference with the diagnosis and treatment process.

[0018] In fintech scenarios, fault data can be categorized into transaction record query anomalies, risk control model warning logs, and user identity authentication material upload errors, making it easier to accurately call different solutions and improve the security of financial services and user experience.

[0019] In this embodiment of the invention, the step of extracting the content type of the fault data to obtain the data content type includes: Identify the file extension of the fault data; The file extensions are mapped to content types according to a preset mapping table to obtain the mapping results; The fault data is standardized and classified according to the mapping results to obtain the data content type.

[0020] In this embodiment of the invention, the identification refers to determining the approximate type category of a file by examining the part after the dot in the filename; the content type mapping refers to finding the corresponding detailed content type description based on the identified file extension according to this mapping table; and the standardized classification refers to classifying and organizing the mapped content types according to certain classification standards and systems based on the content type mapping, so that they conform to unified specifications and formats.

[0021] Specifically, from the fault data collected from the business system, the identifying characters (i.e., file extensions) at the end of each file name are extracted; for example, if the fault data is "Inspection Report.pdf", "Login Error.txt", or "Image.png" reported by users, the extensions ".pdf", "txt", and "png" are identified respectively; for log files automatically generated by the system, such as "error.log", the extension ".log" is also extracted.

[0022] Furthermore, a pre-defined mapping table is invoked, which predefines the correspondence between common file extensions and content types (e.g., .pdf → "document type", .txt → "text type", .png → "image type", .log → "system log type"). The extensions identified in the first step are compared one by one with the entries in the mapping table to match the corresponding content type. For example, the extension ".pdf" corresponds to "document type", and ".png" corresponds to "image type". Finally, the content type mapping result corresponding to each fault data is output.

[0023] Furthermore, based on the mapping results obtained in the second step, all fault data are grouped and categorized according to the mapped content type (such as "document", "text", "image", "system log" etc.) to form a unified classification system. For example, all files mapped to "document" (including extensions such as .pdf, .doc, .docx, etc.) are grouped into the same category, while "image" files (including .png, .jpg, .jpeg, etc.) are grouped into another category, ultimately outputting a structured data content type classification result.

[0024] In this embodiment of the invention, the carrier format of data can be initially distinguished by quick identification, providing a simple and efficient starting point for subsequent classification; standardized mapping is achieved through a preset mapping table to avoid inconsistencies in classification caused by differences in human judgment, and to ensure file extensions from different sources; through standardized classification, scattered fault data is integrated into a clear type system, which facilitates the subsequent invocation of dedicated processing logic for different types of data.

[0025] In this embodiment of the invention, the scattered fault data is made more orderly by classification, which facilitates rapid location and management; at the same time, it provides a basis for subsequent targeted processing, such as using natural language processing for text-based faults and image recognition technology for image-based faults, thereby improving processing efficiency.

[0026] S2. Perform intent analysis on the fault data according to the data content type to obtain the fault data intent.

[0027] In this embodiment of the invention, intent analysis refers to using specific methods and techniques to delve into the true purpose, desired outcome, and possible abnormal motivations of users or business operations hidden behind fault data.

[0028] Specifically, after determining the content type of the fault data, such as text, image, system log, etc., we conduct in-depth interpretation of the characteristics of different types of data, analyze the core demands or purposes behind them, and finally clarify the intention of the fault raiser, such as seeking solutions, reporting system faults, consulting rule information, etc.

[0029] In healthcare scenarios, text-based queries like "How to view electronic medical records" can be identified as "functional consultation intent," while system log queries like "Test result upload failed" can be identified as "system fault feedback intent." This helps to quickly match the corresponding response strategies and ensure efficient doctor-patient communication and healthcare system operation.

[0030] In financial scenarios, text-based messages such as "credit card repayment limit" are identified as "intention to inquire about rules," while image-based messages such as "ID card upload error" are identified as "intention to malfunction." This allows for targeted access to customer service scripts or technical solutions, improving the accuracy of financial service responses and enhancing user trust.

[0031] In this embodiment of the invention, the step of performing intent analysis on the fault data based on the data content type to obtain the fault data intent includes: The classification and labeling system for the fault data is determined based on the data content type and preset business attributes; Convert the aforementioned classification tag system into prompt words; The intent of the fault data is obtained by analyzing the prompt words.

[0032] In this embodiment of the invention, the "determination" refers to conducting a comprehensive and in-depth analysis and judgment of the fault data based on the identified data content type and in combination with preset business attributes, thereby clarifying the fault type to which it belongs and constructing a scientific and reasonable classification label system. The "conversion" refers to converting the various labels in the constructed classification label system into prompt words that can be understood and processed by subsequent intent analysis models or algorithms. The "intent analysis" refers to using the converted prompt words, with the help of specific algorithms, models or rules, to conduct in-depth analysis and interpretation of the fault data, thereby revealing the true purpose of the user or business operation, the expected result, and the possible fault tendencies contained in the fault data.

[0033] First, when S1 determines the content type (text or image) by checking the file type or file extension, if it is text, it directly enters the intent classification stage; if it is an image, it enters the OCR detection stage, using OCR technology to extract the text content as input for subsequent processing. If OCR extraction fails (such as blurry images or unrecognizable text), the system records the error and prompts the user to resubmit the fault.

[0034] Specifically, based on the identified data content types (such as document, image, and text), and according to business attributes, such as medical consultation and system fault reporting in the medical field, and account inquiries, transaction anomaly feedback, and fault content characteristics in the financial field, specific fault type labels are defined, such as "Document - Data Format Error", "Text - Business Rule Consultation", and "Image - Recognition Anomaly Report". Each label is then matched with a corresponding level label, such as "Basic Consultation", "General Fault", and "Emergency Repair", forming a structured classification label system.

[0035] Furthermore, the classification label system is transformed into clear instruction prompts, explicitly informing the DeepSeek model of its role as an "intent classification assistant"—it needs to determine the specific type and level of the fault based on the input fault data content, combined with the data content type and the label system; for example, prompting the model: "When the input data is 'image type', match the most appropriate type from labels such as 'image type - identify abnormal repair report' and 'image type - format incompatibility', and label the level", so that the model clearly understands the classification criteria and output requirements.

[0036] Furthermore, fault data containing specific content, such as "uploaded patient CT images cannot be displayed", with the data content type being "image", is input into the configured DeepSeek model. Based on the injected classification label system, the model analyzes the core intent of the fault, such as "image - identify abnormal repair", and outputs the intent result containing the fault type and level, i.e., the fault data intent.

[0037] In this embodiment of the invention, a targeted classification tag system is constructed by combining the content characteristics of the data itself and the attributes of the business scenario. This ensures that the tags are highly adapted to business needs, avoids ambiguity or redundancy in classification, and provides a clear framework for subsequent intent analysis. Converting the structured classification tag system into natural language prompts allows the intent analysis model to more accurately understand the classification goals and judgment criteria, reduces the model's misinterpretation of the tag system, and improves the guidance of intent analysis. By using prompts to perform intent analysis on fault data, the intent analysis is prevented from deviating from business goals or causing ambiguity.

[0038] In this embodiment of the invention, by combining the data content type with targeted analysis of intent, the core of the fault can be grasped more accurately, avoiding generalized interpretation; at the same time, fault handling can be more focused, ensuring that subsequent solutions match the user's real needs, reducing ineffective operations, and improving fault resolution efficiency and user satisfaction.

[0039] S3. Based on the intended fault data, classify the fault data into consultation-type faults and repair-type faults.

[0040] In this embodiment of the invention, the division refers to the process of clearly classifying fault data into two different categories, namely consultation or repair, based on the user's or business operation's true purpose, expected result, and fault nature reflected in the fault data, according to a predetermined classification standard.

[0041] Specifically, based on the previously analyzed fault data intent (i.e., the core purpose or demand of users in raising faults), the collected fault data is further categorized into two main types: one is consultation-type faults, where the intent of these faults is mainly that users want to obtain information, understand rules, or seek guidance, such as "how to operate a certain function of the system" or "what does this policy mean"; the other is repair-type faults, where the intent focuses on reporting system faults, functional abnormalities, or errors that need to be resolved, such as "the system submit button is not responding" or "garbled characters appear in the data display." This classification clarifies the nature of the faults and provides a basis for taking different handling strategies in the future.

[0042] In specific healthcare scenarios, this classification can quickly distinguish user intent: consultation-related faults can be answered by medical staff or intelligent customer service to ensure the efficiency of patients' information acquisition; repair-related faults will directly trigger the technical team to handle them, avoiding the impact of system failures on the diagnosis and treatment process and ensuring smooth medical services.

[0043] In financial scenarios, this classification can accurately triage faults. Consultation-related faults are responded to by financial advisors or intelligent dialogue systems, improving the user service experience; repair-related faults are quickly transferred to the technical department for investigation, timely resolution of fund-related faults, reduction of user fund risks and trust crises, and protection of the safe operation of financial business.

[0044] S4. Perform semantic matching retrieval on the consultation-type faults using a preset knowledge base to obtain fault responses.

[0045] In this embodiment of the invention, the semantic matching retrieval refers to the comprehensive application of natural language processing technology and information retrieval methods to deeply understand the semantics of the fault and accurately search for semantically related response content in the knowledge base.

[0046] Specifically, for the identified consultation-related faults, a pre-built knowledge base is used, which contains information such as common faults, standard answers, and rule descriptions in the field. Semantic analysis technology is used to compare the similarity between the fault and the content in the knowledge base, and the most matching information is found as the response to the fault. For example, if a user asks "medical insurance reimbursement process", the system will match the corresponding step description from the knowledge base and provide feedback.

[0047] In this embodiment of the invention, the step of performing semantic matching retrieval on the consultation-type fault through a preset knowledge base to obtain a fault response includes: Semantic features are extracted from the aforementioned consultation-related faults to obtain semantic features; The semantic similarity of the semantic features with each historical fault in the preset knowledge base is calculated to obtain the semantic similarity of each historical fault. Extract the historical resolution rate of the historical faults, and perform a weighted summation of the semantic similarity and the historical resolution rate to obtain a comprehensive score; The comprehensive scores are sorted and filtered to obtain fault responses.

[0048] In this embodiment of the invention, the semantic feature extraction refers to using natural language processing techniques and algorithms to mine key information and features that can accurately express the semantic essence of the problem from the text content of consultation-type problems. The semantic similarity calculation refers to measuring the degree of similarity between the extracted semantic features of consultation-type problems and the semantic features of each historical problem in the preset knowledge base. The weighted synthesis refers to comprehensively calculating the semantic similarity and historical resolution rate of each historical problem according to a certain weight to obtain a comprehensive score. The sorting and filtering refers to sorting all historical problems in the preset knowledge base according to the calculated comprehensive score, and filtering out the historical problems and their responses that best meet the current consultation-type problem requirements according to certain rules, as the final problem response.

[0049] Specifically, by using natural language processing technology, key information and semantic features are extracted from the text of consultation-related faults, including core vocabulary such as "reimbursement ratio" and "loan term", grammatical structures such as interrogative sentences and imperative sentences, and implicit intentions such as "inquiry rules" and "seeking steps", and the text is transformed into machine-recognizable feature vectors or structured data.

[0050] Furthermore, semantic similarity algorithms, such as cosine similarity and BERT models, are used to compare the extracted semantic features with the semantic features of historical faults stored in the knowledge base one by one, calculating the degree of matching between the two at the semantic level. The "historical resolution rate" corresponding to each historical fault is retrieved from the knowledge base, that is, the proportion of times the answer to the historical fault has successfully resolved similar inquiries in the past. For example, if the answer to a historical fault is recognized by 90% of users, then by using preset weights, such as similarity accounting for 70% and resolution rate accounting for 30%, the two values ​​are combined to calculate a comprehensive score. For example, if similarity is 0.8 and resolution rate is 0.9, the comprehensive score = 0.8 × 0.7 + 0.9 × 0.3 = 0.83. Finally, all historical faults are sorted from high to low comprehensive scores, and the answers corresponding to the top 1-3 historical faults with the highest scores are selected as the responses to the current inquiry type fault.

[0051] In this embodiment of the invention, the step of sorting and filtering the comprehensive scores to obtain fault responses includes: The comprehensive scores are filtered according to a preset threshold to obtain the retained comprehensive scores; The highest score is extracted from the retained comprehensive scores to obtain the highest comprehensive score; The answer corresponding to the highest comprehensive score will be used as the fault response.

[0052] In this embodiment of the invention, the filtering refers to the process of judging and selecting all calculated comprehensive scores based on a pre-set specific numerical standard, and the extraction of the highest score refers to the process of finding the comprehensive score with the largest value in the set of retained comprehensive scores that have passed the threshold filtering.

[0053] Specifically, if the highest score after sorting is higher than the preset threshold, it means that the matched answer is accurate and effective enough, and the answer corresponding to the result is directly used as the fault response; if the highest score is lower than the threshold, the large model is triggered to intervene and ensure that consultation-type faults are responded to reasonably.

[0054] In this embodiment of the invention, the limitations of literal matching are overcome, and historical faults most relevant to the current fault are found from a semantic level, thereby improving the accuracy of matching. In addition to considering the semantic matching degree, the actual effect of historical faults is also taken into account, avoiding the recommendation of semantically similar but poorly resolved answers. By sorting and filtering, it is ensured that the final answer is the one that is "semantically most matched and has the best historical resolution effect", reducing interference from invalid information and quickly providing users with high-quality solutions.

[0055] In this embodiment of the invention, standardized answers are quickly invoked to respond to inquiries, reducing the cost of manual intervention; semantic matching ensures the accuracy and consistency of responses, avoiding information transmission deviations; at the same time, response speed is improved, allowing users to obtain clear answers in a timely manner, enhancing service efficiency and experience.

[0056] S5. Perform a pre-plan database matching on the repair-type faults to obtain matching results, and determine the solution for the fault data based on the matching results.

[0057] In this embodiment of the invention, the contingency plan matching refers to the process of comparing and analyzing the repair-type faults to be processed with various contingency plans in the pre-built contingency plan library in detail to find the contingency plan that best matches the current fault. The determination refers to the process of selecting the most suitable solution from the matched contingency plans based on the results of the contingency plan matching, combined with the actual situation and business needs.

[0058] Specifically, for the identified repair-related faults, such as system faults and functional abnormalities, they are compared and matched with a pre-built contingency plan library, which contains the characteristics of various common faults and corresponding solution steps. Based on the matched contingency plan content, a specific solution is determined. For example, if a user reports "failed to submit to the registration system", the corresponding contingency plan in the contingency plan library, "network delay caused submission timeout", is matched, and the solution of "refreshing the page and retrying" is adopted.

[0059] In this embodiment of the invention, the step of matching the repair-type faults against the contingency plan database to obtain the matching result includes: The core fault features are extracted from the repair-type faults to obtain the core fault features; The core fault characteristics are compared with the fault type tags in the preset contingency plan library to obtain the fault type tags; The core fault features are repaired and retrieved based on the fault type tags and the preset contingency plan library to obtain matching results.

[0060] In this embodiment of the invention, the core fault feature extraction refers to accurately identifying and extracting the elements and information that most directly and critically reflect the essence of the fault from various phenomena, data and descriptions presented by repair-type faults using professional technical means and analysis methods. The comparison refers to comparing and analyzing the extracted core fault features one by one with the predefined fault type labels in the preset plan library to determine which fault type label best matches the core fault features. The repair retrieval refers to the process of searching for the corresponding repair solutions, operation steps, precautions and other relevant information in the preset plan library based on the fault type labels obtained from the comparison.

[0061] Specifically, the content of repair-related faults is broken down to extract core fault elements, including the business scenario in which the fault occurred, such as the login process and data submission page; specific error manifestations, such as pop-up prompts and function lag; and the system modules involved, such as the user management module and data storage module, to form a structured fault description information.

[0062] Furthermore, the extracted structured fault description information is compared with the preset fault type tags in the contingency plan library, such as tags like "login verification failed", "data synchronization timeout", and "interface display abnormality", to find the tag that best matches the current fault characteristics. For example, if the fault description is "database connection failed when submitting an order", then the "data storage module - connection abnormality" type tag is matched, and this tag is used as the retrieval identifier for the next step of precise matching.

[0063] Furthermore, using the matched fault type tag as an index, all repair plans associated with that tag are retrieved from the plan library. For example, plans for "data storage module - connection failure" may include "restart database service" and "check network connection configuration".

[0064] In this embodiment of the invention, determining the solution for the fault data based on the matching result includes: Determine whether the matching result is valid; If the matching result is valid, then the core fault characteristics are processed according to the matching result to obtain a solution for the fault data; If the matching result is invalid, the core fault characteristics will be submitted for manual processing.

[0065] In this embodiment of the invention, "valid" means that the matching result is practically usable and reasonable for resolving the current core fault characteristics. "Fault handling" means that after the matching result is determined to be valid, the core fault characteristics are actually repaired and resolved according to the repair plan, operation steps and related requirements provided in the matching result.

[0066] Specifically, the matching results of the pre-set validity rules are evaluated, such as whether the overlap between the matched plan and the core fault characteristics of the current repair type exceeds a threshold, whether the plan contains clear and executable solution steps, and whether the success rate of the plan in the past meets the standard. For example, if a user reports "the APP cannot load the bill", the matched plan is "network fluctuation caused loading failure", and the step of "checking the network connection and reloading" in the plan is completely consistent with the fault characteristics and the historical success rate is 95%.

[0067] Furthermore, when the matching result is valid, the standardized solution process in the matching plan is directly invoked to address the core fault characteristics in a targeted manner; when the matching result is invalid, such as no matching plan is found or the plan is not related to the fault, the system automatically transfers the detailed information of the core fault characteristics, such as the fault phenomenon, the occurrence scenario, and user operation records, to the manual processing channel and assigns it to professional technicians.

[0068] In this embodiment of the invention, unstructured natural language faults are transformed into structured features, providing a quantifiable analytical basis for subsequent similarity calculation and matching; it breaks through the limitations of literal matching, finds the historical faults most relevant to the current fault from the semantic level, and improves the accuracy of matching; it not only considers the semantic matching degree, but also incorporates the actual effect of historical faults, avoiding recommending answers that are semantically similar but have poor actual solution effects.

[0069] In this embodiment of the invention, the matching mechanism of the preset plan library can quickly locate the type of repair-type fault and the corresponding solution, reducing the time cost of fault diagnosis; at the same time, the standardized solution can ensure that the processing process is standardized and uniform, avoid inefficiency or incorrect processing caused by differences in human experience, improve the timeliness and accuracy of fault repair, and ensure the stable operation of the system or business.

[0070] S6. Combine the fault response and the solution into a fault handling result.

[0071] In this embodiment of the invention, the fault response obtained by matching the knowledge base for consultation-type faults, such as "the medical insurance reimbursement ratio is 70%", and the solution determined by matching the contingency plan database or manual processing for repair-type faults, such as "data synchronization can be restored after restarting the system", are integrated into a complete fault handling result. This result covers the response to the user's fault and the specific solution, forming a closed-loop handling conclusion. This not only makes it convenient for users to obtain complete information, but also makes it convenient for the business system to record and trace the entire fault handling process.

[0072] As can be seen, in the above solution, for the fault handling result business, fault data from the business system is obtained, and the content type of the fault data is extracted to obtain the data content type; the intent analysis of the fault data is performed on the fault data according to the data content type to obtain the fault data intent; the fault data is divided into consultation-type faults and repair-type faults according to the fault data intent; semantic matching retrieval of the consultation-type faults is performed through a preset knowledge base to obtain fault responses; the repair-type faults are matched with a contingency plan library to obtain matching results, and solutions for the fault data are determined based on the matching results; the fault responses and solutions are aggregated into fault handling results. By performing intent analysis and matching on the fault data, automated fault handling is achieved, reducing the cost of manual intervention.

[0073] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0074] In one embodiment, a business system fault handling device based on a large model is provided, which corresponds one-to-one with the business system fault handling method based on a large model in the above embodiments. For example... Figure 3 As shown, this fault handling device for a business system based on a large model includes an acquisition and extraction module 101, an analysis module 102, a partitioning module 103, a retrieval module 104, a matching and determination module 105, and an aggregation module 106. Detailed descriptions of each functional module are as follows: The acquisition and extraction module 101 is used to acquire fault data from the business system, extract the content type from the fault data, and obtain the data content type. Analysis module 102 is used to perform intent analysis on the fault data according to the data content type to obtain the fault data intent; The classification module 103 is used to classify the fault data into consultation-type faults and repair-type faults according to the fault data intent; The retrieval module 104 is used to perform semantic matching retrieval of the consultation-type faults through a preset knowledge base to obtain fault responses; The matching and determination module 105 is used to match the repair-type faults with the contingency plan library, obtain the matching results, and determine the solution for the fault data based on the matching results. The aggregation module 106 is used to aggregate the fault response and the solution into a fault handling result.

[0075] In one embodiment, the extraction module 101, when extracting the content type of the fault data to obtain the data content type, is used to: Identify the file extension of the fault data; The file extensions are mapped to content types according to a preset mapping table to obtain the mapping results; The fault data is standardized and classified according to the mapping results to obtain the data content type.

[0076] In one embodiment, when the analysis module 102 performs intent analysis on the fault data according to the data content type to obtain the fault data intent, it is used to: The classification and labeling system for the fault data is determined based on the data content type and preset business attributes; Convert the aforementioned classification tag system into prompt words; The intent of the fault data is obtained by analyzing the prompt words.

[0077] In one embodiment, when the retrieval module 104 performs semantic matching retrieval on the consultation-type fault through a preset knowledge base and obtains a fault response, it is used to: Semantic features are extracted from the aforementioned consultation-related faults to obtain semantic features; The semantic similarity of the semantic features with each historical fault in the preset knowledge base is calculated to obtain the semantic similarity of each historical fault. Extract the historical resolution rate of the historical faults, and perform a weighted summation of the semantic similarity and the historical resolution rate to obtain a comprehensive score; The comprehensive scores are sorted and filtered to obtain fault responses.

[0078] When sorting and filtering the comprehensive scores to obtain fault responses, the following is used: The comprehensive scores are filtered according to a preset threshold to obtain the retained comprehensive scores; The highest overall score is obtained by filtering the retained comprehensive scores. The answer corresponding to the highest comprehensive score will be used as the fault response.

[0079] In one embodiment, when the matching determination module 105 performs a matching process against the pre-set plan library for the repair-type faults and obtains a matching result, it is used to: The core fault features are extracted from the repair-type faults to obtain the core fault features; The core fault characteristics are compared with the fault type tags in the preset contingency plan library to obtain the fault type tags; The core fault features are repaired and retrieved based on the fault type tags and the preset contingency plan library to obtain matching results.

[0080] In one embodiment, when determining a solution to the fault data based on the matching result, the method is used to: Determine whether the matching result is valid; If the matching result is valid, then the core fault characteristics are processed according to the matching result to obtain a solution for the fault data; If the matching result is invalid, the core fault characteristics will be submitted for manual processing.

[0081] This invention provides a business system fault handling device based on a large model. For fault handling result business, it acquires fault data from the business system, extracts the content type of the fault data to obtain the data content type, performs intent analysis on the fault data based on the data content type to obtain the fault data intent, classifies the fault data into consultation-type faults and repair-type faults based on the fault data intent, performs semantic matching retrieval on consultation-type faults using a preset knowledge base to obtain fault responses, performs pre-set solution matching on repair-type faults to obtain matching results, and determines solutions for the fault data based on the matching results, and aggregates fault responses and solutions into fault handling results. By performing intent analysis and matching on fault data, it achieves automated fault handling and reduces the cost of manual intervention.

[0082] Specific limitations regarding the fault handling device for a business system based on a large model can be found in the limitations of the fault handling method for a business system based on a large model described above, and will not be repeated here. Each module in the aforementioned fault handling device for a business system based on a large model can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0083] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the server-side functions or steps of a fault handling method for a large-scale business system.

[0084] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of a fault handling method for a large-scale business system.

[0085] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain fault data from the business system, extract the content type from the fault data, and obtain the data content type; Based on the data content type, the intent of the fault data is obtained by performing intent analysis on the fault data. Based on the intent of the fault data, the fault data is divided into consultation-type faults and repair-type faults; The system uses a pre-defined knowledge base to perform semantic matching and retrieval on the consultation-type faults to obtain fault responses. The aforementioned repair-type faults are matched against a contingency plan database to obtain matching results, and solutions for the fault data are determined based on the matching results. The fault response and the solution are combined into a fault handling result.

[0086] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain fault data from the business system, extract the content type from the fault data, and obtain the data content type; Based on the data content type, the intent of the fault data is obtained by performing intent analysis on the fault data. Based on the intent of the fault data, the fault data is divided into consultation-type faults and repair-type faults; The system uses a pre-defined knowledge base to perform semantic matching and retrieval on the consultation-type faults to obtain fault responses. The aforementioned repair-type faults are matched against a contingency plan database to obtain matching results, and solutions for the fault data are determined based on the matching results. The fault response and the solution are combined into a fault handling result.

[0087] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0089] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0090] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. If any software tools or components other than those of our company appear in the embodiments, they are merely illustrative examples and do not represent actual use. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A fault handling method for a business system based on a large model, characterized in that, include: Obtain fault data from the business system, extract the content type from the fault data, and obtain the data content type; Based on the data content type, the intent of the fault data is analyzed to obtain the fault data intent. Based on the intent of the fault data, the fault data is divided into consultation-type faults and repair-type faults; The system uses a pre-defined knowledge base to perform semantic matching and retrieval on the consultation-type faults to obtain fault responses. The aforementioned repair-type faults are matched against a contingency plan database to obtain matching results, and solutions for the fault data are determined based on the matching results. The fault response and the solution are combined into a fault handling result.

2. The business system fault handling method based on a large model as described in claim 1, characterized in that, The step of extracting the content type of the fault data to obtain the data content type includes: Identify the file extension of the fault data; The file extensions are mapped to content types according to a preset mapping table to obtain the mapping results; The fault data is standardized and classified according to the mapping results to obtain the data content type.

3. The business system fault handling method based on a large model as described in claim 1, characterized in that, The step of performing intent analysis on the fault data based on the data content type to obtain the fault data intent includes: The classification and labeling system for the fault data is determined based on the data content type and preset business attributes; Convert the aforementioned classification tag system into prompt words; The intent of the fault data is obtained by analyzing the prompt words.

4. The business system fault handling method based on a large model as described in claim 1, characterized in that, The step of performing semantic matching retrieval on the consultation-type faults through a preset knowledge base to obtain fault responses includes: Semantic features are extracted from the aforementioned consultation-related faults to obtain semantic features; The semantic similarity of the semantic features with each historical fault in the preset knowledge base is calculated to obtain the semantic similarity of each historical fault. Extract the historical resolution rate of the historical faults, and perform a weighted summation of the semantic similarity and the historical resolution rate to obtain a comprehensive score; The comprehensive scores are sorted and filtered to obtain fault responses.

5. The business system fault handling method based on a large model as described in claim 4, characterized in that, The process of sorting and filtering the comprehensive scores to obtain fault responses includes: The comprehensive scores are filtered according to a preset threshold to obtain the retained comprehensive scores; The highest overall score is obtained by filtering the retained comprehensive scores. The answer to the problem corresponding to the highest comprehensive score will be used as the problem response.

6. The business system fault handling method based on a large model as described in claim 1, characterized in that, The matching of the repair-type faults with the contingency plan database to obtain the matching results includes: The core fault features are extracted from the repair-type faults to obtain the core fault features; The core fault characteristics are compared with the fault type tags in the preset contingency plan library to obtain the fault type tags; The core fault features are repaired and retrieved based on the fault type tags and the preset contingency plan library to obtain matching results.

7. The business system fault handling method based on a large model as described in claim 6, characterized in that, The solution for determining the fault data based on the matching result includes: Determine whether the matching result is valid; If the matching result is valid, then the core fault characteristics are processed according to the matching result to obtain a solution for the fault data; If the matching result is invalid, the core fault characteristics will be submitted for manual processing.

8. A fault handling device for a business system based on a large model, characterized in that, include: The acquisition and extraction module is used to acquire fault data from the business system, extract the content type from the fault data, and obtain the data content type. The analysis module is used to perform intent analysis on the fault data according to the data content type to obtain the intent of the fault data; A classification module is used to classify the fault data into consultation-type faults and repair-type faults according to the fault data intent; The retrieval module is used to perform semantic matching retrieval of the consultation-type faults through a preset knowledge base to obtain fault responses; The matching and determination module is used to match the repair-type faults with the contingency plan library, obtain the matching results, and determine the solution for the fault data based on the matching results; The aggregation module is used to aggregate the fault response and the solution into a fault handling result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the business system fault handling method based on a large model as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the business system fault handling method based on a large model as described in any one of claims 1 to 7.

Images