Method for applying large star model to intelligent calculation center inspection analysis, medium and equipment

Abstract

The invention provides a method, medium and equipment for applying a star large model to inspection analysis of an intelligent calculation center, and the method comprises the steps: collecting inspection data of the intelligent calculation center, carrying out the data marking processing, and extracting four dynamic parameters, namely, an inspection index, an inspection target IP, a subordinate user and a user attention degree, from the inspection data; inputting the four dynamic parameters into a star large model for analysis, outputting a semantic text for representing the operation state of each service of the intelligent calculation center, and identifying entity words in the semantic text; analyzing the context correlation of the entity words in the semantic texts, and removing the semantic texts of which the correlation is lower than a set value; and matching the removed semantic text with a semantic database of the business, screening the semantic text according to a matching result, and generating an inspection report associated with the business. According to the invention, a natural language inspection report can be directly generated, which is convenient for a user to understand a business operation state; a calculation mode of context correlation is innovatively provided, and the accuracy of the inspection report is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent large model analysis, and specifically relates to a method, medium and equipment for using a star large model for inspection and analysis in an intelligent computing center. Background Art

[0002] The intelligent computing center, or artificial intelligence computing center, is an information infrastructure that integrates computing power, data, and algorithms. It is like a super-large intelligent brain, providing powerful computing power support for various cutting-edge technology applications and complex digital business scenarios. The intelligent computing center management platform can easily support large model analysis functions based on data. The object of analysis can be a group of people or a single user. The data source of the analysis can be various label data that have been aggregated or behavior detail data. The fast, massive, and multi-business parallel processing mode of the intelligent computing center further poses new challenges to business inspections. Existing technologies often use regular inspections of manual computer rooms for daily operation and maintenance, which not only consumes manpower, but also has low inspection efficiency and lacks focus and intelligent analysis of alarm problems. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a method, medium and equipment for using a large star model for inspection and analysis in an intelligent computing center.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a method for using a large star model for inspection and analysis in an intelligent computing center, comprising:

[0006] Collect inspection data from the intelligent computing center and perform data tagging processing, and extract four dynamic parameters from the inspection data: inspection index, inspection destination IP, user, and user attention;

[0007] Input the four dynamic parameters into the Xingchen big model for analysis, output the semantic text used to represent the operating status of each business in the intelligent computing center, and identify the entity words in the semantic text;

[0008] Analyze the contextual relevance of entity words in semantic texts and remove semantic texts whose relevance is lower than a set value;

[0009] The removed semantic text is matched with the semantic database of the business, and the semantic text is filtered according to the matching result to generate an inspection report associated with the business.

[0010] Optionally, the data marking process includes data cleaning, format normalization, and marking the business corresponding to the inspection data.

[0011] Optionally, the four dynamic parameters are first operated as follows before being input into the star model:

[0012] The inspection indicators are normalized according to type, and the normalized inspection indicators are associated with the inspection destination IP, the user to which they belong, and the user's attention to generate vector data for input into the Xingchen big model.

[0013] Optionally, the inspection indicators include delay rate and packet loss rate.

[0014] Optionally, identifying entity words in the semantic text is specifically:

[0015] Entity words in the semantic text are tagged using a part-of-speech tagging tool, and stop words are filtered for the tagged entity words. Optionally, the context relevance of the entity words is calculated using the following formula:

[0016] rel k =min-max(s k )*abs(Cos(CR(X i ), c k ))c k

[0017] In the formula, rel k represents the contextual relevance coefficient of entity word k, min-max represents the normalization function, s k is the confidence score of entity word k, abs represents the absolute value function, cos represents the cosine distance function, c k Represents the semantic representation of the semantic text where the entity word k is located, CR(X i ) represents the context X of the semantic text where the entity word k is located i The average semantic representation of .

[0018] Optionally, the semantic text after removal is matched with a semantic database of the business, and the semantic text is filtered according to the matching result, specifically:

[0019] A similarity algorithm based on word vectors is used to calculate the comprehensive similarity between the removed semantic text and the business-related historical semantic texts stored in the semantic database, and the semantic text with the greatest comprehensive similarity is selected.

[0020] Optionally, multiple service-related inspection reports of the same user are pushed according to the order of user attention.

[0021] In a second aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method of using a large star model for inspection and analysis in an intelligent computing center as described above.

[0022] In a third aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method of using a large star model for inspection and analysis in an intelligent computing center as described above is implemented.

[0023] The beneficial effects of the present invention are as follows: the present invention makes full use of the semantic analysis characteristics of the Xingchen big model, combines the four dynamic parameters of inspection indicators, inspection target IP set, users and user attention for calculation, and can directly generate natural language inspection reports, which is convenient for users to directly understand the business operation status without further parsing the data; the present invention also innovatively proposes a method for calculating context relevance, and obtains semantic text for generating inspection reports through double screening of context relevance and semantic database matching, which effectively improves the accuracy of inspection reports and reduces the content of non-related information in inspection reports. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is an overall flow chart of the method of using the Star Big Model for inspection and analysis in the intelligent computing center.

[0025] Figure 2 It is a data processing architecture diagram of the method of using the Star Big Model for inspection and analysis in the intelligent computing center. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0027] In one embodiment, the present invention proposes a method for using a large star model for inspection and analysis in an intelligent computing center, the process is as follows: Figure 1 As shown, the data architecture is as follows Figure 2 As shown, including:

[0028] Step 1: Collect inspection data from the intelligent computing center and perform data labeling.

[0029] In this embodiment, the collected intelligent computing center inspection data can be processed by data marking through the log layering data mining method of the intelligent computing center management platform, including data cleaning, format normalization, and marking the business corresponding to the inspection data, so as to match the inspection data with the business, and facilitate the subsequent generation of business-related inspection reports.

[0030] Step 2: Extract four dynamic parameters from the inspection data: inspection index, inspection destination IP, user and user attention.

[0031] Inspection indicators can be set according to the type and characteristics of the service or the needs of the user. In this embodiment, the service operation status is mainly measured by the network status, and the inspection indicators are set to include the delay rate and the packet loss rate.

[0032] Step 3: Input the four dynamic parameters into the Xingchen big model for analysis, output the semantic text used to represent the operating status of each business of the intelligent computing center, and identify the entity words in the semantic text.

[0033] The Xingchen Big Model is a product called Xingchen Big Model Software Factory released by China Telecom Artificial Intelligence Technology Co., Ltd. and China Telecom Artificial Intelligence Research Institute (TeleAI). It is the first breakthrough AI development tool in China. Small models usually refer to machine learning or deep learning models with fewer parameters and shallower layers. They have the advantages of being lightweight, efficient, and easy to deploy, and can run efficiently in resource-limited environments such as mobile devices or embedded systems. In contrast to small models, large models refer to machine learning models with large-scale parameters and complex computing structures. They are the product of the combination of "big data + large computing power + strong algorithms". They have a large parameter scale and can efficiently handle multi-user and multi-service inspection data processing tasks.

[0034] In the training process of large models, unsupervised learning methods are usually used. This means that the model does not require manually labeled data during the training process, but learns language rules and knowledge directly from the original text.

[0035] Before inputting the four dynamic parameters of inspection index, inspection target IP, user and user attention into the Star Big Model, the inspection index is first normalized according to the type, and the normalized inspection index is associated with the inspection target IP, user and user attention to generate vector data for input into the Star Big Model.

[0036] The specific process of semantic analysis of the Xingchen big model is as follows:

[0037] 1) Enter inspection-related information:

[0038] -##Start quoting##Inspection configuration resource requirements: 8 cores, 16G, 100T, inspection indicators: delay, packet loss, inspection destination IP: ip1, ip2, device owner: XXX, user attention: 30%, business weight: 80%##End quoting##-

[0039] 2) Based on the content of the inspection data dictionary structure and the vector database marking, the answer points / fragments are output in the form of natural language text in sequence, and then the answer points / fragments generated by the model in the previous step are organized into coherent sentences, i.e. semantic text, according to the preset template to ensure that the information is accurate and the logic is clear.

[0040] For the output semantic text used to represent the operating status of each business of the intelligent computing center, the part-of-speech tagging tool is used to tag the entity words in the semantic text, and the stop words are filtered out for the marked entity words.

[0041] Step 4: Analyze the contextual relevance of entity words in the semantic text and remove semantic texts with relevance lower than the set value.

[0042] Context relevance refers to the strength of association between the entity word retrieved from the context and the conversation context. In this embodiment, the context relevance of the entity word is calculated by the following formula:

[0043] rel k =min-max(s k )*abs(cos(CR(X i ), c k ))c k

[0044] In the formula, rel k represents the contextual relevance coefficient of entity word k, min-max represents the normalization function, s k is the confidence score of entity word k, abs represents the absolute value function, cos represents the cosine distance function, c k Represents the semantic representation of the semantic text where the entity word k is located, CR(X i ) represents the context X of the semantic text where the entity word k is located i The average semantic representation of .

[0045] Step 5: Match the removed semantic text with the semantic database of the business, and filter the semantic text according to the matching result to generate an inspection report associated with the business.

[0046] In this embodiment, a similarity algorithm based on word vectors is used to calculate the comprehensive similarity between the removed semantic text and the business-related historical semantic texts stored in the semantic database, and the semantic text with the greatest comprehensive similarity is screened out to generate the inspection report. Through the dual screening of context relevance and comprehensive similarity, it can effectively ensure that the semantic text finally used to generate the inspection report is the best text output by the model, and is not much different from the historical semantic text, which greatly reduces the error rate. In addition, for multiple business-related inspection reports of the same user, they are pushed according to the user's attention ranking.

[0047] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute a method of using a large star model for inspection and analysis in an intelligent computing center according to the aforementioned embodiment.

[0048] In another embodiment, the present invention proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method of using a large star model in the aforementioned embodiment for inspection and analysis in an intelligent computing center is implemented.

[0049] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0050] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0051] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for using a large star model for inspection and analysis in an intelligent computing center, characterized in that: include: Collect inspection data from the intelligent computing center and perform data tagging processing, and extract four dynamic parameters from the inspection data: inspection index, inspection destination IP, user, and user attention; Input the four dynamic parameters into the Xingchen big model for analysis, output the semantic text used to represent the operating status of each business in the intelligent computing center, and identify the entity words in the semantic text; Analyze the contextual relevance of entity words in semantic texts and remove semantic texts whose relevance is lower than a set value; The removed semantic text is matched with the semantic database of the business, and the semantic text is filtered according to the matching result to generate an inspection report associated with the business.

2. The method for using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 1, characterized in that: The data marking process includes data cleaning, format standardization and marking the business corresponding to the inspection data.

3. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 1, characterized in that: Before the four dynamic parameters are input into the star model, the following operations are performed: The inspection indicators are normalized according to type, and the normalized inspection indicators are associated with the inspection destination IP, the user to which they belong, and the user's attention to generate vector data for input into the Star Big Model.

4. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 3, characterized in that: The inspection indicators include delay rate and packet loss rate.

5. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 1, characterized in that: The identifying entity words in the semantic text is specifically: Use part-of-speech tagging tools to tag entity words in semantic texts, and filter stop words from the tagged entity words.

6. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 1, characterized in that: The context relevance of the entity word is calculated by the following formula: rel k =min-max(s k )*abs(Cos(CR(X i ),c k ))c k In the formula, rel k represents the contextual relevance coefficient of entity word k, min-max represents the normalization function, s k is the confidence score of entity word k, abs represents the absolute value function, cos represents the cosine distance function, c k Represents the semantic representation of the semantic text where the entity word k is located, CR(X i ) represents the context X of the semantic text where the entity word k is located i The average semantic representation of .

7. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 1, characterized in that: The semantic text after removal is matched with the semantic database of the business, and the semantic text is filtered according to the matching result, specifically: A similarity algorithm based on word vectors is used to calculate the comprehensive similarity between the removed semantic text and the business-related historical semantic texts stored in the semantic database, and the semantic text with the greatest comprehensive similarity is selected.

8. The method of using a large star model for inspection and analysis in an intelligent computing center as claimed in claim 7, characterized in that: For multiple business-related inspection reports of the same user, they are pushed according to the user's attention ranking.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables the computer to execute the method of using a large star model for inspection and analysis in an intelligent computing center as described in any one of claims 1-8.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method for using a large star model for inspection and analysis in an intelligent computing center as described in any one of claims 1 to 8 is implemented.

Images