An alarm fingerprint intelligent merging method based on a memory matrix

Abstract

The application discloses an alarm fingerprint intelligent merging method based on a memory matrix, which comprises the following stages: an alarm preprocessing stage, a fingerprint optimization stage and an alarm merging stage. The above technical scheme has the following advantages: 1) the cost of intelligent alarm merging is reduced; 2) the memory matrix mode is used in the merging stage, the customizability of alarm merging is improved while the user configuration cost is reduced, and the demand of users in actual business scenarios can be better met; and 3) the processing speed is faster, and compared with a large number of neural network calculations, matrix calculation can greatly improve the speed of merging calculation.

Description

Technical Field

[0001] This invention relates to the field of alarm merging, specifically to an intelligent alarm fingerprint merging method based on a memory matrix. Background Technology

[0002] The main problems with existing SLS alarm intelligent merging technology lie in two aspects: alarm text clustering and clustering result merging.

[0003] (1) Alarm Text Clustering: Among the clustering algorithms used in the current SLS intelligent alarm merging technology, the MinJoin clustering algorithm calculates the text edit distance by using the local minimum hash value. This algorithm is expensive in terms of computation and storage requirements, and may have problems in noise reduction and high-dimensional data processing. At the same time, it cannot handle clusters of different sizes well.

[0004] (2) Clustering result merging: The clustering results do not depend on historical alarms and are automatically learned. Therefore, the merged results may not meet the user's needs, and the merged results may overfit, resulting in unsatisfactory noise reduction effect.

[0005] For the reasons mentioned above, a method for intelligent merging of alarm fingerprints based on memory matrices has become a technical problem that urgently needs to be solved by the whole society. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution provided by this invention is: an intelligent alarm fingerprint merging method based on a memory matrix, comprising the following stages,

[0007] (1) Alarm preprocessing stage: The alarm processing stage mainly extracts important alarm fields. You can use the default dynamic alarm fingerprint or customize the key fields of the alarm sample to generate a custom alarm fingerprint.

[0008] (2) In the fingerprint optimization stage, Redis is used to store the memory matrix, creating a highly scalable alarm merging scheme. Key fields are extracted from the input alarm data to generate hash values, and then the hash values ​​are converted into feature matrices. The expectation matrix is ​​multiplied by the inverse of the feature matrix to obtain the memory matrix array. This matrix array records the user's expectation information for the alarm. By continuously inputting the feature matrix and expectation matrix of the alarm data to train the memory matrix array, the memory matrix can be optimized to meet the actual needs of the customer's scenario. In the user's production environment, the newly entered matrix can be multiplied by the memory matrix to obtain the expectation matrix of the alarm. By calculating the expectation matrix, the probability of the alarm belonging to each category can be obtained.

[0009] (3) Alarm merging stage: In this stage, we convert newly entered alarms into feature matrices, multiply the feature matrices by the memory matrix group to obtain the expected matrix, and finally merge alarms of the same category in the expected matrix.

[0010] Furthermore, step (1) includes the following steps:

[0011] (1) Classify alarms according to user needs to form an expectation matrix;

[0012] (2) Redis stores the expectation matrix;

[0013] (3) Generate a hash value as an alarm fingerprint by using alarm key fields or user-defined fields;

[0014] (4) Convert the alarm fingerprint into an alarm feature matrix.

[0015] The advantages of the invention compared to existing technologies are:

[0016] The beneficial effects of this invention are as follows:

[0017] 1) Reduced the cost of intelligent alarm merging;

[0018] 2) Using a memory matrix during the merging phase reduces user configuration costs while improving the customizability of alarm merging, better meeting user needs in actual business scenarios.

[0019] 3) Faster processing speed: Compared to a large number of neural network calculations, matrix calculations can significantly improve the speed of combined calculations. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of a module of an alarm fingerprint intelligent merging method based on a memory matrix according to the present invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings.

[0022] The present invention will be described in detail with reference to the accompanying drawings.

[0023] In its specific implementation, this invention provides an intelligent alarm fingerprint merging method based on a memory matrix, comprising the following stages:

[0024] (1) Alarm preprocessing stage

[0025] The alarm processing phase primarily extracts key alarm fields to facilitate better data handling during optimization. In this invention's alarm processing phase, we can use the default dynamic alarm fingerprint or generate a custom alarm fingerprint by defining key fields of the alarm sample, allowing the data to better reflect the user's desired merging results.

[0026] Classify alarms according to user needs to form an expectation matrix -> Store the expectation matrix in Redis -> Generate hash values ​​as alarm fingerprints using alarm key fields or user-defined fields -> Convert alarm fingerprints into alarm feature matrices.

[0027] (2) Fingerprint optimization stage

[0028] We use Redis to store the memory matrix, creating a highly scalable alarm merging solution that effectively solves the aforementioned problems of high cost and inability to customize for users. Key fields are extracted from the input alarm data to generate hash values, which are then converted into feature matrices. The expected matrix is ​​multiplied by the inverse of the feature matrix to obtain a memory matrix array, which records the user's expected alarm information. By continuously inputting feature matrices and expected matrices from alarm data, the memory matrix array can be trained to meet the actual needs of the customer's scenario. In the user's production environment, a newly entered matrix is ​​multiplied by the memory matrix to obtain the expected matrix of the alarm. By calculating the expected matrix, the probability of the alarm belonging to each category can be obtained. Alarms merged using SLS intelligent alarm merging technology may not perfectly match the user's expectations. This method optimizes based on the user's expected result matrix, thus the final merged alarms better meet the user's expectations.

[0029] The reasons for using Redis with a memoization matrix are as follows:

[0030] 1) Redis's key-value pair storage structure is easy to read, write, and store, and is relatively fast.

[0031] 2) Using a memory matrix can reduce costs and improve user customization.

[0032] (3) Alarm merging stage

[0033] In this stage, we convert newly entered alarms into feature matrices, multiply the feature matrices by the memory matrix group to obtain the expected matrix, and finally merge alarms of the same category in the expected matrix.

[0034] The improvements are as follows:

[0035] (1) Use the memory matrix to calculate the alarm category;

[0036] (2) Use hash values ​​as alarm fingerprints to generate feature matrices;

[0037] (2) Record alarm category information through a memory matrix;

[0038] (3) Use Redis to store the matrix and alarm fingerprint information for fast reading and writing, thus improving efficiency.

[0039] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for intelligent merging of alarm fingerprints based on a memory matrix, characterized in that: It includes the following stages: (1) Alarm preprocessing stage: The alarm preprocessing stage mainly extracts important alarm fields. You can use the default dynamic alarm fingerprint or customize the key fields of the alarm sample to generate a custom alarm fingerprint. (2) In the fingerprint optimization stage, Redis is used to store the memory matrix. The expected matrix is ​​multiplied by the inverse of the feature matrix to obtain the memory matrix array. This matrix array records the user's expected information for the alarm. By continuously inputting the feature matrix and expected matrix of the alarm data, the memory matrix array can be trained to meet the actual needs of the customer. In the user's production environment, the newly entered matrix can be multiplied by the memory matrix to obtain the expected matrix of the alarm. By calculating the expected matrix, the probability of the alarm belonging to each category can be obtained. (3) Alarm merging stage: In this stage, newly entered alarms are converted into feature matrices. The feature matrices are multiplied by the memory matrix array to obtain the expected matrix. Finally, alarms of the same category in the expected matrix are merged. The alarm preprocessing stage includes the following steps: (1) Classify alarms according to user needs to form an expectation matrix; (2) Redis stores the expectation matrix; (3) Generate a hash value as an alarm fingerprint using alarm key fields or user-defined fields; (4) Convert the alarm fingerprint into an alarm feature matrix.

Images