Cloud communication abnormity diagnosis method, system and device and storage medium

Abstract

The invention discloses a cloud communication abnormity diagnosis method, system and device and a storage medium, belongs to the technical field of communication abnormity diagnosis, and solves the technical problem of low efficiency of manual cloud communication abnormity diagnosis. The method comprises the following steps: analyzing and preprocessing an abnormal case, and performing feature extraction and data cleaning after preprocessing to obtain a cleaned case; performing feature clustering on the cleaning cases in the same batch; similar TopN cases are retrieved in a vector knowledge base according to the cases obtained through feature clustering, and if the similar TopN cases are matched, the name, description, classification, root cause and solution content of the similar cases are returned; if the retrieval is not successful, performing root cause reasoning and problem classification recommendation by using a large language model, and finally returning an abnormal case root cause diagnosis result; displaying the processing result of the abnormal case to the user at the front end; and obtaining feedback of the user, and if the feedback is that the current problem can be effectively solved, adding / refreshing the case to the vector knowledge base in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of communication anomaly diagnosis, and more specifically, to a cloud communication anomaly diagnosis method, system, device and storage medium. Background Art

[0002] As cloud communication technology continues to develop, users have higher requirements for the service quality and reliability of cloud communications. Existing cloud communication capabilities cannot quickly and accurately locate the cause of abnormalities such as poor terminal signals and network jitters, resulting in poor customer experience, business impact, complaints, and even worse social impacts.

[0003] These types of anomalies are usually sporadic cases, and often require professional personnel to investigate and handle them, which has problems of poor accuracy and low efficiency. This not only wastes customers' time, but also increases various costs for operators. How to reduce costs, increase efficiency, and improve service quality has become an important research topic in the cloud communications industry. Summary of the invention

[0004] The technical problem to be solved by the present invention is to address the above-mentioned deficiencies in the prior art. A first object of the present invention is to provide a method for diagnosing cloud communication anomalies that can improve accuracy and efficiency.

[0005] A second objective of the present invention is to provide a cloud communication anomaly diagnosis system.

[0006] A third object of the present invention is to provide a computer device.

[0007] A fourth objective of the present invention is to provide a computer storage medium.

[0008] In order to achieve the above-mentioned objective 1, the present invention provides a method for diagnosing cloud communication anomalies, comprising the following steps:

[0009] Step 1. Obtain and save abnormal cases generated during cloud communication;

[0010] Step 2. Scan and collect the saved abnormal cases regularly, analyze and preprocess the collected abnormal cases, and after the preprocessing is completed, perform feature extraction and data cleaning to obtain cleaned cases;

[0011] Step 3. Perform feature clustering on the cleaning cases of the same batch to avoid repeated analysis of abnormal cases;

[0012] Step 4. Search the vector knowledge base for similar TopN cases obtained by feature clustering. If a match is found, the name, description, classification, root cause, and solution of the similar case are returned. If the search is unsuccessful, a large language model is used to perform root cause reasoning and problem classification recommendations, and finally the root cause diagnosis results of the abnormal case are returned.

[0013] Step 5. Classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end;

[0014] Step 6. Obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time to improve the accuracy and efficiency of the next diagnosis.

[0015] As a further improvement, the methods for obtaining abnormal cases include:

[0016] Obtain abnormal cases of abnormal push notifications from the privacy account platform;

[0017] Obtain abnormal cases of user feedback on the use of self-service systems;

[0018] Obtain abnormal cases collected and reported by operation and maintenance personnel.

[0019] Furthermore, in step 2, the process of parsing and preprocessing the abnormal case is to load the abnormal case, fill in the corresponding template, parse the content of the abnormal case according to the preset rules, and save the abnormal case to the database.

[0020] Furthermore, the elements for feature extraction from the parsed content include case caller, case called party, business model, business account, reporting scenario, recording quality inspection, and ring tone recognition.

[0021] Furthermore, data cleaning is to reduce noise and filter the elements of feature extraction.

[0022] Furthermore, in step 3, the process of feature clustering is to use the K-Means unsupervised learning clustering algorithm to calculate the cleaned cases, use the Silhouette Coefficient to calculate the highest score and select the best K value, and perform secondary classification and outlier removal through features to achieve accurate clustering of different cases.

[0023] Further, in step 4, the vector knowledge base includes a historical case knowledge base, a typical case knowledge base, and an expert rule knowledge base;

[0024] The search process includes:

[0025] First-level search: search in the expert rule knowledge base according to expert rule matching;

[0026] Secondary search: searching the historical case knowledge base based on historical case matching;

[0027] Level 3 search: searching in the typical case knowledge base based on typical case matching;

[0028] The detection content includes: similar case name / number, similar case anomaly description, similar case anomaly classification, similar case anomaly root cause description, and similar case anomaly resolution measures;

[0029] It also includes adjusting the parameters of the large language model according to the output results of the large language model.

[0030] In order to achieve the above-mentioned second objective, the present invention provides a cloud communication abnormality diagnosis system, comprising:

[0031] The acquisition module is used to acquire and save abnormal cases generated during the cloud communication process;

[0032] The data cleaning module is used to regularly scan and collect the saved abnormal cases, analyze and preprocess the collected abnormal cases, and perform feature extraction and data cleaning to obtain cleaned cases after the preprocessing is completed;

[0033] Feature clustering module, used to perform feature clustering on cleaning cases in the same batch to avoid repeated analysis of abnormal cases;

[0034] The diagnosis module is used to search the vector knowledge base for similar TopN cases obtained by feature clustering. If a match is found, the name, description, classification, root cause, and solution of the similar case are returned. If the search is unsuccessful, a large language model is used to perform root cause reasoning and problem classification recommendations, and finally the root cause diagnosis results of the abnormal case are returned.

[0035] The display module is used to classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end;

[0036] The feedback module is used to obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time to improve the accuracy and efficiency of the next diagnosis.

[0037] In order to achieve the third objective above, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method for diagnosing cloud communication anomalies when executing the computer program.

[0038] In order to achieve the above-mentioned fourth objective, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned method for diagnosing cloud communication anomalies is implemented.

[0039] Beneficial Effects

[0040] Compared with the prior art, the present invention has the following advantages:

[0041] 1. The present invention can reduce operating costs: through automated abnormality diagnosis, enterprises can reduce their reliance on manual customer service, thereby significantly reducing operating costs.

[0042] 2. The present invention can improve service efficiency: users can quickly obtain abnormal diagnosis results without waiting for manual customer service, thereby improving service response speed and efficiency.

[0043] 3. The present invention can provide accurate diagnosis: as the usage time increases and the scope of use expands, the diagnostic results of the model will become more accurate and can better meet user needs.

[0044] 4. The invention can be continuously improved: by collecting customer feedback and making fine adjustments, the model can continuously optimize its own performance and provide higher quality services. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of the present invention;

[0046] Figure 2 Flowchart for the collection and cleaning of abnormal cases;

[0047] Figure 3 Flowchart for clustering features of abnormal cases;

[0048] Figure 4 Flowchart for vector knowledge base retrieval;

[0049] Figure 5 This is a software architecture diagram. DETAILED DESCRIPTION

[0050] The present invention will be further described below with reference to the specific embodiments in the accompanying drawings.

[0051] See also Figure 1 to Figure 5 In the scenario where the user uses a privacy number, the business party performs AXB binding in advance. When user A dials privacy number X, it will normally connect to B. However, when encountering abnormalities such as poor terminal signal and network jitter, abnormal cases may occur during the connection process of the privacy number.

[0052] In order to provide users with self-service abnormality diagnosis services, or cloud communication services to automatically diagnose abnormal cases, the present invention provides a cloud communication abnormality diagnosis method, comprising the following steps:

[0053] Step 1. Obtain and save abnormal cases generated during cloud communication;

[0054] Step 2. Scan and collect the saved abnormal cases regularly (by time period or batch), analyze and preprocess the collected abnormal cases, and after the preprocessing is completed, perform feature extraction and data cleaning to obtain cleaned cases;

[0055] Step 3. Perform feature clustering on the cleaning cases of the same batch to avoid repeated analysis of abnormal cases;

[0056] Step 4. The cases obtained by feature clustering are searched for similar TopN cases (i.e., the top N cases) in the vector knowledge base. If a match is found, the name, description, classification, root cause, and solution of the similar case are returned. If the search is unsuccessful, the large language model LLM is used for root cause reasoning and problem classification recommendation, and finally the root cause diagnosis result of the abnormal case is returned. The LLM technology is used to realize the automation and intelligence of abnormal diagnosis, reducing the possibility of human error.

[0057] Step 5. Classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end. Users can diagnose abnormalities without professional knowledge, which improves the user experience;

[0058] Step 6. Obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time, and the vector knowledge base can be continuously optimized to continuously adapt to new abnormal situations, maintain its diagnostic accuracy and timeliness, and be used to improve the accuracy and efficiency of the next diagnosis.

[0059] Ways to obtain abnormal cases include: obtaining abnormal cases pushed abnormally by the privacy account platform; obtaining abnormal cases reported by users using the self-service system; and obtaining abnormal cases collected and reported by operation and maintenance personnel.

[0060] In step 2, the process of parsing and preprocessing the abnormal case is to load the abnormal case, fill in the corresponding template, parse the content of the abnormal case according to the preset rules, and save the abnormal case to the database.

[0061] The elements extracted from the parsed content include the case caller, case callee, business model, business account, reporting scenario, recording quality inspection, and ring tone recognition. Data cleaning is to reduce noise and filter the extracted elements. Through data cleaning, the characteristic elements related to the cause of the abnormality of the case are retained, ensuring the accuracy of vector knowledge base retrieval and similar problem query and the completeness of the necessary information for large model reasoning input.

[0062] In step 3, the process of feature clustering is to use the K-Means unsupervised learning clustering algorithm to calculate the cleaned cases, use the Silhouette Coefficient to calculate the highest score and select the best K value, and perform secondary classification and outlier removal through features to achieve accurate clustering of different cases.

[0063] In step 4, the vector knowledge base includes a historical case knowledge base, a typical case knowledge base, and an expert rule knowledge base.

[0064] The retrieval process includes: first-level retrieval, searching in the expert rule knowledge base according to expert rule matching; second-level retrieval, searching in the historical case knowledge base according to historical case matching; third-level retrieval, searching in the typical case knowledge base according to typical case matching. The detection content includes: similar case name / number, similar case abnormal description, similar case abnormal classification, similar case abnormal root cause description, similar case abnormal solution measures. The accuracy and completeness of the data can be achieved through the three-level retrieval.

[0065] The method also includes adjusting the parameters of the large language model LLM according to the output results of the large language model LLM. When the large language model LLM is processing data, it is also training. With the increase of data volume and the expansion of the scope of use, as well as the optimization of parameters, the performance and accuracy of the large language model LLM will continue to improve, and it has good scalability.

[0066] A cloud communication abnormality diagnosis system, comprising:

[0067] The acquisition module is used to acquire and save abnormal cases generated during the cloud communication process;

[0068] The data cleaning module is used to regularly scan and collect the saved abnormal cases, analyze and preprocess the collected abnormal cases, and perform feature extraction and data cleaning to obtain cleaned cases after the preprocessing is completed;

[0069] Feature clustering module, used to perform feature clustering on cleaning cases in the same batch to avoid repeated analysis of abnormal cases;

[0070] The diagnosis module is used to search the vector knowledge base for similar TopN cases obtained by feature clustering. If a match is found, the name, description, classification, root cause, and solution of the similar case are returned. If the search is unsuccessful, the large language model LLM is used to perform root cause reasoning and problem classification recommendations, and finally the root cause diagnosis results of the abnormal case are returned.

[0071] The display module is used to classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end;

[0072] The feedback module is used to obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time to improve the accuracy and efficiency of the next diagnosis.

[0073] A computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the above-mentioned method for diagnosing cloud communication anomalies is implemented.

[0074] A computer-readable storage medium stores a computer program, which implements the above-mentioned method for diagnosing cloud communication anomalies when executed by a processor.

[0075] The software architecture of the system of the present invention is as follows Figure 5 The hierarchical design is described as follows:

[0076] (1) The application layer serves as a bridge between the communication business and the service layer. Its main functions are to provide valid data to the service layer, provide an interactive interface to users, and display the results processed by the service layer.

[0077] (2) The service layer is the intermediary between the application layer and the training layer. It is responsible for data interaction, storage, and management, and provides an API interface for abnormal case root cause diagnosis. Its main functions include abnormal case feature aggregation, similar problem retrieval, and abnormal root cause recommendation.

[0078] (3) The role of the training layer is to train and optimize the machine learning model, process input data and generate corresponding outputs, adapt to different application scenarios, and update the model's knowledge base through continuous optimization training, learning and evaluation to ensure that it adapts to business changes.

[0079] (4) The model layer mainly uses advanced large language models (LLM), embedding models, and clustering models, which can provide powerful text processing and generation capabilities.

[0080] The above are only preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the structure of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for diagnosing cloud communication anomalies, characterized in that: The following steps are involved: Step 1. Obtain and save abnormal cases generated during cloud communication; Step 2. Scan and collect the saved abnormal cases regularly, analyze and preprocess the collected abnormal cases, and after the preprocessing is completed, perform feature extraction and data cleaning to obtain cleaned cases; Step 3. Perform feature clustering on the cleaning cases of the same batch to avoid repeated analysis of abnormal cases; Step 4. Search the vector knowledge base for similar TopN cases obtained by feature clustering. If a match is found, the name, description, classification, root cause, and solution of the similar case are returned. If the search fails, the large language model (LLM) is used to perform root cause reasoning and problem classification recommendations, and finally the root cause diagnosis results of the abnormal case are returned; Step 5. Classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end; Step 6. Obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time to improve the accuracy and efficiency of the next diagnosis.

2. A method for diagnosing cloud communication anomalies according to claim 1, characterized in that: Ways to obtain exception cases include: Obtain abnormal cases of abnormal push notifications from the privacy account platform; Obtain abnormal cases of user feedback on the use of self-service systems; Obtain abnormal cases collected and reported by operation and maintenance personnel.

3. A method for diagnosing cloud communication anomalies according to claim 1, characterized in that: In step 2, the process of parsing and preprocessing the abnormal case is to load the abnormal case, fill in the corresponding template, parse the content of the abnormal case according to the preset rules, and save the abnormal case to the database.

4. A method for diagnosing cloud communication anomalies according to claim 3, characterized in that: The elements for feature extraction from the parsed content include case caller, case called party, business model, business account, reporting scenario, recording quality inspection, and ring tone recognition.

5. A method for diagnosing cloud communication anomalies according to claim 4, characterized in that: Data cleaning is to reduce noise and filter the features extracted.

6. A method for diagnosing cloud communication anomalies according to claim 1, characterized in that: In step 3, the process of feature clustering is to use the K-Means unsupervised learning clustering algorithm to calculate the cleaned cases, use SilhouetteCoefficient to calculate the highest score and select the best K value, and perform secondary classification and outlier removal through features to achieve accurate clustering of different cases.

7. A method for diagnosing cloud communication anomalies according to claim 1, characterized in that: In step 4, the vector knowledge base includes a historical case knowledge base, a typical case knowledge base, and an expert rule knowledge base; The search process includes: First-level search: search in the expert rule knowledge base according to expert rule matching; Secondary search: searching the historical case knowledge base based on historical case matching; Level 3 search: searching in the typical case knowledge base based on typical case matching; The detection content includes: similar case name / number, similar case anomaly description, similar case anomaly classification, similar case anomaly root cause description, and similar case anomaly resolution measures; The method also includes adjusting parameters of the large language model (LLM) according to the output result of the large language model (LLM).

8. A cloud communication abnormality diagnosis system, characterized in that: include: The acquisition module is used to acquire and save abnormal cases generated during the cloud communication process; The data cleaning module is used to regularly scan and collect the saved abnormal cases, analyze and preprocess the collected abnormal cases, and perform feature extraction and data cleaning to obtain cleaned cases after the preprocessing is completed; Feature clustering module, used to perform feature clustering on cleaning cases in the same batch to avoid repeated analysis of abnormal cases; The diagnosis module is used to search the vector knowledge base for similar TopN cases obtained by feature clustering. If a match is found, the name, description, classification, root cause and solution of the similar case are returned. If the search fails, the large language model (LLM) is used to perform root cause reasoning and problem classification recommendations, and finally the root cause diagnosis results of the abnormal case are returned; The display module is used to classify, trim and judge the returned content and results, and display the processing results of this abnormal case to the user on the front end; The feedback module is used to obtain feedback from the user after manual confirmation of the processing result. If the feedback is that the current problem can be effectively solved, the case will be added / refreshed to the vector knowledge base in real time to improve the accuracy and efficiency of the next diagnosis.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method for diagnosing cloud communication anomalies described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a method for diagnosing cloud communication anomalies as described in any one of claims 1 to 7 is implemented.

Images