CN-115269591-B - Intelligent alarm fingerprint merging method based on memory matrix

Abstract

The invention discloses an intelligent alarm fingerprint merging method based on a memory matrix, which comprises the following stages of alarm preprocessing, fingerprint optimizing and alarm merging. The technical scheme has the advantages that 1) the cost of intelligent alarm merging is reduced, 2) the memory matrix mode is used in the merging stage, the configuration cost of a user is reduced, meanwhile, the customization of alarm merging is improved, and the requirements of the user in an actual service scene can be met. 3) The processing speed is faster, and compared with a large number of neural network calculations, the matrix calculation can greatly improve the speed of merging calculation.

Inventors

• Ou Xiaoquan
• XIONG YAJUN
• ZHU PINYAN

Assignees

• 北京云集智造科技有限公司

Dates

Publication Date

20260512

Application Date

20220712

Claims (1)

1. 1. The intelligent alarm fingerprint merging method based on the memory matrix is characterized by comprising the following steps: (1) The alarm preprocessing stage mainly extracts important alarm fields, and can use default dynamic alarm fingerprints and also can customize key fields of alarm samples to generate customized alarm fingerprints; (2) In the fingerprint optimization stage, a redis is used for storing a memory matrix, and an expected matrix is multiplied by an inverse matrix of a feature matrix to obtain a memory matrix array, wherein the matrix array records expected information of a user for alarming; in the user production environment, the newly entered matrix can obtain the expected matrix of the alarm by multiplying the memory matrix, and the probability that the alarm belongs to each category can be obtained by calculating the expected matrix; (3) An alarm merging stage, in which the newly entered alarm is converted into a feature matrix, the feature matrix is multiplied by a memory matrix array to obtain an expected matrix, and finally alarms of the same category of the expected matrix are merged; The alarm preprocessing stage comprises the following steps: (1) Classifying the alarms according to the user demands to form a desired matrix; (2) redis stores the expected matrix; (3) Generating a hash value as an alarm fingerprint through an alarm key field or a user-defined field; (4) And converting the alarm fingerprint into an alarm feature matrix.

Description

Intelligent alarm fingerprint merging method based on memory matrix Technical Field The invention relates to the field of alarm merging, in particular to an intelligent alarm fingerprint merging method based on a memory matrix. Background The main problems of the existing SLS alarm intelligent merging technology are that alarm text clustering and clustering result merging are performed in two aspects. (1) And (3) clustering the alarm text, namely performing text editing distance calculation clustering through a local minimum hash value according to MinJoin clustering algorithm in a clustering algorithm adopted by the current SLS alarm intelligent merging technology. The algorithm is expensive in terms of computational effort and memory requirements, can be problematic in terms of noise reduction, high-dimensional data processing, and cannot handle clusters of different sizes well. (2) And the clustering results are combined, wherein the clustering results do not depend on historical alarms and are automatically learned, so that the combined results do not meet the requirements of users, and the noise reduction effect is not ideal due to the fact that the combined results are possibly overfitted. Based on the above reasons, an intelligent alarm fingerprint merging method based on a memory matrix becomes a technical problem to be solved urgently in the whole society. Disclosure of Invention In order to solve the technical problems, the technical proposal provided by the invention is that an intelligent alarm fingerprint merging method based on a memory matrix comprises the following stages, (1) The alarm preprocessing stage mainly extracts important alarm fields, and can use default dynamic alarm fingerprints and also can customize key fields of alarm samples to generate customized alarm fingerprints; (2) In the fingerprint optimization stage, a high-expansibility alarm merging scheme is created by using a redis memory matrix, key fields of input alarm data are extracted to generate hash values, the hash values are converted into feature matrices, and an expected matrix is multiplied by an inverse matrix of the feature matrices to obtain a memory matrix array, wherein the matrix array records expected information of a user on alarms; (3) And in the alarm merging stage, the newly entered alarms are converted into feature matrixes, the feature matrixes are multiplied by the memory matrix group to obtain expected matrixes, and finally the alarms with the same category of the expected matrixes are merged. Further, the step (1) includes the following steps: (1) Classifying the alarms according to the user demands to form a desired matrix; (2) redis stores the expected matrix; (3) Generating a hash value as an alarm fingerprint through an alarm key field or a user-defined field; (4) And converting the alarm fingerprint into an alarm feature matrix. Compared with the prior art, the invention has the advantages that: the beneficial effects of the invention are as follows: 1) The cost of intelligent alarm merging is reduced; 2) The memory matrix mode is used in the merging stage, so that the configuration cost of the user is reduced, the customization of alarm merging is improved, and the requirements of the user in an actual service scene can be met. 3) The processing speed is faster, and compared with a large number of neural network calculations, the matrix calculation can greatly improve the speed of merging calculation. Drawings FIG. 1 is a schematic diagram of a memory matrix-based intelligent alarm fingerprint merging method. Detailed Description The present invention will be described in further detail with reference to the accompanying drawings. The present invention will be described in detail with reference to the accompanying drawings. The invention provides an intelligent alarm fingerprint merging method based on a memory matrix in specific implementation, which comprises the following stages, (1) Alarm preprocessing stage And in the alarm processing stage, important alarm fields are mainly extracted, so that data can be better processed in the optimization process. In the alarm processing stage of the invention, a default dynamic alarm fingerprint can be used, and a custom alarm fingerprint can be generated by customizing key fields of an alarm sample, so that the data can better embody the combination result expected by a user. The method comprises the steps of classifying alarms according to user requirements to form a desired matrix- > redis storage desired matrix- > generating hash values through alarm key fields or user-defined fields to serve as alarm fingerprints- > converting the alarm fingerprints into alarm feature matrices. (2) Fingerprint optimization stage The redis memory matrix is used, so that a high-expansibility alarm merging scheme is created, and the problems that the cost is high and customization cannot be realized for users are well solved. The input alarm data extracts key fields