Alarm classification method and device, electronic equipment and medium

Abstract

The application provides an alarm classification method and device, electronic equipment and medium. The method comprises: obtaining feature data of to-be-processed alarm information, the feature data comprising a first feature variable representing alarm duration, a second feature variable representing customer type, a third feature variable representing alarm site type, and a fourth feature variable; wherein the fourth feature variable is used to represent the importance of the alarm type to which the to-be-processed alarm information belongs; inputting the feature data of the to-be-processed alarm information into an alarm classification model to obtain an alarm level of the to-be-processed alarm information output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to represent the urgency of the to-be-processed alarm information. The method of the application efficiently realizes accurate classification of the urgency of the to-be-processed alarm information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to an alarm classification method and device, electronic equipment and medium. BACKGROUND

[0002] With the full spread of 5G network and cloud infrastructure construction, the number of network resources will increase exponentially, and the number of network alarm information is also increasing. Therefore, in order to improve the efficiency of fault handling, it is necessary to distinguish the urgency of alarm information.

[0003] In the prior art, after the system receives network management alarm information, the dispatch personnel judges the urgency of the alarm information, and generates a fault work order to be distributed to the maintenance user.

[0004] In the above scheme, the alarm classification is performed by manual operation, which has the problem of low efficiency. SUMMARY

[0005] The present application provides an alarm classification method, device, electronic equipment and medium, which solves the problem of low efficiency of classifying the urgency of alarm information.

[0006] In one aspect, the present application provides an alarm classification method, comprising: obtaining feature data of to-be-processed alarm information, the feature data comprising a first feature variable representing alarm duration, a second feature variable representing customer type, a third feature variable representing alarm site type, and a fourth feature variable; wherein the fourth feature variable is used to represent the importance of the alarm type to which the to-be-processed alarm information belongs; inputting the feature data of the to-be-processed alarm information into an alarm classification model to obtain an alarm level of the to-be-processed alarm information output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to represent the urgency of the to-be-processed alarm information.

[0007] In one possible implementation, the fourth feature variable of the to-be-processed alarm information is obtained by: performing priority sorting on the alarm type to which the to-be-processed alarm information belongs, and performing inverse index subtraction uniformization standardization processing on the sorted result; obtaining an accumulated importance factor corresponding to the alarm type to which the to-be-processed alarm information belongs, the accumulated importance factor representing the importance of the alarm type; and obtaining the fourth feature variable according to the result of the standardization processing and the accumulated importance factor.

[0008] In a possible implementation manner, the method further includes: obtaining training data, the training data including feature data of historical alarm information and alarm levels corresponding to the historical alarm information; creating a clustering-KNN classifier according to the training data; classifying the historical alarm information according to alarm levels to obtain a plurality of classification sets, and determining initial clustering centers of each classification set; and obtaining extended clustering centers of each classification set by using a K-MEANS clustering algorithm; calculating similarity of the to-be-processed alarm information and the historical alarm information, and determining a predetermined number of historical alarm information with the largest similarity as neighbor alarm information; calculating a degree factor of the neighbor alarm information belonging to an extended clustering center of a classification set in which the neighbor alarm information is located; and determining an alarm level of the to-be-processed alarm information according to the degree factor.

[0009] In a possible implementation manner, after the feature data of the to-be-processed alarm information is input into the alarm classification model to obtain the alarm level of the to-be-processed alarm information output by the alarm classification model, the method further includes: generating a fault work order according to the alarm level of the to-be-processed alarm information and a network resource comprehensive account, wherein the network resource comprehensive account includes state data of a device corresponding to the to-be-processed alarm information; determining a small division attribution of the fault work order, and dispatching the fault work order to a maintenance user according to a result of the small division attribution.

[0010] In a possible implementation manner, the dispatching the fault work order to the maintenance user according to the result of the small division attribution includes: according to an order from high to low of the alarm level of the to-be-processed alarm information, sequentially dispatching corresponding fault work orders according to the result of the small division attribution.

[0011] In a possible implementation manner, after the dispatching the fault work order to the maintenance user according to the result of the small division attribution, the method further includes: in response to detecting that the fault work order has not been checked off after a predetermined time period elapses since the dispatching of the fault work order is completed, determining a superior user of the maintenance user, and dispatching the fault work order to the superior user.

[0012] In one possible implementation, before acquiring the feature data of the alarm information to be processed, the method further includes: detecting whether there is a first alarm information in the alarm information to be processed that meets a predetermined condition, wherein the predetermined condition includes multiple base stations experiencing instantaneous outages exceeding a predetermined number; acquiring the feature data of the alarm information to be processed includes: acquiring the feature data of the alarm information to be processed for the alarm information to be processed other than the first alarm information; after detecting whether there is a first alarm information in the alarm information to be processed that meets the predetermined condition, the method further includes: determining the alarm level of the alarm information to be processed for the first alarm information, wherein the alarm level is higher than the alarm level of the alarm information to be processed output by the alarm classification model.

[0013] On the other hand, this application provides an alarm classification device, comprising: a feature extraction module, used to acquire feature data of alarm information to be processed, the feature data including a first feature variable representing the duration of the alarm, a second feature variable representing the customer type, a third feature variable representing the alarm site type, and a fourth feature variable; wherein the fourth feature variable is used to represent the importance of the alarm type to which the alarm information to be processed belongs; and a classification module, used to input the feature data of the alarm information to be processed into an alarm classification model to obtain the alarm level of the alarm information to be processed output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to represent the urgency of the alarm information to be processed.

[0014] In one possible implementation, the feature extraction module is specifically used to: prioritize the alarm types to which the alarm information to be processed belongs, and perform inverse index subtraction standardization processing on the sorting result; obtain the cumulative importance factor corresponding to the alarm type to which the alarm information to be processed belongs, wherein the cumulative importance factor characterizes the importance of the alarm type; and obtain the fourth feature variable based on the result of the standardization processing and the cumulative importance factor.

[0015] In one possible implementation, the apparatus further includes: a model training module, configured to: acquire training data, the training data including feature data of historical alarm information and alarm levels corresponding to the historical alarm information; and: create a KNN classifier based on the training data; classify the historical alarm information according to alarm levels to obtain multiple classification sets, and determine the initial cluster center of each classification set; and: use the K-MEANS clustering algorithm to obtain the extended cluster center of each classification set, obtaining the cluster center of each classification set, the cluster center including the initial cluster center and the extended cluster center; calculate the similarity between the alarm information to be processed and the historical alarm information, determine the top predetermined number of historical alarm information with the highest similarity as the nearest neighbor alarm information; calculate the degree factor of the nearest neighbor alarm information belonging to the cluster center of its classification set, and adjust the KNN classifier according to the degree factor to obtain the alarm classification model.

[0016] In one possible implementation, the device further includes: a work order scheduling module, configured to: generate a fault work order based on the alarm level of the pending alarm information and the network resource comprehensive ledger; wherein the network resource comprehensive ledger includes the status data of the device corresponding to the pending alarm information; determine the sub-category of the fault work order, and dispatch the fault work order to the maintenance user according to the sub-category result.

[0017] In one possible implementation, the work order scheduling module is specifically used to: dispatch corresponding fault work orders according to the order of alarm levels of the pending alarm information from high to low, and according to the result of the subdivision and affiliation.

[0018] In one possible implementation, the work order scheduling module is further configured to: in response to detecting that the fault work order has not been cleared after a predetermined period of time since its dispatch, determine the superior user of the maintenance user, and dispatch the fault work order to the superior user.

[0019] In one possible implementation, the device further includes: a pre-judgment module, configured to detect whether there is a first alarm message among the alarm messages to be processed that meets predetermined conditions, the predetermined conditions including multiple base stations experiencing instantaneous outages exceeding a predetermined number; a feature extraction module, specifically configured to acquire feature data of the alarm messages to be processed, including: acquiring feature data of the alarm messages to be processed for alarm messages other than the first alarm message; the pre-judgment module is further configured to determine the work order level of the fault work order for the first alarm message, wherein the work order level is higher than the work order level of the fault work order output by the alarm classification model.

[0020] In another aspect, this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method described above.

[0021] In another aspect, this application provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement the method described above.

[0022] The alarm classification method, apparatus, electronic device, and medium provided in this application, after obtaining the alarm information to be processed, first extracts feature data and inputs the feature data into the alarm classification model to obtain the alarm level of the alarm information to be processed. The urgency of the alarm information to be processed is comprehensively evaluated from four dimensions: alarm duration, customer type, alarm site type, and alarm type. A clustering-KNN classifier is used as the alarm classification model to obtain the alarm information classification result, thereby efficiently and accurately determining the urgency of the alarm information to be processed and improving fault handling capabilities. Attached Figure Description

[0023] 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.

[0024] Figure 1 A flowchart illustrating the alarm classification method provided in Embodiment 1 of this application;

[0025] Figure 2 This is a flowchart illustrating the operation of the clustering-KNN classifier provided in Embodiment 1 of this application;

[0026] Figure 3 Example diagram of the maintenance work order upgrade provided in Embodiment 1 of this application;

[0027] Figure 4 Example diagram of upgrading a smaller maintenance work order as provided in Embodiment 1 of this application;

[0028] Figure 5 Example diagram of upgrading a property management work order based on Embodiment 1 of this application;

[0029] Figure 6 This is an example diagram of the tower maintenance work order upgrade provided in Embodiment 1 of this application;

[0030] Figure 7 Example diagram of handling difficult fault work orders provided in Embodiment 1 of this application;

[0031] Figure 8This is a schematic diagram of the alarm classification device provided in Embodiment 2 of this application;

[0032] Figure 9 This is a schematic diagram of the alarm classification electronic device provided in Embodiment 3 of this application.

[0033] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0034] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0035] In this application, a module refers to a functional module or a logical module. It can be in software form, where its function is implemented by a processor executing program code; or it can be in hardware form. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "OR" relationship.

[0036] With the comprehensive deployment of the national "new infrastructure" strategy, the scale of 5G networks is rapidly expanding, and the number of network alarm messages is increasing accordingly, adding new pressure to network maintenance and scheduling. Therefore, how to efficiently process and respond to alarm messages has become particularly important. In practical applications, the number of received alarm messages is large, so they are usually classified according to their urgency, and fault solutions are then scheduled and executed based on the urgency level. Thus, the ability to achieve accurate and reliable alarm classification has become a crucial factor in improving the efficiency of alarm feedback and processing.

[0037] The technical solutions of this application are illustrated below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0038] Example 1

[0039] Figure 1 This is a flowchart illustrating an alarm classification method provided in an embodiment of this application. Figure 1 As shown, the alarm classification method provided in this embodiment may include:

[0040] S101, Obtain feature data of the alarm information to be processed. The feature data includes a first feature variable representing the duration of the alarm, a second feature variable representing the customer type, a third feature variable representing the alarm site type, and a fourth feature variable. The fourth feature variable is used to represent the importance of the alarm type to which the alarm information to be processed belongs.

[0041] S102, input the feature data of the alarm information to be processed into the alarm classification model to obtain the alarm level of the alarm information to be processed output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to characterize the urgency of the alarm information to be processed.

[0042] In practical applications, the execution subject of this embodiment can be an alarm classification device, which can be implemented by a computer program, such as application software; or it can be implemented as a medium storing relevant computer programs, such as a USB flash drive or cloud drive; or it can be implemented by a physical device that integrates or installs relevant computer programs, such as a chip or server.

[0043] Specifically, after acquiring the alarm information to be processed, the alarm classification device extracts the feature data of the alarm information, which includes a first feature variable, a second feature variable, a third feature variable, and a fourth feature variable. The first feature variable indicates whether the alarm duration has reached a predetermined time; the second feature variable indicates the customer type; the third feature variable indicates whether it is a critical site alarm; and the fourth feature variable indicates the importance of the alarm type mentioned in the alarm information. The feature data of the alarm information is extracted from four dimensions: alarm duration, customer type, alarm site, and alarm type. This allows for a comprehensive judgment of the urgency of the alarm information. The feature data of the alarm information to be processed is input into a pre-trained clustering-KNN classifier to obtain the alarm level of the alarm information, which is used to characterize the urgency of the alarm information.

[0044] Table 1. Examples of partial characteristic data for alarm information.

[0045] Number C1 C2 C3 1 0 0 1 2 1 0 0 3 0 1 0 4 0 0 0 5 0 0 0 6 0 0 0 7 0 0 0 8 0 1 0 9 1 1 0 10 0 0 0

[0046] For example, Table 1 shows an example of obtaining the first, second, and third feature variables provided in Embodiment 1 of this application. As shown in Table 1, the first feature variable is denoted as C1, representing whether the alarm persists for 5 minutes; the second feature variable is denoted as C2, representing whether it is a large customer; the third feature variable is denoted as C3, representing whether it is a critical protection site, with 0 representing no and 1 representing yes. The fourth feature variable is denoted as C4, representing the importance of the alarm type to which the alarm information belongs. Thus, the feature data of the alarm information to be processed can be obtained. The feature data of the alarm information to be processed is input into the trained clustering-KNN classifier to obtain the alarm level of the alarm information to be processed. In this example, the alarm level output by the clustering-KNN classifier is three categories: -1, 0, and 1. -1 represents that the urgency and importance of the alarm information to be processed is low, and no further processing is performed; 0 represents that the urgency of the alarm information to be processed is relatively low, and it can be marked as a normal order when dispatching a fault work order; 1 represents that the urgency of the alarm information to be processed is relatively high, and it can be marked as a secondary order when dispatching a fault work order.

[0047] The fourth characteristic variable can be calculated in several ways. In one example, S101 may specifically include:

[0048] The alarm types to which the alarm information to be processed belong are prioritized, and the sorting results are subjected to inverse index subtraction for standardization.

[0049] Obtain the cumulative importance factor corresponding to the alarm type to which the alarm information to be processed belongs, wherein the cumulative importance factor characterizes the importance of the alarm type;

[0050] The fourth feature variable is obtained based on the results of the standardization process and the accumulated importance factor.

[0051] Specifically, the alarm information describes different alarm types, which are ranked according to their processing priority in actual applications. Since smaller values ​​indicate higher priority, to ensure consistency in the direction of all indicators (i.e., larger values ​​indicate higher importance), the ranking results undergo inverse indicator subtraction standardization processing. This involves subtracting the priority value from the number of alarm types to obtain the standardized result. Considering that alarm types with higher priority have a greater impact, a cumulative importance factor is assigned to each alarm type based on its importance. The fourth characteristic variable of the alarm information is calculated based on the standardized result and the cumulative importance factor.

[0052] Table 2 Example of calculating the fourth characteristic variable

[0053] Alarm name Priority x i ]]> Important factor alpha i ]] Fourth characteristic variable A Network element connection interruption 1 15 0.98 Cell unavailable alarm 2 10 0.88 Radio frequency unit service unavailable alarm 3 10 0.87 Cell not established 4 10 0.85 Radio frequency unit maintenance link abnormal alarm 5 10 0.83 Single board occlusion alarm 6 10 0.82 Radio frequency unit DC power failure alarm 7 5 0.72 Single board hardware failure alarm 8 5 0.70 Cell occlusion alarm 9 5 0.68 Radio frequency unit standing wave alarm 10 5 0.67 Cell service capacity decline alarm 11 5 0.65 Transmission optical interface abnormal alarm 12 5 0.63 Radio frequency unit CPRI interface abnormal alarm 13 5 0.62 ……

[0054] For example, Table 2 provides an example of calculating a fourth feature variable according to Embodiment 1 of this application. As shown in Table 2, assuming that the current alarm types include 45 types such as network element connection interruption alarm, cell unavailability alarm, and radio frequency unit service unavailability alarm, the alarm types are set to M=45, and the different alarm types are prioritized and the result is recorded as x. i It also performs inverse index subtraction for standardization and consistency processing; and assigns an additive importance factor α to different alarm types. i (here α) max =15), and finally the fourth characteristic variable of the alarm information is calculated. The calculation formula is as follows:

[0055] A = [(Mx i )+α i ] / (M+α max )

[0056] There are various ways to implement the clustering-KNN classifier. In one example, the method may also include:

[0057] Acquire training data, which includes feature data of historical alarm information and alarm levels corresponding to the historical alarm information; and create a clustering-KNN classifier based on the training data.

[0058] The historical alarm information is classified according to the alarm level to obtain multiple classification sets, and the initial cluster center of each classification set is determined; and the extended cluster center of each classification set is obtained by using the K-MEANS clustering algorithm.

[0059] Calculate the similarity between the alarm information to be processed and the historical alarm information, and determine the top predetermined number of historical alarm information with the highest similarity as the nearest neighbor alarm information;

[0060] Calculate the degree factor of the nearest neighbor alarm information belonging to the extended cluster center of its classification set; and determine the alarm level of the alarm information to be processed based on the degree factor.

[0061] The traditional KNN algorithm, given a training dataset with predefined categories, predicts the classification of new samples based on the categories of their nearest neighbors through methods such as majority voting. However, the traditional KNN algorithm does not consider the importance of each nearest neighbor to its respective category, leading to decreased classification accuracy. Therefore, clustering algorithms can be used to improve the traditional KNN algorithm, resulting in the clustering-KNN algorithm. Before KNN classification, the clustering algorithm calculates the cluster centers for each category and the importance factor of each nearest neighbor belonging to its cluster center. This calculated importance factor is then used to modify the traditional KNN algorithm. Several clustering algorithms can be selected; this solution uses the K-MEANS clustering algorithm as an example.

[0062] Figure 2 This is a schematic diagram illustrating the operation of a clustering-KNN classifier provided in one embodiment of this application. Figure 2 As shown, since clustering algorithms perform clustering analysis without any prior data information, they sometimes fail to obtain effective clustering results. Therefore, this scheme utilizes the known alarm levels of the training samples to guide the K-MEANS clustering algorithm in finding initial cluster centers. That is, among the training samples with known alarm levels, training samples with the same alarm level are grouped into one class, and a training sample is randomly selected from each class as the initial cluster center for that class. This overcomes the blindness of the K-MEANS clustering algorithm in selecting initial cluster centers and helps to improve clustering quality.

[0063] Let the training sample dataset be X = {x} i |x i =(x i1 x i2 , ..., x iN ) T Let i = 1, 2, ..., n, where n is the number of training samples, N is the dimension of the feature vector of the training samples, and there are M target alarm levels ω1, ω2, ..., ωn. M Therefore, the number of clusters in the K-MEANS clustering algorithm is M. Referring to the previous example, the alarm levels of the alarm information are divided into three categories: -1, 0, and 1. Therefore, M is set to 3. The alarm information samples to be processed are y = (y1, y2, ..., y...). N ) T ;

[0064] For a given training sample dataset X = {x i |x i =(x i1 x i2 , ..., x iN ) TThe training samples are divided into M classes based on their alarm levels, i = 1, 2, ..., n. A training sample is randomly selected from each class as the initial cluster center for that class. in Corresponding to categories ω1, ω2, ..., ω M The initial cluster centers;

[0065] The K-MEANS clustering algorithm is used to find the M cluster centers C1, C2, ..., C3 of the training data. M C r =(C r1 C r2 C rN ), r = 1, 2, ..., M;

[0066] Find the sample to be processed, y = (y1, y2, ..., y3). N ) T The similarity SIM(y, X) with each training sample is shown in the following formula:

[0067]

[0068] Find the training samples corresponding to the K maximum values ​​in the calculated similarity SIM(y, X) and use them as the K nearest neighbors g of the sample to be processed. m =(g m1 g m2 , ..., g mN ) T m = 1, 2, ..., K;

[0069] Calculate the degree factor μ(g) of the K nearest neighbor samples belonging to its cluster center. m ω j j = 1, 2, ..., M, μ(g m ω j The calculation method is when Otherwise, calculate using the following formula:

[0070]

[0071] Calculate the discriminant functions and generate decisions. First, calculate K discriminant functions F(y, ω). j Given K, j = 1, 2, ..., K, the alarm level class(y) of the sample y to be processed is determined as the alarm level corresponding to the largest of the calculated K discriminant function values. The specific calculation method is as follows:

[0072]

[0073] class(y)=argmax{F(y,ω1),F(y,ω2),F(y,ω3),...,F(y,ω K )}

[0074] After alarm classification is implemented, fault work orders can be dispatched based on the classification results. In one example, after inputting the feature data of the alarm information to be processed into the alarm classification model and obtaining the alarm level of the alarm information to be processed output by the alarm classification model, it may further include:

[0075] A fault work order is generated based on the alarm level of the pending alarm information and the comprehensive network resource ledger; wherein, the comprehensive network resource ledger includes the status data of the device corresponding to the pending alarm information;

[0076] The fault work order is assigned to a smaller subclass, and based on the subclass assignment result, the fault work order is dispatched to the maintenance user.

[0077] Specifically, the comprehensive network resource ledger organizes data center ledgers, physical site ledgers, logical site ledgers, community-level service ledgers, and transmission service ledgers. It integrates basic site information, site power information, network topology of each site, comprehensive maintenance status, service status, and property status from various systems, consolidating the network foundation data of various systems and platforms. This unifies data source standards and formats, solving the problem of users having to work with multiple systems, leading to low efficiency. Based on the alarm level of the pending alarm information obtained from the classification, and combined with the status data of the corresponding devices in the comprehensive network resource ledger, fault work orders are generated. To improve dispatch efficiency, each region, such as the city company, is subdivided into smaller regions. After determining the sub-regional affiliation of fault work orders, they are dispatched within the sub-regions.

[0078] For example, after classifying the alarm levels of pending alarm information, a fault work order is generated by combining relevant equipment data and information such as site area, affiliation, protection level, critical scenario, maintenance personnel, and responsible professional from the network resource comprehensive ledger. After determining its sub-affiliation, the fault work order is dispatched to the maintenance personnel within that sub-affiliation. Upon receiving the fault work order, the maintenance personnel conduct a preliminary verification of the fault's attribution and then dispatch the work order to the maintenance user in the department where the problem originated.

[0079] Dividing areas into smaller zones facilitates refined management and improves operational efficiency. Simultaneously, it provides network monitoring visualization capabilities, enabling detailed visualization and precise scheduling and control of network operation quality at a finer granularity. This includes overall fault status, fault trends, cluster analysis, and site alarm query functionality, including site topology interfaces and alarm information. Targeted network quality improvement functions are also available, providing indicator query capabilities based on network quality metrics for each smaller area. This supports real-time analysis of network metrics by area, horizontal and month-on-month benchmarking, and indicator early warning functions. Targeted alerts are sent, and fault lists are linked synchronously, providing professional improvement suggestions to guide targeted optimization and improvement within each smaller area. The system also supports collaborative request functions, allowing smaller areas to initiate work orders, dispatching support requests to relevant professionals or issuing collaborative requests to other smaller areas. Time-limited processing, scheduled upgrades, and closed-loop evaluation functions are provided, offering a score for collaboration between smaller areas and other smaller areas. This enables online collaborative linkage between smaller teams across areas and professions. Smaller requests include technical support, spare parts, network resources (pipeline resources, supporting resources, platform resources, etc.), and on-site coordination (property coordination, etc.).

[0080] The alarm level of pending alarm information can indicate the dispatch of fault work orders. In one example, dispatching the fault work order to the maintenance user based on the result of the subclassification may include:

[0081] Based on the alarm levels of the pending alarm information in descending order, and according to the results of the subdivision and attribution, the corresponding fault work orders are dispatched.

[0082] Specifically, the higher the alarm level of the pending alarm information, the greater its importance and urgency. Therefore, based on the alarm level of the pending alarm information, fault work orders with higher alarm levels can be prioritized for dispatch, thus achieving orderly fault handling. For example, the aforementioned level 2 work orders are dispatched before ordinary work orders.

[0083] There are various strategies for dispatching fault work orders. One strategy is to combine the number of fault work orders and alarm levels of each maintenance user to determine their workload and dispatch fault work orders in order of workload from low to high. There are no restrictions on this strategy.

[0084] To improve the processing efficiency of fault work orders, a work order escalation strategy can be set. In one example, after dispatching the fault work order to the maintenance user based on the subclassification result, the process further includes:

[0085] In response to the detection that the fault work order has not been cleared after a predetermined period of time since it was dispatched, the superior user of the maintenance user is identified, and the fault work order is dispatched to the superior user.

[0086] Specifically, to improve the efficiency of fault work order processing for maintenance users, a work order processing time can be set. If a work order is not cleared after the predetermined time, an overdue reminder will be sent to the maintenance user. If the maintenance user cannot automatically escalate the fault work order to their superior user or a high-weight user within a smaller defined scope, the escalation method can include automatic escalation and manual escalation, which are not restricted here.

[0087] For example, Figure 3 This is an example diagram illustrating the upgrade of a maintenance work order provided in Embodiment 1 of this application. Assume that a fault work order is dispatched to a maintenance service provider based on the fault issue. If the maintenance service provider fails to resolve the fault within 24 hours, a timeout message is sent to the maintenance supervisor and the dispatch department administrator in the DingTalk group, prompting the maintenance service provider to process the work order. Work orders that the maintenance service provider cannot handle can be forwarded to the maintenance supervisor. If the maintenance supervisor fails to resolve the fault within 48 hours, a timeout message is sent to the sub-CEO and the dispatch department administrator in the DingTalk group, prompting the maintenance supervisor to process the work order. Work orders that the maintenance supervisor cannot handle can be forwarded to the sub-CEO. If the sub-CEO fails to resolve the fault within 48 hours, a timeout message is sent to the dispatch department administrator in the DingTalk group, prompting the sub-CEO to follow the sub-CEO's process for dispatching difficult fault work orders. The person handling the fault work order can modify the "Fault Type (Power, Fiber Optic Cable, Equipment, Other)," "Fault Cause," "Responsibility," and "Unrepaired" fields. Other fields in the fault work order cannot be modified.

[0088] For another example, Figure 4 This is an example diagram illustrating the upgrade of a subdivided maintenance work order according to Embodiment 1 of this application. Assume that, based on the fault issue, a fault work order is dispatched to the subdivided maintenance team. If the subdivided maintenance team fails to resolve the fault within 24 hours, a timeout message is sent to the subdivided CEO and the scheduling department administrator in the DingTalk group, prompting the subdivided maintenance team to process the work order. Work orders that the subdivided maintenance team cannot handle can be forwarded to the subdivided CEO. If the subdivided CEO fails to resolve the fault within 48 hours, a timeout message is sent to the scheduling department administrator in the DingTalk group, prompting the subdivided CEO to follow the process of dispatching complex fault work orders.

[0089] For another example, Figure 5 This is an example diagram illustrating the upgrade of work orders for smaller property management units as provided in Embodiment 1 of this application. Assuming the fault belongs to a third party, the fault work order is dispatched to the smaller property management unit. If the smaller property management unit fails to resolve the fault within 24 hours, a timeout message is sent to the smaller CEO and the dispatch department administrator in the DingTalk group, prompting the smaller property management unit to process the work order. Work orders that the smaller property management unit cannot handle can be forwarded to the smaller CEO. If the smaller CEO fails to resolve the fault within 48 hours, a timeout message is sent to the dispatch department administrator in the DingTalk group, prompting the smaller CEO to follow the process of dispatching work orders for complex faults.

[0090] For another example, Figure 6 This is an example diagram illustrating the upgrade of tower maintenance work orders provided in Embodiment 1 of this application. Assuming the fault is a tower-related issue, the work order is dispatched to the tower maintenance provider. If the tower maintenance provider fails to resolve the fault within 24 hours, a timeout message is sent to the tower area maintenance manager and the dispatch department administrator via DingTalk, prompting the tower maintenance provider to process the work order. Work orders that the tower maintenance provider cannot handle can be forwarded to the tower area maintenance manager. If the tower area maintenance manager fails to resolve the fault within 48 hours, a timeout message is sent to the sub-CEO and the dispatch department administrator via DingTalk, prompting the tower area maintenance manager to process the work order. Work orders that the tower area maintenance manager cannot handle can be forwarded to the sub-CEO. If the sub-CEO fails to resolve the fault within 48 hours, a timeout message is sent to the dispatch department administrator via DingTalk, prompting the sub-CEO to follow the sub-CEO's process for dispatching complex fault work orders.

[0091] For another example, Figure 7 This is an example diagram illustrating the handling of complex fault work orders provided in Embodiment 1 of this application. Upon receiving a fault work order, the regional fault dispatcher must verify the attribution of the fault and assign the work order to the responsible professional maintenance personnel. The responsible professional maintenance personnel must provide feedback within 24 hours; otherwise, a timeout message will be sent to the CEO, supervisor, and dispatch department administrator in the DingTalk group, prompting the responsible professional maintenance personnel to process the work order. There are three feedback options: the first is to arrange processing, requiring recovery within 36 hours; the second is to schedule processing, requiring a clear processing plan and repair time (plan setting text box, time setting time selection panel), requiring recovery within one month; the third is that it cannot be processed, forwarding it to the supervising leader for approval, who must initiate the sub-site dismantling process within 4 hours. Once confirmed, these three feedback options cannot be modified.

[0092] Because some pending alarm messages correspond to extremely high urgency levels and the judgment rules are simple, in order to quickly identify high alarm levels, in one example, before obtaining the feature data of the pending alarm messages, the following may also be included:

[0093] Detect whether there is a first alarm message in the alarm messages to be processed that meets a predetermined condition, the predetermined condition including multiple base stations experiencing instantaneous outages exceeding a predetermined number;

[0094] The acquisition of feature data for alarm information to be processed includes: acquiring feature data for alarm information to be processed other than the first alarm information.

[0095] After detecting whether there is a first alarm message that meets a predetermined condition among the alarm messages to be processed, the method further includes: determining the alarm level of the alarm message to be processed for the first alarm message, wherein the alarm level is higher than the alarm level of the alarm message to be processed output by the alarm classification model.

[0096] Specifically, if multiple base stations experience instantaneous outages in the alarm information, the urgency level can be determined to be extremely high. Instead of classifying and outputting alarm levels through a classifier, the alarm level is set to a higher level than the alarm level output by the alarm classification model to characterize its urgency. For example, if the alarm levels output by the aforementioned clustering-KNN classifier are -1, 0, and 1, and if 10 base stations experience instantaneous outages or other urgent alarm information at a certain point in time, the alarm level is determined to be 2, and a Level 1 emergency alert is generated for immediate handling.

[0097] In the alarm classification method provided in this embodiment, after obtaining the alarm information to be processed, feature data is first extracted and then input into the alarm classification model to obtain the alarm level of the alarm information to be processed. The urgency of the alarm information to be processed is comprehensively evaluated from four dimensions: alarm duration, customer type, alarm site type, and alarm type. The clustering-KNN classifier is used as the alarm classification model to obtain the classification result of the alarm information, thereby efficiently realizing the accurate judgment of the urgency of the alarm information to be processed and improving the fault handling capability.

[0098] Example 2

[0099] Figure 8 This is a schematic diagram of the structure of an alarm classification device provided in an embodiment of this application. Figure 8 As shown, the alarm classification device provided in this embodiment may include:

[0100] The feature extraction module 81 is used to obtain feature data of the alarm information to be processed. The feature data includes a first feature variable representing the duration of the alarm, a second feature variable representing the customer type, a third feature variable representing the alarm site type, and a fourth feature variable. The fourth feature variable is used to represent the importance of the alarm type to which the alarm information to be processed belongs.

[0101] The classification module 82 is used to input the feature data of the alarm information to be processed into the alarm classification model to obtain the alarm level of the alarm information to be processed output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to characterize the urgency of the alarm information to be processed.

[0102] In practical applications, the execution subject of this embodiment can be an alarm classification device, which can be implemented by a computer program, such as application software; or it can be implemented as a medium storing relevant computer programs, such as a USB flash drive or cloud drive; or it can be implemented by a physical device that integrates or installs relevant computer programs, such as a chip or server.

[0103] Specifically, after acquiring the alarm information to be processed, the feature extraction module 81 extracts the feature data of the alarm information, which includes a first feature variable, a second feature variable, a third feature variable, and a fourth feature variable. The first feature variable represents whether the alarm duration has reached the predetermined time; the second feature variable represents the customer type; the third feature variable represents whether it is a critical site alarm; and the fourth feature variable represents the importance of the alarm type mentioned in the alarm information. The feature data of the alarm information is extracted from four dimensions: alarm duration, customer type, alarm site, and alarm type. This allows for a comprehensive judgment of the urgency of the alarm information. The classification module 82 inputs the feature data of the alarm information to be processed into a pre-trained clustering-KNN classifier to obtain the alarm level of the alarm information to be processed, which is used to represent the urgency of the alarm information.

[0104] The fourth feature variable can be calculated in several ways. In one example, feature extraction module 81 can specifically be used for:

[0105] The alarm types to which the alarm information to be processed belong are prioritized, and the sorting results are subjected to inverse index subtraction for standardization.

[0106] Obtain the cumulative importance factor corresponding to the alarm type to which the alarm information to be processed belongs, wherein the cumulative importance factor characterizes the importance of the alarm type;

[0107] The fourth feature variable is obtained based on the results of the standardization process and the accumulated importance factor.

[0108] Specifically, the alarm information describes different alarm types, which are ranked according to their processing priority in actual applications. Since smaller values ​​indicate higher priority, to ensure consistency in the direction of all indicators (i.e., larger values ​​indicate higher importance), the ranking results undergo inverse indicator subtraction standardization processing. This involves subtracting the priority value from the number of alarm types to obtain the standardized result. Considering that alarm types with higher priority have a greater impact, a cumulative importance factor is assigned to each alarm type based on its importance. The fourth characteristic variable of the alarm information is calculated based on the standardized result and the cumulative importance factor.

[0109] There are various training methods for the clustering-KNN classifier. In one example, the device may further include:

[0110] Model training module: used to acquire training data, which includes feature data of historical alarm information and alarm levels corresponding to the historical alarm information; and to create a clustering-KNN classifier based on the training data;

[0111] The historical alarm information is classified according to the alarm level to obtain multiple classification sets, and the initial cluster center of each classification set is determined; and the extended cluster center of each classification set is obtained by using the K-MEANS clustering algorithm.

[0112] Calculate the similarity between the alarm information to be processed and the historical alarm information, and determine the top predetermined number of historical alarm information with the highest similarity as the nearest neighbor alarm information;

[0113] Calculate the degree factor of the nearest neighbor alarm information belonging to the extended cluster center of its classification set; and determine the alarm level of the alarm information to be processed based on the degree factor.

[0114] The traditional KNN algorithm, given a training dataset with predefined categories, predicts the classification of new samples based on the categories of their nearest neighbors through methods such as majority voting. However, the traditional KNN algorithm does not consider the importance of each nearest neighbor to its respective category, leading to decreased classification accuracy. Therefore, clustering algorithms can be used to improve the traditional KNN algorithm, resulting in the clustering-KNN algorithm. Before KNN classification, the clustering algorithm calculates the cluster centers for each category and the importance factor of each nearest neighbor belonging to its cluster center. This calculated importance factor is then used to modify the traditional KNN algorithm. Several clustering algorithms can be selected; this solution uses the K-MEANS clustering algorithm as an example.

[0115] Specifically, since clustering algorithms perform clustering analysis without any prior data information, they sometimes fail to obtain effective clustering results. Therefore, this solution utilizes the known alarm levels of the training samples to guide the K-MEANS clustering algorithm in finding initial cluster centers. That is, among the training samples with known alarm levels, training samples with the same alarm level are grouped into one class, and a training sample is randomly selected from each class as the initial cluster center for that class. This overcomes the blindness in the selection of initial cluster centers by the K-MEANS clustering algorithm and helps to improve the clustering quality.

[0116] Let the training sample dataset be X = {x} i |x i =(x i1 xi2 , ..., x iN ) T Let i = 1, 2, ..., n, where n is the number of training samples, N is the dimension of the feature vector of the training samples, and there are M target alarm levels ω1, ω2, ..., ωn. M Therefore, the number of clusters in the K-MEANS clustering algorithm is M. Referring to the previous example, the alarm levels of the alarm information are divided into three categories: -1, 0, and 1. Therefore, M is set to 3. The alarm information samples to be processed are y = (y1, y2, ..., y...). N ) T ;

[0117] For a given training sample dataset X = {x i |x i =(x i1 x i2 , ..., x iN ) T The training samples are divided into M classes based on their alarm levels, i = 1, 2, ..., n. A training sample is randomly selected from each class as the initial cluster center for that class. in Corresponding to categories ω1, ω2, ..., ω M The initial cluster centers;

[0118] The K-MEANS clustering algorithm is used to find the M cluster centers C1, C2, ..., C3 of the training data. M C r =(C r1 C r2 C rN ), r = 1, 2, ..., M;

[0119] Find the sample to be processed, y = (y1, y2, ..., y3). N ) T The similarity SIM(y, x) with each training sample is shown in the following formula:

[0120]

[0121] Find the training samples corresponding to the K maximum values ​​in the calculated similarity SIM(y, X) and use them as the K nearest neighbors g of the sample to be processed. m =(g m1 g m2 , ..., g mN ) T m = 1, 2, ..., K;

[0122] Calculate the degree factor μ(g) of the K nearest neighbor samples belonging to its cluster center.m ω j j = 1, 2, ..., M, μ(g m ω j The calculation method is when Otherwise, calculate using the following formula:

[0123]

[0124] Calculate the discriminant functions and generate decisions. First, calculate K discriminant functions F(y, ω). j Given K, j = 1, 2, ..., K, the alarm level class(y) of the sample y to be processed is determined as the alarm level corresponding to the largest of the calculated K discriminant function values. The specific calculation method is as follows:

[0125]

[0126] class(y)=argmax{F(y,ω1),F(y,ω2),F(y,ω3),...,F(y,ω K )}

[0127] After alarm classification is implemented, fault work orders can be dispatched based on the classification results. In one example, the device may further include a work order scheduling module for:

[0128] A fault work order is generated based on the alarm level of the pending alarm information and the comprehensive network resource ledger; wherein, the comprehensive network resource ledger includes the status data of the device corresponding to the pending alarm information;

[0129] The fault work order is assigned to a smaller subclass, and based on the subclass assignment result, the fault work order is dispatched to the maintenance user.

[0130] Specifically, the comprehensive network resource ledger organizes data center ledgers, physical site ledgers, logical site ledgers, community-level service ledgers, and transmission service ledgers. It integrates basic site information, site power information, network topology of each site, comprehensive maintenance status, service status, and property status from various systems, consolidating the network foundation data of various systems and platforms. This unifies data source standards and formats, solving the problem of users having to work with multiple systems, leading to low efficiency. Based on the alarm level of the pending alarm information obtained from the classification, and combined with the status data of the corresponding devices in the comprehensive network resource ledger, fault work orders are generated. To improve dispatch efficiency, each region, such as the city company, is subdivided into smaller regions. After determining the sub-regional affiliation of fault work orders, they are dispatched within the sub-regions.

[0131] Dividing areas into smaller zones facilitates refined management and improves operational efficiency. Simultaneously, it provides network monitoring visualization capabilities, enabling detailed visualization and precise scheduling and control of network operation quality at a finer granularity. This includes overall fault status, fault trends, cluster analysis, and site alarm query functionality, including site topology interfaces and alarm information. Targeted network quality improvement functions are also available, providing indicator query capabilities based on network quality metrics for each smaller area. This supports real-time analysis of network metrics by area, horizontal and month-on-month benchmarking, and indicator early warning functions. Targeted alerts are sent, and fault lists are linked synchronously, providing professional improvement suggestions to guide targeted optimization and improvement within each smaller area. The system also supports collaborative request functions, allowing smaller areas to initiate work orders, dispatching support requests to relevant professionals or issuing collaborative requests to other smaller areas. Time-limited processing, scheduled upgrades, and closed-loop evaluation functions are provided, offering a score for collaboration between smaller areas and other smaller areas. This enables online collaborative linkage between smaller teams across areas and professions. Smaller requests include technical support, spare parts, network resources (pipeline resources, supporting resources, platform resources, etc.), and on-site coordination (property coordination, etc.).

[0132] The alarm level of pending alarm information can indicate the dispatch of fault work orders. In one example, the work order scheduling module can specifically be used for:

[0133] Based on the alarm levels of the pending alarm information in descending order, and according to the results of the subdivision and attribution, the corresponding fault work orders are dispatched.

[0134] Specifically, the higher the alarm level of the pending alarm information, the higher its urgency. Therefore, based on the alarm level of the pending alarm information, fault work orders with higher alarm levels can be dispatched first to achieve orderly fault handling.

[0135] There are various strategies for dispatching fault work orders. One strategy is to combine the number of fault work orders and alarm levels of each maintenance user to determine their workload and dispatch fault work orders in order of workload from low to high. There are no restrictions on this strategy.

[0136] To improve the processing efficiency of fault work orders, a work order escalation strategy can be set. In one example, the work order scheduling module can also be used for:

[0137] In response to the detection that the fault work order has not been cleared after a predetermined period of time since it was dispatched, the superior user of the maintenance user is identified, and the fault work order is dispatched to the superior user.

[0138] Specifically, to improve the efficiency of fault work order processing for maintenance users, a work order processing time can be set. If a work order is not cleared after the predetermined time, an overdue reminder will be sent to the maintenance user. If the maintenance user cannot automatically escalate the fault work order to their superior user or a high-weight user within a smaller defined scope, the escalation method can include automatic escalation and manual escalation, which are not restricted here.

[0139] Because some of the pending alarm information corresponds to extremely high importance and the judgment rules are simple, in order to quickly identify high alarm levels, in one example, the device may further include:

[0140] The pre-judgment module is used to detect whether there is a first alarm message that meets a predetermined condition among the alarm messages to be processed. The predetermined condition includes multiple base stations experiencing instantaneous outages in quantities exceeding a predetermined number.

[0141] The feature extraction module 81 is specifically used to obtain the feature data of the alarm information to be processed, including: obtaining the feature data of the alarm information to be processed for alarm information other than the first alarm information;

[0142] The pre-judgment module is further configured to determine the alarm level of the alarm information to be processed for the first alarm information, wherein the alarm level is higher than the alarm level of the alarm information to be processed output by the alarm classification model.

[0143] Specifically, if multiple base stations experience instantaneous outages in the alarm information, the urgency level can be determined to be extremely high. Instead of classifying and outputting alarm levels through a classifier, the alarm level is set to a higher level than that output by the alarm classification model to characterize its urgency.

[0144] In the alarm classification device provided in this embodiment, after obtaining the alarm information to be processed, feature data is first extracted and then input into the alarm classification model to obtain the alarm level of the alarm information to be processed. The urgency of the alarm information to be processed is comprehensively evaluated from four dimensions: alarm duration, customer type, alarm site type, and alarm type. The clustering-KNN classifier is used as the alarm classification model to obtain the classification result of the alarm information, thereby efficiently realizing the accurate judgment of the urgency of the alarm information to be processed and improving the fault handling capability.

[0145] Example 3

[0146] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure, such as... Figure 9 As shown, the electronic device includes:

[0147] The electronic device includes a processor 291 and a memory 292; it may also include a communication interface 293 and a bus 294. The processor 291, memory 292, and communication interface 293 can communicate with each other via the bus 294. The communication interface 293 can be used for information transmission. The processor 291 can invoke logical instructions stored in the memory 292 to execute the methods of the above embodiments.

[0148] Furthermore, the logic instructions in the aforementioned memory 292 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0149] The memory 292, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 291 executes functional applications and data processing by running the software programs, instructions, and modules stored in the memory 292, thereby implementing the methods in the above-described method embodiments.

[0150] The memory 292 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 292 may include high-speed random access memory and may also include non-volatile memory.

[0151] This disclosure provides a non-transitory computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement the methods described in the foregoing embodiments.

[0152] Example 4

[0153] This disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the methods provided in any of the embodiments described above.

[0154] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0155] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. An alarm classification method, characterized in that, include: The feature data of the alarm information to be processed is obtained. The feature data includes a first feature variable representing the duration of the alarm, a second feature variable representing the customer type, a third feature variable representing the alarm site type, and a fourth feature variable. The fourth feature variable is used to represent the importance of the alarm type to which the alarm information to be processed belongs. Input the feature data of the alarm information to be processed into the alarm classification model to obtain the alarm level of the alarm information to be processed output by the alarm classification model; The first, second, third, and fourth feature variables constitute a specific feature combination for comprehensively evaluating the urgency of alarms; the calculation formula for the fourth feature variable is as follows: A=[(Mx i )+a i ] / (M+α max ), Where M is the number of alarm types, x i The result of prioritizing different alarm types, α i The cumulative importance factor for different alarm types; α max For the maximum cumulative importance factor; The alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to characterize the urgency of the alarm information to be processed. Obtaining the fourth feature variable of the alarm information to be processed includes: prioritizing the alarm types to which the alarm information to be processed belongs, and performing inverse index subtraction standardization on the sorting result; obtaining the cumulative importance factor corresponding to the alarm type to which the alarm information to be processed belongs, wherein the cumulative importance factor characterizes the importance of the alarm type; and obtaining the fourth feature variable based on the result of the standardization and the cumulative importance factor. Acquire training data, which includes feature data of historical alarm information and alarm levels corresponding to the historical alarm information; and create a clustering-KNN classifier based on the training data. The historical alarm information is classified according to the alarm level to obtain multiple classification sets, and the initial cluster center of each classification set is determined; and the extended cluster center of each classification set is obtained by using the K-MEANS clustering algorithm. Calculate the similarity between the alarm information to be processed and the historical alarm information, and determine the top predetermined number of historical alarm information with the highest similarity as the nearest neighbor alarm information; Calculate the degree factor of the nearest neighbor alarm information belonging to the extended cluster center of its classification set; and determine the alarm level of the alarm information to be processed based on the degree factor.

2. The method according to claim 1, characterized in that, After inputting the feature data of the alarm information to be processed into the alarm classification model to obtain the alarm level of the alarm information to be processed output by the alarm classification model, the method further includes: A fault work order is generated based on the alarm level of the pending alarm information and the comprehensive network resource ledger; wherein, the comprehensive network resource ledger includes the status data of the device corresponding to the pending alarm information; The fault work order is assigned to a smaller subclass, and based on the subclass assignment result, the fault work order is dispatched to the maintenance user.

3. The method according to claim 2, characterized in that, The step of dispatching the fault work order to the maintenance user based on the result of the subdivision includes: Based on the alarm levels of the pending alarm information in descending order, and according to the results of the subdivision and attribution, the corresponding fault work orders are dispatched.

4. The method according to claim 2, characterized in that, After dispatching the fault work order to the maintenance user based on the result of the subdivision, the process further includes: In response to the detection that the fault work order has not been cleared after a predetermined period of time since it was dispatched, the superior user of the maintenance user is identified, and the fault work order is dispatched to the superior user.

5. The method according to any one of claims 1-4, characterized in that, Before acquiring the feature data of the alarm information to be processed, the method further includes: Detect whether there is a first alarm message in the alarm messages to be processed that meets a predetermined condition, the predetermined condition including multiple base stations experiencing instantaneous outages exceeding a predetermined number; The acquisition of feature data for alarm information to be processed includes: acquiring feature data for alarm information to be processed other than the first alarm information. After detecting whether there is a first alarm message that meets a predetermined condition among the alarm messages to be processed, the method further includes: determining the alarm level of the alarm message to be processed for the first alarm message, wherein the alarm level is higher than the alarm level of the alarm message to be processed output by the alarm classification model.

6. An alarm classification device, characterized in that, include: Feature extraction module: used to obtain feature data of alarm information to be processed. The feature data includes a first feature variable representing the duration of the alarm, a second feature variable representing the customer type, a third feature variable representing the alarm site type, and a fourth feature variable; wherein, the fourth feature variable is used to represent the importance of the alarm type to which the alarm information to be processed belongs. The classification module is used to input the feature data of the alarm information to be processed into the alarm classification model, and obtain the alarm level of the alarm information to be processed output by the alarm classification model; wherein the alarm classification model is a pre-trained clustering-KNN classifier, and the alarm level is used to characterize the urgency of the alarm information to be processed; wherein the first feature variable, the second feature variable, the third feature variable, and the fourth feature variable constitute a specific feature combination for comprehensively evaluating the urgency of the alarm; the calculation formula of the fourth feature variable is as follows: A=[(Mx i )+a i ] / (M+α max ), Where M is the number of alarm types, x i The result of prioritizing different alarm types, α i The cumulative importance factor for different alarm types; α max For the maximum cumulative importance factor; The feature extraction module is specifically used to prioritize the alarm types to which the alarm information to be processed belongs, and to perform inverse index subtraction standardization processing on the sorting results; to obtain the cumulative importance factor corresponding to the alarm type to which the alarm information to be processed belongs, wherein the cumulative importance factor represents the importance of the alarm type; and to obtain the fourth feature variable based on the standardization process and the cumulative importance factor. The model training module is used to acquire training data, including feature data of historical alarm information and the alarm levels corresponding to the historical alarm information; and to create a clustering-KNN classifier based on the training data; to classify the historical alarm information according to the alarm level to obtain multiple classification sets, and to determine the initial cluster center of each classification set; and to obtain the extended cluster center of each classification set using the K-MEANS clustering algorithm; to calculate the similarity between the alarm information to be processed and the historical alarm information, and to determine the top predetermined number of historical alarm information with the highest similarity as the nearest neighbor alarm information; to calculate the degree factor of the degree factor of the nearest neighbor alarm information belonging to the extended cluster center of its classification set; and to determine the alarm level of the alarm information to be processed based on the degree factor.

7. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-5.

Images