Method for extracting effective alarm data based on natural language features

Abstract

The invention discloses a method for extracting effective alarm data based on natural language features, which comprises a locality sensitive hash module and an editing distance algorithm module, and is characterized in that the locality sensitive hash algorithm module comprises a word segmentation module, a hash module, a weighting module, a merging module and a dimension reduction module; the word segmentation module is used for giving a section of statement, carrying out word segmentation to obtain effective feature vectors, and then setting weights of 5 levels such as 1-5 for each feature vector; the hash module calculates a hash value of each feature vector through a hash function; the weighting module is used for weighting all the feature vectors on the basis of the hash value; the merging module accumulates weighting results of the feature vectors to form only one sequence string; the dimension reduction module is used for obtaining a simhash value of a statement; and the editing distance algorithm module performs similarity clustering on the hash values.

Description

technical field [0001] The invention relates to the technical field of special equipment, in particular to a method for extracting effective alarm data based on natural language features. Background technique [0002] The access alarm methods of the traditional operation and maintenance system are complex, and there is no standard format for the alarm types sent by various business monitoring platforms. Under normal circumstances, the number of alarms that can be processed by natural persons is about 3 per minute. If a large number of alarms occur too frequently, an alarm storm will be formed, which will far exceed the limit of processing by natural persons, and key alarms will be lost, resulting in a delay in the processing of key issues. There will be serious production accidents and other problems. [0003] The current existing implementation schemes include: the alarm threshold setting generally adopts the method of traffic control. When the alarm data is below the thre...

AI Technical Summary

Problems solved by technology

[0005] Existing defects are realized through the above technologies: a large number of matching rules will seriously affect performance problems; after the alarm storm arrives, when the processing volume of alarms per second reaches the level of one thousand, the matching rules will cause a large amount of time delay, and the maximum delay may exceed 5-10 minutes, a large number of alarm processing lags will lead to catastrophic failures in the application system

[0007] The existing defects are realized through the above technologies: within the effective alarm processing range, the system processes alarms normally. When the alarm processing threshold exceeds the limit, the system does not receive alarms. Users of different types of subsequent alarms will not be able to receive them normally. If a critical alarm occurs, it will be discarded. It will lead to the interruption of subsequent processing, resulting in the loss of the basis for problem location, and the application system may have a catastrophic failure

Images