Alarm level classification methods, devices and equipment

By constructing feature function models and unsupervised models to process alarm data in banking business scenarios, the problem of alarm data processing delay caused by the complexity of banking business is solved, and rapid and automatic alarm level identification and risk response are achieved.

CN119106365BActive Publication Date: 2026-11-17CHINA MERCHANTS BANK
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411122776.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2026-11-17
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

In existing technologies, the complex and dynamic nature of banking business scenarios makes it difficult for operations and maintenance personnel to handle a large number of alarms in a timely manner, resulting in the inability to process alarm data in a timely manner.

Method used

By pre-constructing a feature function model to calculate the alarm data, target feature values ​​are obtained. An unsupervised model is used to calculate the alarm level reference threshold. Based on the feature values ​​and threshold, the alarm level is identified, thus achieving automatic classification of alarm levels.

Benefits of technology

It enables real-time processing and identification of alarm data, quickly generates alarm level information, and helps operations and maintenance personnel respond to potential risks and problems in a timely manner, adapting to the complexity and dynamic changes of banking business.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119106365B_ABST
    Figure CN119106365B_ABST
Patent Text Reader

Abstract

This application discloses an alarm level classification method, apparatus, and device, relating to the field of intelligent operation and maintenance technology. The method includes: acquiring alarm data; calculating the alarm data using a pre-built feature function model to obtain target feature values; calculating the target feature values ​​using a pre-built unsupervised model to obtain an alarm level reference threshold; and performing level identification based on the target feature values ​​and the alarm level reference threshold to obtain alarm level information. This application can process alarm data in a timely manner.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology, and in particular to a method, device and equipment for alarm level classification. Background Technology

[0002] The time-series data anomaly alarm level classification assesses the severity of alarms by comprehensively analyzing characteristics such as the data volume, range, and degree of change within the alarm interval. This alarm level classification helps operations and maintenance personnel quickly locate and resolve more serious problems when indicators show abnormalities.

[0003] Currently, alarm level classification is usually based on predefined level rules after manually collecting various monitoring indicator data. However, banking business scenarios are diverse, many of which are complex and dynamically changing. Moreover, as business indicators grow, alarm types become more diverse, generating a large number of alarms every day, which makes it impossible for operations and maintenance personnel to handle them in a timely manner.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, apparatus and equipment for classifying alarm levels, which aims to solve the technical problem that alarm data cannot be processed in a timely manner by relying solely on human labor.

[0006] To achieve the above objectives, this application proposes an alarm level classification method, which includes:

[0007] Obtain alarm data;

[0008] The alarm data is calculated using a pre-built feature function model to obtain target feature values;

[0009] The alarm level reference threshold is obtained by calculating the target feature values ​​through a pre-built unsupervised model.

[0010] Alarm level information is obtained by identifying the level based on the target feature value and the alarm level reference threshold.

[0011] In one embodiment, the characteristic function model includes at least one of a stable decreasing alarm characteristic function model, a stable increasing alarm characteristic function model, and a discrete decreasing alarm characteristic function model.

[0012] The alarm data includes an alarm interval feature vector, which comprises the average percentage increase or decrease of anomalies within the alarm interval.

[0013] When the feature function model is the stability decrease alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model to calculate the alarm data includes:

[0014] The first intermediate feature value is calculated based on the feature vector of the alarm interval;

[0015] The first intermediate feature value is normalized.

[0016] The average percentage increase or decrease of abnormal points within the alarm interval is processed by segmented weighting.

[0017] The target feature value is calculated by combining the normalized first intermediate feature value and the average percentage increase or decrease of abnormal points within the alarm interval after segmented weighting.

[0018] In one embodiment, when the feature function model is the stable rising alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model on the alarm data includes:

[0019] The second intermediate feature value is calculated based on the feature vector of the alarm interval;

[0020] The second intermediate feature value is normalized to obtain the target feature value.

[0021] In one embodiment, when the feature function model is the discrete-type decreasing alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model on the alarm data includes:

[0022] Define a zero-value interval feature in the alarm interval feature vector to obtain a new alarm interval feature;

[0023] The dispersion ratio of the current alarm interval relative to its historical performance is calculated based on the new alarm interval characteristics.

[0024] Scaling the average percentage increase or decrease of abnormal points within the alarm interval;

[0025] The third intermediate feature value is calculated based on the ratio of the dispersion of the current alarm interval relative to the historical performance and the average percentage increase or decrease of abnormal points within the scaled alarm interval.

[0026] The third intermediate feature value is normalized to obtain the target feature value.

[0027] In one embodiment, the step of calculating the target feature value using a pre-built unsupervised model to obtain the alarm level reference threshold includes:

[0028] The target feature values ​​are calculated using a pre-constructed Gaussian mixture model to obtain the first cluster.

[0029] Labels are added to each of the first clusters based on the mean of each first cluster, resulting in the first clusters after labeling.

[0030] The bounding values ​​of the mean of the first cluster after adding labels are supplemented based on the preset fine-grained interval to obtain the expanded dataset;

[0031] The second cluster is obtained by calculating the augmented dataset using the Gaussian mixture model.

[0032] The alarm level reference threshold is calculated based on the second cluster.

[0033] In one embodiment, the step of calculating the alarm level reference threshold based on the second cluster includes:

[0034] Determine whether the alarm level reference threshold reaches the preset alarm level classification standard;

[0035] If not, after retraining the Gaussian mixture model, the following steps are performed: calculate the target feature values ​​using the pre-built Gaussian mixture model to obtain the first cluster.

[0036] In one embodiment, the alarm level reference threshold includes a maximum feature value and a minimum feature value, and the step of performing level identification based on the target feature value and the alarm level reference threshold to obtain alarm level information includes:

[0037] Determine whether the target feature value is greater than the maximum feature value;

[0038] If so, output the severity level information;

[0039] If not, then determine whether the target feature value is less than the minimum feature value;

[0040] If so, output the level information for mild severity;

[0041] If not, output the level information for the moderate level.

[0042] In one embodiment, the alarm data includes alarm history information, and after the step of performing level identification based on the target feature value and the alarm level reference threshold to obtain alarm level information, the method further includes:

[0043] Determine whether a persistent alarm level was marked the previous day;

[0044] If so, and an alarm occurs on the same day, then the previous persistent flag will be inherited, and the persistent alarm level will be marked.

[0045] If so, and no alarm occurs on that day, the continuous alarm level is lifted, and the continuous flag is set to a non-continuous state value;

[0046] If not, then determine whether there are alarms of a severity level in the alarm history information;

[0047] If the alarm history information contains alarm information of a severe level, and an alarm occurs on the same day, then the persistence flag is set to a severe persistence state value to mark the persistent alarm level;

[0048] If there are no serious alarms in the alarm history information, and an alarm occurs on the same day, the continuous flag is incremented by 1. When the accumulated continuous flag exceeds a preset continuous flag threshold, a continuous alarm level is marked.

[0049] Furthermore, to achieve the above objectives, this application also proposes an alarm level classification device, which includes:

[0050] The data acquisition module is used to acquire alarm data;

[0051] The feature calculation module is used to calculate the target feature value by using a pre-built feature function model on the alarm data;

[0052] The threshold calculation module is used to calculate the target feature value using a pre-built unsupervised model to obtain the alarm level reference threshold.

[0053] The alarm level identification module is used to identify the alarm level based on the target feature value and the alarm level reference threshold to obtain alarm level information.

[0054] In addition, to achieve the above objectives, this application also proposes an alarm level classification device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the alarm level classification method as described above.

[0055] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the alarm level classification method described above.

[0056] One or more technical solutions proposed in this application have at least the following technical effects:

[0057] This application calculates target feature values ​​from the alarm data using a pre-built feature function model, and then calculates alarm level reference thresholds from these target feature values ​​using a pre-built unsupervised model. Based on the target feature values ​​and the alarm level reference thresholds, alarm level information is obtained through level identification. This allows for real-time processing and identification of alarm data, and rapid generation of alarm level information. This is particularly important for alarm scenarios requiring rapid response and processing, helping to take timely measures to address potential risks and problems. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating an embodiment of the alarm level classification method in this application.

[0061] Figure 2 This is a schematic diagram of the stationary sequence of the alarm level classification method in this application;

[0062] Figure 3 This is a schematic diagram of the discrete sequence of alarm level classification method in this application. Figure 1 ;

[0063] Figure 4 This is a schematic diagram of the discrete sequence of alarm level classification method in this application. Figure 2 ;

[0064] Figure 5 This is a flowchart illustrating the process of calculating the alarm level reference threshold using the alarm level classification method of this application.

[0065] Figure 6 This is a flowchart illustrating the alarm level classification method and level identification process in this application.

[0066] Figure 7 This is a schematic diagram of the overall process of the alarm level classification method in this application;

[0067] Figure 8 This is a schematic diagram of the module structure of the alarm level classification device according to an embodiment of this application;

[0068] Figure 9This is a schematic diagram of the device structure of the hardware operating environment involved in the alarm level classification method in this application embodiment.

[0069] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0071] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0072] The main solution of this application embodiment is: to acquire alarm data; to calculate the alarm data using a pre-built feature function model to obtain target feature values; to calculate the target feature values ​​using a pre-built unsupervised model to obtain alarm level reference thresholds; and to perform level identification based on the target feature values ​​and the alarm level reference thresholds to obtain alarm level information.

[0073] In this embodiment, for ease of description, the alarm level classification device will be used as the subject of the explanation below.

[0074] Due to the diverse range of banking business scenarios, many of which are complex and dynamic, and the increasing number of alarm types as business metrics grow, a large number of alarms are generated every day, making it impossible for operations and maintenance personnel to handle them in a timely manner.

[0075] This application provides a solution that calculates target feature values ​​from alarm data using a pre-built feature function model, then calculates alarm level reference thresholds from these target feature values ​​using a pre-built unsupervised model, and finally identifies alarm level information based on the target feature values ​​and the reference thresholds. This approach enables real-time processing and identification of alarm data and rapid generation of alarm level information. This is particularly important for alarm scenarios requiring rapid response and processing, facilitating timely measures to address potential risks and problems.

[0076] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or alarm level classification device capable of performing the above functions. The following description uses an alarm level classification device as an example to illustrate this embodiment and the subsequent embodiments.

[0077] Based on this, the embodiments of this application provide a method for classifying alarm levels, referring to... Figure 1 , Figure 1This is a flowchart illustrating the alarm level classification method of this application in Implementation Example 1.

[0078] In this embodiment, the alarm level classification method includes steps S10 to S40:

[0079] Step S10: Obtain alarm data;

[0080] It should be noted that alarm data is a detailed record captured and logged by the monitoring system regarding abnormal operating status of the system, network, application, or device. This includes, but is not limited to, the time and location of the alarm (e.g., IP address, device ID), type (e.g., hardware failure, software error, security threat), severity, relevant parameters (e.g., performance metrics, error codes), and possible solutions or recommendations. Furthermore, alarm data may also include feature vector data specific to the alarm scenario for further analysis and alarm pattern identification, as well as historical alarm data for time series analysis, trend prediction, and pattern recognition.

[0081] Alarm data also includes performance metric alarms, security alarms, log alarms, dependency alarms, custom alarms, related alarms, and predictive alarms.

[0082] Step S20: Calculate the alarm data using a pre-built feature function model to obtain the target feature value;

[0083] It should be noted that this embodiment defines multiple function models to adapt to complex timing alarms in a variety of business scenarios.

[0084] One approach is to use pre-built feature function models to address complex timing alarms generated by diverse business scenarios.

[0085] Furthermore, the feature function models include a falling alarm feature function model for stationary sequences, a rising alarm feature function model, and a falling alarm feature function model for discrete sequences.

[0086] In another implementation, based on the alarm interval feature vector data, multiple algorithms can be combined to fit an S-shaped function for mapping processing. This improves the influence of important features while reducing the vector dimension and the training cost of the unsupervised model. Subsequently, the alarm level reference threshold output by the function model is transformed by the mapping function in the online level recognition module, and the final level information is output.

[0087] Step S30: Calculate the target feature value using a pre-built unsupervised model to obtain the alarm level reference threshold;

[0088] The unsupervised approach eliminates the need for manually labeled training data, and the feature value calculation stage fully considers the data performance characteristics of various sequence types. Furthermore, during the training and calculation of alarm level thresholds, the threshold distribution can be fine-tuned by expanding alarm data of specific severity levels according to actual business needs, thus optimizing the grading effect. Compared to manual labeling or customized rules, the unsupervised model-based adaptive division of alarm level threshold ranges is more efficient.

[0089] As one implementation method, the clusters obtained from training the offline module can be used as a pre-built unsupervised model, and the alarm level reference threshold can be calculated through the clusters.

[0090] Step S40: Based on the target feature value and the alarm level reference threshold, perform level identification to obtain alarm level information.

[0091] It can perform real-time level recognition, continuous level recognition, or a combination of real-time level recognition and continuous level recognition.

[0092] One implementation method is to combine an online alarm level identification module with a continuous alarm level analysis component to achieve the final alarm level identification and score description output.

[0093] This embodiment provides a method for classifying alarm levels. The method calculates target feature values ​​from the alarm data using a pre-built feature function model, then calculates alarm level reference thresholds from these target feature values ​​using a pre-built unsupervised model. Based on the target feature values ​​and the alarm level reference thresholds, level identification is performed to obtain alarm level information. This method enables real-time processing and identification of alarm data, and rapid generation of alarm level information. This is particularly important for alarm scenarios requiring rapid response and processing, helping to take timely measures to address potential risks and problems.

[0094] Furthermore, this embodiment summarizes and analyzes the rich types and characteristics of time series data in banking business scenarios. By integrating various features, a general and adaptive time series alarm level classification method is designed and implemented, which can adapt to various time series types, including stationary, irregular, and discrete ones. Alarm level information is obtained by fitting level scores using a custom feature function model and then training the classification using an unsupervised clustering model.

[0095] Based on Embodiment 1 of this application, in Embodiment 2 of this application, the content that is the same as or similar to that in Embodiment 1 can be referred to the above description, and will not be repeated hereafter. In addition, the characteristic function model includes at least one of a stable decreasing alarm characteristic function model, a stable increasing alarm characteristic function model, and a discrete decreasing alarm characteristic function model.

[0096] The alarm data includes an alarm interval feature vector, which includes the average percentage increase or decrease of abnormal points within the alarm interval. The alarm interval feature vector can be represented as AlarmVec = [num, count, percent], where num is the magnitude of the average statistical value of abnormal points within the alarm interval, count is the number of abnormal points within the alarm interval, and percent is the average percentage increase or decrease of abnormal points within the alarm interval.

[0097] When the feature function model is the stability decrease alarm feature function model, step S20, calculating the alarm data using the pre-constructed feature function model to obtain the target feature value, further includes steps S21 to S24:

[0098] Step S21: Calculate the first intermediate feature value based on the feature vector of the alarm interval;

[0099] It should be noted that stationary sequences have continuous values, clear periods and trends, relatively small fluctuations, and large overall numerical magnitudes. Stationary sequences exhibit the following characteristics: Figure 2 As shown.

[0100] Step S22: Normalize the first intermediate feature value;

[0101] Step S23: Perform segmented weighted processing on the average percentage increase or decrease of abnormal points within the alarm interval;

[0102] Step S24: Combine the normalized first intermediate feature value and the average percentage increase or decrease of abnormal points within the alarm interval after segmented weighting to calculate the target feature value.

[0103] The stationary decreasing alarm feature function model is applicable to the calculation of feature values ​​for decreasing alarm intervals in stationary sequences. The specific steps and methods are as follows:

[0104] (1) Input the alarm interval feature vector AlarmVec=[num,count,percent], and calculate the first intermediate feature value of area according to (Formula 1).

[0105]

[0106] (2) Normalization is performed. The MinMaxScalar model trained on the training set is used to scale the first intermediate feature value to the (0,1) range to avoid numerical out-of-bounds issues.

[0107] (3) According to (Formula 2), the percent value is processed by segment weighting.

[0108]

[0109] (4) Combine the first intermediate feature value of the normalized area with the percent′ value, and calculate the target feature value size according to (Formula 3).

[0110] size=percent′×area (formula 3)

[0111] This application embodiment calculates the target feature value by using a stability decline alarm function model to calculate the matched alarm data, thereby providing a more accurate calculation model and rationally allocating resources.

[0112] Based on Embodiment 2 of this application, in Embodiment 3 of this application, the content that is the same as or similar to Embodiment 2 can be referred to the above description, and will not be repeated hereafter. On this basis, when the feature function model is the stable rising alarm feature function model, step S20, calculating the alarm data using the pre-constructed feature function model to obtain the target feature value, further includes steps S25 to S26:

[0113] Step S25: Calculate the second intermediate feature value based on the feature vector of the alarm interval;

[0114] Step S26: Normalize the second intermediate feature value to obtain the target feature value.

[0115] Among them, the stationary rising alarm feature function model can be applied to the calculation of feature values ​​for rising alarm intervals in stationary sequences. The stationary rising alarm feature function model can be understood as fitting the volume of the alarm interval; the larger the volume, the more severe the alarm.

[0116] The specific steps and methods are as follows:

[0117] (1) Input the alarm interval feature vector AlarmVec=[num,count,percent], and calculate the second intermediate feature value size′ according to (Formula 4).

[0118] size′=(0.5×log 100 (min(count,1440)+1)×(log 100 (num+1)) 3 +0.5×log 10 percent) 2 (Formula 4)

[0119] (2) Normalize size′ and use the MinMaxScalar model trained on the training set to scale the second intermediate feature value to the (0,1) interval and output the target feature value of size.

[0120] This application embodiment calculates the target feature value by using a steady-state rising alarm function model to match alarm data, thereby providing a more accurate calculation model and rationally allocating resources.

[0121] Based on Embodiment 2 of this application, in Embodiment 4 of this application, the content that is the same as or similar to Embodiment 2 can be referred to the above description, and will not be repeated hereafter. On this basis, when the feature function model is the discrete decreasing alarm feature function model, step S20, calculating the alarm data using the pre-constructed feature function model to obtain the target feature value, further includes steps S27 to S211:

[0122] Step S27: Define a zero-value interval feature in the alarm interval feature vector to obtain a new alarm interval feature;

[0123] It should be noted that discrete sequences exhibit a discrete distribution overall, with no obvious pattern or periodicity, and strong random fluctuations. The characteristics of discrete sequences are as follows: Figure 3 and Figure 4 As shown.

[0124] For this type of sequence, this module defines a zero-value interval feature in the alarm interval feature vector. cur zero ref Replace the statistics num and the number of outliers within the alarm interval in the stationary vector.

[0125] Among them, zero cur The actual length of the continuous interval of zero values, zero ref To predict the length of the zero-value interval. The zero-value interval feature can be used to measure and analyze abnormal states of discrete sequences, such as service interruptions and failures. The zero-value interval is the interval of the sequence at time t. i The length of the consecutive 0-value region contained in the preceding time step, counted in minutes.

[0126] Step S28: Calculate the dispersion ratio of the current alarm interval relative to historical performance based on the new alarm interval characteristics;

[0127] Step S29: Scale the average percentage increase or decrease of abnormal points within the alarm interval;

[0128] Step S210: Calculate the third intermediate feature value based on the ratio of the dispersion of the current alarm interval relative to the historical performance and the average percentage increase or decrease of abnormal points in the scaled alarm interval.

[0129] Step S211: Normalize the third intermediate feature value to obtain the target feature value.

[0130] The specific steps and methods are as follows:

[0131] (1) Input the discrete sequence alarm interval feature vector AlarmVec = [zero ref ,zero cur [,percent], calculate the area feature value according to (Formula 5). area is the ratio of the dispersion of the current alarm interval relative to the historical performance.

[0132]

[0133] (2) Further scale percent and zero ref Input the numerical value (Formula 6) to calculate and generate percent′.

[0134]

[0135] (3) Calculate and generate the third intermediate feature value size′ according to (Formula 7).

[0136] size′=(0.8×area+0.2×percent′) 2 (Formula 7)

[0137] (4) Normalize the third intermediate feature value size′. Using the MinMaxScalar model trained on the training set, scale the third intermediate feature value to the (0,1) interval and output the target feature value size.

[0138] This application proposes a level classification solution based on a small-scale strategy to address the pain points of irregular and discrete alarm level classification problems.

[0139] Based on Embodiment 1 of this application, in Embodiment 5 of this application, the content that is the same as or similar to that in Embodiment 1 can be referred to the above description, and will not be repeated hereafter. Based on this, step S30, calculating the target feature value using a pre-built unsupervised model to obtain the alarm level reference threshold, further includes steps S31 to S35:

[0140] Step S31: Calculate the target feature values ​​using a pre-constructed Gaussian mixture model to obtain the first cluster;

[0141] In this embodiment, a Gaussian mixture model (GMM) is used to perform offline and unsupervised clustering on the target feature values ​​of various alarm intervals. Then, the alarm level reference threshold is calculated using the boundaries of each cluster to provide a basis for judgment for the online alarm level identification module.

[0142] It should be noted that Gaussian mixture models can accurately quantify things using Gaussian density functions, decomposing a thing into several sub-distributions based on Gaussian probability density functions, and have good mathematical properties and good computational performance.

[0143] Before using the Gaussian mixture model to calculate the target feature value, it is necessary to train the Gaussian mixture model GmmModel. The initialization parameters of the Gaussian mixture model are generated based on the K-means algorithm, which converges faster and performs better than the original randomization method.

[0144] Step S32: Add labels to each of the first clusters based on the mean of each first cluster to obtain the first clusters after adding labels;

[0145] Step S33: Based on a preset fine-grained interval, supplement the boundary values ​​of the mean of the first cluster after adding labels to obtain an expanded dataset.

[0146] Step S34: Calculate the augmented dataset using the Gaussian mixture model to obtain the second cluster;

[0147] Step S35: Calculate the alarm level reference threshold based on the second cluster.

[0148] The algorithm steps for training and calculating the alarm level threshold range using a Gaussian mixture model are as follows:

[0149] (1) Compare the mean values ​​of each cluster, add labels to each cluster dataset in ascending order, and output the set of each cluster, i.e., the first cluster: C light C medium C serious And extract the mean of each of the first clusters: mean light mean medium mean serious .

[0150] (2) Based on (Formula 8) to (Formula 10), expand the values ​​between the means boundaries of the first cluster. Where interval = 0.001, supplement the data in a fine-grained manner at intervals to calculate a more accurate rank threshold.

[0151] S light = np.arange(mean) light ,mean medium ,int erval) (Formula 8)

[0152] S serious = np.arange(mean) medium ,meanserious ,int erval) (Formula 9)

[0153] S extend =S+S light +S serious (Formula 10)

[0154] (3) Expand the dataset S extend Input the GmmModel model again, and the second cluster is calculated: C′ light C′ medium C′ serious .

[0155] (4) Based on the second cluster, calculate the output alarm level threshold range levelThreshold according to (Formula 11), i.e., C′ medium The maximum and minimum size values ​​in the set.

[0156] levelThreshold=[C′ medium .size.min,C′ medium .size.max](Formula 11)

[0157] In this embodiment, the target feature value is calculated using a Gaussian mixture model to obtain an alarm level reference threshold. Since the Gaussian mixture model can accurately quantify things using a Gaussian density function, decomposing a thing into several sub-distributions based on the Gaussian probability density function, and has good mathematical properties and good computational performance, the accuracy of the alarm level reference threshold can be improved.

[0158] Based on Embodiment 5 of this application, in Embodiment 6 of this application, the content that is the same as or similar to Embodiment 5 can be referred to the above description, and will not be repeated hereafter. Based on this, after step S35, which calculates the alarm level reference threshold based on the second cluster, steps S36 to S37 are further included:

[0159] Step S36: Determine whether the alarm level reference threshold reaches the preset alarm level classification standard;

[0160] Step S37: If not, after retraining the Gaussian mixture model, perform the following step: calculate the target feature value using the pre-constructed Gaussian mixture model to obtain the first cluster.

[0161] Specifically, the alarm level reference threshold levelThreshold is verified to meet the preset alarm level classification standard. If it meets the requirements, the alarm level reference threshold is applied to the online level classification module. Otherwise, the sample data is expanded according to the actual business needs, the Gaussian mixture model is retrained, and then step S31 is executed based on the retrained Gaussian mixture model to fine-tune the level threshold classification effect.

[0162] To aid in understanding Embodiments 5 and 6 of this application, please refer to Figure 5 , Figure 5 This is a flowchart illustrating the process of calculating the alarm level reference threshold using the alarm level classification method of this application.

[0163] Step S311: Obtain the Size feature value training dataset (i.e., the intermediate feature value training dataset);

[0164] Step S312: Input the Size feature value training dataset into the GMM model based on the initial parameters of the K-means algorithm to train the GMM model.

[0165] Step S313: Calculate the target feature values ​​using the trained GMM model, and obtain the level threshold range (i.e., alarm level reference threshold) based on the cluster boundary value set. At this point, the sample weight of the training dataset for intermediate feature values ​​can be adjusted according to actual business needs, thereby retraining the GMM model and fine-tuning the level threshold distribution through the GMM model.

[0166] Step S314: Verify whether the alarm level reference threshold meets the reasonable alarm level classification standard. If so, output the alarm level threshold range (i.e., the alarm level reference threshold).

[0167] Step S315: If not, readjust the training dataset for intermediate feature values ​​and then proceed to step S311.

[0168] This application embodiment improves the accuracy of the alarm level reference threshold by determining whether the alarm level reference threshold meets the preset alarm level classification standard after calculating the alarm level reference threshold.

[0169] Based on Embodiment 1 of this application, in Embodiment 7 of this application, the content that is the same as or similar to that in Embodiment 1 can be referred to the above description, and will not be repeated hereafter. Based on this, the alarm level reference threshold includes a maximum feature value and a minimum feature value. Step S40, based on the target feature value and the alarm level reference threshold, performs level identification to obtain alarm level information, and further includes steps S41 to S45:

[0170] Step S41: Determine whether the target feature value is greater than the maximum feature value;

[0171] Step S42: If yes, output the severity level information;

[0172] Step S43: If not, determine whether the target feature value is less than the minimum feature value;

[0173] Step S44: If yes, output the level information of the mild level;

[0174] Step S45: If not, output the level information for the moderate level.

[0175] Specifically, after real-time alarms are triggered by anomaly detection in time series data, feature vectors of the alarm interval are extracted, and target feature values ​​are calculated by adapting them to the corresponding feature function model based on the sequence type.

[0176] The alarm level reference threshold is obtained by calculating the target feature values ​​using a pre-built unsupervised model.

[0177] Then, based on the alarm level reference threshold comparison, the interval level is identified, and the level description and level score are output. The specific calculation process is as follows: Figure 6 As shown:

[0178] Step S411: Obtain the alarm interval feature vector AlamVec.

[0179] Step S412: Calculate the size feature value (i.e., the target feature value) based on the alarm interval feature vector.

[0180] Step S413: Calculate the target feature value using a pre-built unsupervised model to obtain the alarm level reference threshold, including the maximum value C.size.max and the minimum value C.size.min.

[0181] Step S414: Determine whether the target feature value size is greater than C.size.max.

[0182] Step S415: If yes, output (severity level, size).

[0183] Step S416: If not, determine whether the target feature value size is less than C.size.min.

[0184] Step S417, if yes, output (mild level, size).

[0185] Step S418: If not, output (moderate level, size).

[0186] This application embodiment identifies the alarm level range by comparing alarm level reference thresholds, and outputs the level description and level score, thereby completing a simple and efficient online alarm level identification.

[0187] Based on Embodiment 7 of this application, in Embodiment 8 of this application, the content that is the same as or similar to that in Embodiment 7 can be referred to the above description, and will not be repeated hereafter. Based on this, the alarm data includes alarm history information. After step S40, which involves level identification based on the target feature value and the alarm level reference threshold to obtain alarm level information, steps S50 to S100 are further included:

[0188] Step S50: Determine whether the continuous alarm level was marked the previous day;

[0189] In particular, Embodiment 8 of this application describes the steps of performing continuous level identification. Based on the historical level of the sequence, continuous level identification performs a time-series continuous analysis to identify alarm intervals with frequent fluctuations. This level indicates that the sequence has frequently experienced abnormal alarms recently rather than new ones, which can reduce attention.

[0190] Specifically, the persistence level analysis determines whether the current alarm interval belongs to the long-term continuous alarm type based on alarm history information. The persistence level is the level that identifies the duration of the alarm.

[0191] Furthermore, the continuous level analysis determines whether the current alarm interval belongs to the long-term continuous alarm type based on the alarm history information of the previous day.

[0192] Step S60: If so, and an alarm occurs on the same day, then inherit the previous persistent flag and mark the persistent alarm level;

[0193] If a continuous alarm level has already been marked and an alarm has occurred on the same day, the historical continuous flag bit of the sequence will be inherited and the alarm level will continue to be marked as continuous.

[0194] Step S70: If yes, and no alarm occurs on that day, then the continuous alarm level is lifted, and the continuous flag is set to a non-continuous state value.

[0195] If a continuous alarm level has been marked and no alarms have occurred on the current day, the continuous alarm level will be cleared and the continuous flag will be set to a non-continuous state value, such as 0. This allows us to know the current alarm status based on the continuous flag.

[0196] Step S80: If not, determine whether there are alarm messages of a severity level in the alarm history information;

[0197] Step S90: If the alarm history information contains alarm information of a severe level and an alarm occurs on the same day, then set the persistence flag to a severe persistence state value to mark the persistent alarm level.

[0198] The value for severe persistent state can be 5.

[0199] Step S100: If there are no serious alarms in the alarm history information and an alarm occurs on the same day, then increment the continuous flag by 1. When the accumulated continuous flag is greater than the preset continuous flag threshold, mark the continuous alarm level.

[0200] For example, when the accumulated continuous flags exceed the preset continuous flag threshold of 3, a continuous alarm level is marked.

[0201] This application's embodiments design a time-dimensional persistent level classification scheme, which can identify persistent alarms that do not require special attention due to frequent fluctuations.

[0202] For example, to help understand the implementation process of the alarm level classification method obtained by combining this embodiment with the above embodiment one, please refer to 7. Figure 7 A complete flowchart of an alarm level classification method is provided.

[0203] The alarm level classification method includes modules such as a custom multi-class feature function model, offline clustering model training, online real-time score conversion and level recognition, and continuous level analysis.

[0204] Specifically, the complete process of alarm level classification includes the following steps:

[0205] Step S101: Obtain alarm data through the data interface. The alarm data includes alarm feature vectors and historical alarm information.

[0206] Step S102: Calculate the target feature value based on the alarm feature vector adapted to the corresponding feature function model. The corresponding feature function models include the stable decreasing alarm size function model (i.e., the stable decreasing alarm feature function model), the stable increasing alarm size function model (i.e., the stable increasing alarm feature function model), and the discrete decreasing alarm size function model (i.e., the discrete decreasing alarm feature function model).

[0207] Step S103: The target feature value is calculated by the offline unsupervised level classification module to obtain the alarm level reference threshold.

[0208] Step S104: The online level recognition module calculates the target feature value and alarm level reference threshold, and outputs the alarm level, level score, and level description. The online level recognition module includes online real-time score conversion and level recognition, and continuous level analysis. The alarm levels include severe, moderate, mild, and continuous levels.

[0209] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the alarm level classification method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0210] This application also provides an alarm level classification device, please refer to... Figure 8 The alarm level classification device includes:

[0211] Data acquisition module 10 is used to acquire alarm data;

[0212] Feature calculation module 20 is used to calculate the target feature value by using a pre-built feature function model on the alarm data;

[0213] The threshold calculation module 30 is used to calculate the target feature value through a pre-built unsupervised model to obtain the alarm level reference threshold;

[0214] The alarm level identification module 40 is used to identify the alarm level based on the target feature value and the alarm level reference threshold to obtain alarm level information.

[0215] The alarm level classification device provided in this application, employing the alarm level classification method in the above embodiments, can solve the technical problem that alarms cannot be processed in a timely manner by manpower alone. Compared with the prior art, the beneficial effects of the alarm level classification device provided in this application are the same as those of the alarm level classification method provided in the above embodiments, and other technical features in the alarm level classification device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0216] This application provides an alarm level classification device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the alarm level classification method in the above embodiment 1.

[0217] The following is for reference. Figure 9The diagram illustrates a structural schematic suitable for implementing an alarm level classification device according to embodiments of this application. The alarm level classification device in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 9 The alarm level classification device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0218] like Figure 9 As shown, the alarm level classification device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the alarm level classification device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the alarm level classification device to communicate wirelessly or wiredly with other devices to exchange data. Although alarm level classification devices with various systems are shown in the figure, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0219] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0220] The alarm level classification device provided in this application, employing the alarm level classification method in the above embodiments, can solve the technical problem that current methods cannot handle alarms in a timely manner relying solely on manpower. Compared with the prior art, the beneficial effects of the alarm level classification device provided in this application are the same as those of the alarm level classification method provided in the above embodiments, and other technical features of this alarm level classification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0221] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0222] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0223] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the alarm level classification method in the above embodiments.

[0224] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0225] The aforementioned computer-readable storage medium may be included in the alarm level classification device; or it may exist independently and not be assembled into the alarm level classification device.

[0226] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the alarm level classification device, the alarm level classification device causes the following actions: acquiring alarm data; calculating the alarm data using a pre-built feature function model to obtain target feature values; calculating the target feature values ​​using a pre-built unsupervised model to obtain alarm level reference thresholds; and performing level identification based on the target feature values ​​and the alarm level reference thresholds to obtain alarm level information.

[0227] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0228] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0229] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0230] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described alarm level classification method, which can solve the technical problem that alarms cannot be processed in a timely manner by human intervention alone. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the alarm level classification method provided in the above embodiments, and will not be repeated here.

[0231] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method of alarm level classification, characterized by, The method includes: Obtain alarm data; The alarm data is calculated using a pre-built feature function model to obtain target feature values; The alarm level reference threshold is obtained by calculating the target feature values ​​using a pre-built unsupervised model. Specifically, this includes: calculating the target feature values ​​using a pre-built Gaussian mixture model to obtain a first cluster; adding labels to each of the first clusters based on their mean values ​​to obtain labeled first clusters; supplementing the boundary values ​​of the labeled first clusters with finer-grained intervals based on preset intervals to obtain an expanded dataset; calculating the expanded dataset using the Gaussian mixture model to obtain a second cluster; and calculating the alarm level reference threshold based on the second clusters. Alarm level information is obtained by identifying the level based on the target feature value and the alarm level reference threshold.

2. The method of claim 1, wherein, The characteristic function model includes at least one of the following: a stable decreasing alarm characteristic function model, a stable increasing alarm characteristic function model, and a discrete decreasing alarm characteristic function model. The alarm data includes an alarm interval feature vector, which comprises the average percentage increase or decrease of anomalies within the alarm interval. When the feature function model is the stable decreasing alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model to calculate the alarm data includes: The first intermediate feature value is calculated based on the feature vector of the alarm interval; The first intermediate feature value is normalized. The average percentage increase or decrease of abnormal points within the alarm interval is processed by segmented weighting. The target feature value is calculated by combining the normalized first intermediate feature value and the average percentage increase or decrease of abnormal points within the alarm interval after segmented weighting.

3. The method of claim 2, wherein, When the feature function model is the stable rising alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model to calculate the alarm data includes: The second intermediate feature value is calculated based on the feature vector of the alarm interval; The second intermediate feature value is normalized to obtain the target feature value.

4. The method of claim 2, wherein, When the feature function model is the discrete-type decreasing alarm feature function model, the step of calculating the target feature value by using the pre-constructed feature function model to calculate the alarm data includes: Define a zero-value interval feature in the alarm interval feature vector to obtain a new alarm interval feature; The dispersion ratio of the current alarm interval relative to its historical performance is calculated based on the new alarm interval characteristics. Scaling the average percentage increase or decrease of abnormal points within the alarm interval; The third intermediate feature value is calculated based on the ratio of the dispersion of the current alarm interval relative to the historical performance and the average percentage increase or decrease of abnormal points within the scaled alarm interval. The third intermediate feature value is normalized to obtain the target feature value.

5. The method of claim 1, wherein, The step of calculating the alarm level reference threshold based on the second cluster includes: Determine whether the alarm level reference threshold reaches the preset alarm level classification standard; If not, after retraining the Gaussian mixture model, the following steps are performed: calculate the target feature values ​​using the pre-built Gaussian mixture model to obtain the first cluster.

6. The method of claim 1, wherein, The alarm level reference threshold includes a maximum feature value and a minimum feature value. The step of identifying the alarm level based on the target feature value and the alarm level reference threshold to obtain alarm level information includes: Determine whether the target feature value is greater than the maximum feature value; If so, output the severity level information; If not, then determine whether the target feature value is less than the minimum feature value; If so, output the level information for mild severity; If not, output the level information for the moderate level.

7. The method of claim 6, wherein, The alarm data includes alarm history information. Following the step of obtaining alarm level information by performing level identification based on the target feature value and the alarm level reference threshold, the method further includes: Determine whether a persistent alarm level was marked the previous day; If so, and an alarm occurs on the same day, then the previous persistent flag will be inherited, and the persistent alarm level will be marked. If so, and no alarm occurs on that day, the continuous alarm level is lifted, and the continuous flag is set to a non-continuous state value; If not, then determine whether there are alarms of a severity level in the alarm history information; If the alarm history information contains alarm information of a severe level, and an alarm occurs on the same day, then the persistence flag is set to a severe persistence state value to mark the persistent alarm level; If there are no serious alarms in the alarm history information, and an alarm occurs on the same day, the continuous flag is incremented by 1. When the accumulated continuous flag exceeds a preset continuous flag threshold, a continuous alarm level is marked.

8. An alarm level dividing apparatus characterized by comprising: The device includes: The data acquisition module is used to acquire alarm data; The feature calculation module is used to calculate the target feature value by using a pre-built feature function model on the alarm data; The threshold calculation module is used to calculate the target feature values ​​using a pre-built unsupervised model to obtain an alarm level reference threshold. Specifically, it includes: calculating the target feature values ​​using a pre-built Gaussian mixture model to obtain a first cluster; adding labels to each of the first clusters based on their mean values ​​to obtain labeled first clusters; supplementing the boundary values ​​of the labeled first clusters with finer-grained intervals based on preset intervals to obtain an expanded dataset; calculating the expanded dataset using the Gaussian mixture model to obtain a second cluster; and calculating the alarm level reference threshold based on the second clusters. The alarm level identification module is used to identify the alarm level based on the target feature value and the alarm level reference threshold to obtain alarm level information.

9. An alarm level classification device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the alarm level classification method as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Alarm data processing method and device, electronic equipment and readable storage medium

    CN115460065A

  • Cloud database intelligent alarm method based on machine learning

    CN117851185A