Method for determining a fault image, processing device and computer-readable storage medium

Abstract

The application discloses a fault image determination method, a processing device and a computer readable storage medium. The method comprises the following steps: constructing a heterogeneous network based on historical fault data, wherein the heterogeneous network takes faults as nodes and takes the correlation between faults as edges; determining the correlation between the faults based on the heterogeneous network; determining a fault group according to the correlation; when a fault to be processed is received, determining the distance between the fault to be processed and the fault group; and determining the fault image of the fault to be processed according to the distance. The effect of improving the fault scheduling efficiency is achieved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method for determining fault profiles, a processing device, and a computer-readable storage medium. Background Technology

[0002] Fault dispatching refers to the rapid response and mobilization of support teams to restore business operations affected by a fault, triggered by feedback from frontline departments or the detection of anomalies. In related technologies, fault dispatching typically requires staff to rely on experience to determine the appropriate dispatching method for a given fault. Therefore, fault dispatching demands a high level of experience from decision-makers, who need relevant fault-handling experience to react quickly. This leads to a relatively low efficiency in fault dispatching. Summary of the Invention

[0003] This application provides a method for determining fault profiles, a processing device, and a computer-readable storage medium, which solves the problem of low fault scheduling efficiency in related technologies and achieves the effect of improving fault scheduling efficiency.

[0004] This application provides a method for determining a fault profile, the method comprising the following steps:

[0005] A heterogeneous network is constructed based on historical fault data, wherein faults are the nodes and the relationships between faults are the edges.

[0006] The correlation between the faults is determined based on the heterogeneous network;

[0007] The fault groups are determined based on the aforementioned relationships;

[0008] Upon receiving a fault to be processed, determine the distance between the fault to be processed and the fault group;

[0009] The fault profile of the fault to be processed is determined based on the distance.

[0010] Optionally, the step of determining the correlation between the faults based on the heterogeneous network includes:

[0011] Based on the heterogeneous network, node sequences are obtained through a random walk network representation learning method.

[0012] The association is determined based on the node sequence.

[0013] Optionally, the step of obtaining the node sequence based on the heterogeneous network using a random walk network representation learning method includes:

[0014] Each node in the heterogeneous network is taken as the starting point for the walk;

[0015] Based on the preset edge relationship combination and the preset walk length, determine the node sequence corresponding to each walk starting point.

[0016] Optionally, the edge type of the heterogeneous network includes at least one of service impact, fault cause, and fault resolution time.

[0017] Optionally, the step of determining the node sequence corresponding to each starting point of the walk based on a preset combination of edge relationships and a preset walk length includes:

[0018] Based on the preset edge relationship combination, determine the candidate nodes associated with the starting point of the walk;

[0019] Based on the relationship weight between the starting point and the candidate nodes, the next hop node of the starting point is selected from the candidate nodes;

[0020] Repeatedly execute the process of determining candidate nodes for the next hop node based on the preset edge relationship combination, and select the next hop node from the candidate nodes of the next hop node according to the relationship weight;

[0021] When the walking length is greater than or equal to the preset walking length, the node sequence is determined according to the walking path.

[0022] Optionally, before the step of determining the fault profile of the fault to be processed based on the distance, the method further includes:

[0023] Determine the fault group profile;

[0024] The step of determining the fault profile of the fault to be processed based on the distance includes:

[0025] Determine the target fault group corresponding to the fault to be processed, wherein the target fault group is the fault group with the smallest distance to the fault to be processed;

[0026] Based on the fault group profile of the target fault group, the fault profile of the fault to be processed is determined.

[0027] Optionally, before the step of constructing a heterogeneous network based on historical fault data, the method further includes:

[0028] When the fault database is updated, fault data within a preset time window is identified in the fault database.

[0029] The fault data is used as the historical data.

[0030] This application embodiment also provides a processing device, the processing device comprising:

[0031] The construction module is used to build a heterogeneous network based on historical fault data. The heterogeneous network uses faults as nodes and the relationships between faults as edges.

[0032] A determination module is configured to determine the correlation between the faults based on the heterogeneous network; and to determine a fault group based on the correlation.

[0033] A calculation module is used to determine the distance between the fault to be processed and the fault group when a fault to be processed is received;

[0034] The analysis module is used to determine the fault profile of the fault to be processed based on the distance.

[0035] This application embodiment also provides a processing device, including a memory, a processor, and a fault profile determination program stored in the memory and executable on the processor. When the processor executes the fault profile determination program, it implements the method described above.

[0036] This application also provides a computer-readable storage medium storing a fault profile determination program thereon, which, when executed by a processor, implements the method described above.

[0037] One or more technical solutions provided in this application, by profiling historical faults and incorporating effective information such as the risk status and handling solutions of historical faults, match the current fault with existing faults and predict the trend of the current fault through the window of historical faults. This allows for the prediction of fault risks through a combination of subjective experience and objective data, timely dispatch of relevant personnel when the risks are controllable, and promotion of fault handling according to risk priority. It also provides insights into locating the causes of faults and simulates fault solutions, thereby improving the efficiency of fault dispatching. Attached Figure Description

[0038] Figure 1 A schematic flowchart illustrating an embodiment of the method for determining fault profiles in this application;

[0039] Figure 2 A heterogeneous network diagram involved in one embodiment of the method for determining fault profiles in this application;

[0040] Figure 3 Another schematic diagram of the process involved in an embodiment of the method for determining fault profiles in this application;

[0041] Figure 4 Another schematic diagram of a process involved in an embodiment of the method for determining fault profiles in this application;

[0042] Figure 5 This is a schematic diagram of the modules involved in the method for determining the fault profile in this application;

[0043] Figure 6 This is a schematic diagram of the terminal hardware structure involved in the embodiments of this application;

[0044] Figure 7 This is a modular schematic diagram of the processing device involved in the embodiments of this application. Detailed Implementation

[0045] In related technologies, fault dispatching typically requires staff to rely on experience to determine how to handle a given fault. Therefore, fault dispatching demands a high level of experience from decision-makers, who need relevant fault-handling experience to react quickly. This results in relatively low fault dispatching efficiency.

[0046] In addition, there are other related technologies that can monitor the operational status and working environment of various communication stations (including communication equipment rooms, base stations, branch offices, and modular offices), or monitor the operation of business systems. This allows for the acquisition of indicators across various dimensions of equipment, environment, and system, and the calculation of the current global environment and local system health using different methods. This health level is then compared with historical health levels under normal conditions, and warnings are issued by setting thresholds. Dispatchers can receive warnings immediately and then manually match these warnings with existing faults, incorporating the warning information into the fault dispatching process.

[0047] This solution improves the accuracy of health assessment and the timeliness of early warning, but it mainly plays a role in the field of early warning. If early warning is to provide value in the field of fault dispatching, it cannot provide solutions for actual faults such as dispatching.

[0048] In summary, the relevant technical solutions all have shortcomings, reflected in labor costs, reference value, and depth of analysis. Currently, users have increasingly stringent requirements for systems, and system failures can lead to user dissatisfaction. How to quickly and reasonably schedule failures to ensure rapid business recovery is a problem that needs optimization. Optimizing this problem can mitigate losses and win back users. Traditional failure scheduling processes are influenced by subjective factors such as experience, placing high demands on scheduling personnel, and may result in untimely scheduling and inaccurate risk prediction. This application proposes a method for determining failure profiles. By mining failure-related features collected in failure scheduling scenarios, it constructs failure profiles for similar groups. This achieves objective data-driven assistance in failure scheduling, matching current failures with existing failures, predicting the trend of current failures through historical failure windows, and promptly scheduling relevant personnel when risks are controllable, thereby improving the timeliness and effectiveness of failure scheduling.

[0049] To better understand the above technical solutions, exemplary embodiments of this application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0050] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0051] Example 1

[0052] Please refer to Figure 1 In this embodiment, the method for determining the fault profile includes the following steps:

[0053] Step S10: Construct a heterogeneous network based on historical fault data, wherein the heterogeneous network uses faults as nodes and the relationships between faults as edges;

[0054] Step S20: Determine the correlation between the faults based on the heterogeneous network;

[0055] Step S30: Determine the faulty group based on the aforementioned correlation;

[0056] Step S40: Upon receiving a fault to be processed, determine the distance between the fault to be processed and the fault group;

[0057] Step S50: Determine the fault profile of the fault to be processed based on the distance.

[0058] In this embodiment, multi-dimensional attributes related to faults in the fault scheduling scenario can be determined first based on historical fault data. These attributes include, but are not limited to, fault discovery time, affected services, affected groups, overall fault duration, fault escalation criteria, cause of fault occurrence, fault handler, and responsible party. A feature set for calculating fault correlations is extracted, which may include one or more of the multi-dimensional attributes. Since faults are correlated on certain features, such as faults A and B affecting the same services, or faults B and C having the same cause, fault groups are then determined based on these correlations.

[0059] In some implementations, a heterogeneous information network of faults can be constructed, and relationships between faults can be mined on this network using a random walk approach. Then, clustering algorithms are used to calculate the similarity between faults, resulting in closely related fault groups. A set of group labels is then constructed based on these similar fault groups to determine a fault group profile, including but not limited to text display, radar chart display, and bar chart display. This allows, upon receiving a fault to be processed, the fault profile corresponding to the fault to be processed can be determined based on the distance between the fault to be processed and the fault group.

[0060] Optionally, in one implementation, a heterogeneous network can be constructed first based on historical data. The nodes in this heterogeneous network are of the fault type, while the edge types can be multiple, with different edge types representing different semantic information between faults. For example, ... Figure 2 The heterogeneous network shown can include faults at nodes A, B, C, D, and E. The relationships (edges) between faults A and B include their service impact and causes. The relationships between faults A and E include their service impact and processing time. The relationships between faults A and C also include their service impact. Similarly, the diagram shows relationships between other faults, which will not be elaborated upon further.

[0061] After constructing the heterogeneous network, a node sequence can be obtained based on the heterogeneous network using a random walk network representation learning method, and the association relationship can be determined based on the node sequence.

[0062] For example, in heterogeneous networks, network representation learning methods based on meta-path random walks can be used to mine the relationships between faults and thus calculate their similarity. Specifically, the method of random walks in heterogeneous networks for network representation learning is guided by meta-paths. Random walks are performed according to the order of the meta-paths to collect and store heterogeneous nodes in the network. Then, representation learning methods are used to project the nodes into a vector space of a certain dimension, allowing the nodes to be represented by vectors.

[0063] As an optional implementation, it can be based on the Metapath2vec method. The Metapath2vec method first selects the metapath of the network, and then performs a random walk on the heterogeneous network guided by the metapath. The walk process is essentially sampling the nodes in the network. After the walk is complete, a node sequence guided by the metapath is obtained. It should be noted that this embodiment can also be based on other optional methods to determine the node sequence. The Metapath2vec method is not the only implementation. The main idea of ​​the above approach is to first determine the relationship combinations between faulty nodes in the heterogeneous network, linking the faulty nodes through different relationships, so that the node relationships in the network can be captured through random walks based on the metapath. Therefore, the heterogeneous network diagram can be as follows: Figure 2 As shown, faults are linked together by attributes such as business impact, fault cause, and fault handling time. In other words, the node type in this heterogeneous network is fault, while the edge type can be multiple, with different edge types representing different semantic information between faults. Once the node sequence is determined, it can be used as the training corpus for a neural network model, allowing the model to be trained to uncover the relationships between nodes.

[0064] Optionally, during the random walk, the relationship weights between faulty nodes will be considered. The node with the greater relationship weight with the current faulty node is more likely to become the next node to be walked. For example, if the similarity of the impact on business between fault A and fault B is greater than that between fault A and fault C, then fault B is more likely to become the next node to be walked.

[0065] For example, each faulty node in the heterogeneous network can be selected sequentially and used as the starting point for the traversal. The traversal length is preset to N (i.e., the default traversal length is N). The traversal proceeds according to a certain combination of edge relationships, such as a business impact-cause relationship. The current node will search for the next node based on the business impact relationship, and the next node will search for the node after that based on the cause relationship. This process is repeated until a meta-path of length N, guided by the faulty node, is generated.

[0066] It should be noted that in this example, the relation corresponding to the t-th step (t < N) of the walk is R. t Then node v i to v i+1 State transition probability p(v i+1 |v i ) can be defined as:

[0067]

[0068] Wherein, N(v) i ) represents node vi The set of neighboring nodes, w(v i →v i+1 |R t ) represents node v i With v i+1 In relation R t The connection weights in w(v) i →v i+1 |R t The following is represented:

[0069] w(v i →v i+1 |R t ) = sim(v i →v i+1 |R t )*w(R t )

[0070] Where sim(v) i →v i+1 |R t ) represents node v i With v i+1 In relation R t The degree of similarity, for each relation R t The similarity calculations are different. For example, when relation R... t If it is text, it can be calculated using the classic Levenshtein distance and Jaccard similarity, or the word2vec method can be used. When relation R t When it is a discrete feature, it can be calculated using the Tanimoto method.

[0071] w(R t ) refers to relation R t The weights associated with a particular technology, such as the weight of R related to the occurrence of a fault within a specific time period due to the use of certain new technologies, are as follows: tThe weighting of the similarity between faults will be greater. For example, during the development of network technology, strategic goals are proposed in stages. For instance, for information systems, many new technologies are proposed to expand online, intelligent, and cloud-based capabilities. These new technologies have brought significant benefits to the production system, but during periods of unstable technology use, they inevitably have some impact on the system. In other words, in the early stages of applying certain technologies to information systems, the causes of some faults are directly related to the application of the technology. Therefore, when the causes of two faults are related to the currently used technology, the credibility of the correlation between the two faults is higher. Thus, it is necessary to adjust the connection weights between nodes in the relationship of fault causes. Optionally, as an implementation method, it is necessary to first record the technological updates for each time period. The recording method can be a table or a mapping relationship. Taking the table recording as an example, the record table contains the weight of the technology's impact on the fault, as well as the impact surface and weight value updated according to the iteration of the technology. The weight value can be determined based on actual project experience and is a preset fixed value. For example, the record table can be as shown in Table 1:

[0072]

[0073]

[0074] If the timing of a historical failure coincides with an unstable phase of a technology's launch, the initial weight can be multiplied by the technology's weight within a specific time period when traversing the relationships in dimensions related to that technology.

[0075] Furthermore, for calculating the similarity of fault handling time, it's necessary to hierarchically categorize the time consumption of different handling methods. For example, based on the different handling methods, the processing time can be divided into fault handling time caused by platform-related issues and fault handling time caused by code-related issues. Obviously, the average processing time for platform-related issues should be lower than that for code-related issues, because platform-related issues can mostly be resolved through self-healing or standard solutions. Code-related issues, on the other hand, require code modification and re-release, a process that is time-consuming.

[0076] Optionally, after determining the node sequence based on the meta-paths using the above method, to make the collected network node relationships more stable, the number of meta-paths guided by each faulty node can be set to count, where count ≥ 1, thereby reducing jitter during the acquisition process. For example, in... Figure 2 On the heterogeneous network shown, a random walk starting from fault A can yield a possible set of metapaths, Paths, as follows, where the size of Paths is count.

[0077]

[0078] Once the meta-path set `Paths` is determined, the paths taken for random walks starting from each node can be saved, with each path node serving as a fault identifier. This data can then be used as a training corpus. A neural network model can then be trained to mine the relationships between nodes.

[0079] For example, in some implementations, a word2vec model can be used to learn from this data and uncover the relationships that exist within it. Finally, a relationship representation vector corresponding to each fault is obtained. This results in a matrix A, where each row represents a representation vector for a fault. Therefore, calculating the similarity between faults can be transformed into calculating the distance between corresponding elements of matrix A.

[0080] When calculating the distance between corresponding elements of matrix A, the pairwise distances function provided by the sklearn package can be used to calculate the distance between matrices. Then, a clustering algorithm is used to calculate the similarity between faults, forming multiple fault groups, such that the similarity (distance) of samples within the same group is greater than the similarity (distance) of samples between different groups. The clustering algorithm includes, but is not limited to, the K-Means algorithm.

[0081] Furthermore, when a fault occurs, i.e. when a fault to be processed is received, in order to ensure the timeliness of the profile output, in the technical solution disclosed in this embodiment, the distance between the current fault (i.e. the fault to be processed) and the historical fault group can be directly calculated based on the tag dimension.

[0082] For example, refer to Figure 3 First, based on the analysis of historical fault profiles, a label set L = {label1, label2, ...} corresponding to the current fault can be extracted; then, the distance between the current fault and the existing fault group can be calculated. The calculation method includes, but is not limited to, using the Tanimoto method.

[0083] For example, suppose the label set of the existing historical fault group 1 is G = {label} a label b If the distance d1 between the two groups is {d1, d2, ..., dn}, then the distance d1 between them can be calculated using Tanimoto(L, G) = len(L∩G) / len(L∪G). Therefore, the distance D between the current fault and the n historical fault groups is obtained as follows: D = {d1, d2, ..., dn}. n}

[0084] Then you can select the fault group m (d = d) that is closest to the current fault. m ) is selected as a candidate. Then, it is determined whether the distance exceeds the threshold d. threshold Optionally, the threshold is the maximum distance between existing groups.

[0085] When d m <d threshold If the current fault profile is not determined by the profile of the most recent faulty group m, then the current fault profile is represented by the profile of the most recent faulty group m; otherwise, the current fault is treated as a separate group. Finally, the output of the above fault profiles is used as an objective basis for fault scheduling. This achieves the goal of reducing computational load while ensuring the accuracy of the results.

[0086] Finally, the fault profile corresponding to the current fault, i.e. the fault to be processed, can be used as a scheduling reference to quickly execute fault scheduling actions.

[0087] Alternatively, please refer to Figure 4 Before step S10, the procedure further includes:

[0088] Step S60: When updating the fault database, determine the fault data in the fault database that is within a preset time window, and use the fault data as the historical data.

[0089] Because historical faults provide a time-dependent representation of current faults, older faults are less relevant to the current fault profile than more recent faults. Therefore, time needs to be factored into the fault profile calculation. For example, a sliding window algorithm can be used to update the profile of a group of faults. The time window can be set to dimensions such as 1 hour or 1 day, as required. When the fault database is updated, the time window is moved, old candidate itemsets are deleted, and new candidate itemsets are added before the group fault profile is calculated again.

[0090] In the technical solution disclosed in this embodiment, by profiling historical faults and incorporating effective information such as the risk status and handling solutions of historical faults, the current fault is matched with existing faults, and the trend of the current fault is predicted through the window of historical faults. This allows for the prediction of fault risks through a combination of subjective experience and objective data. When the risks are controllable, relevant personnel can be promptly dispatched to promote the handling of faults according to risk priority, providing ideas for locating the causes of faults and rehearsing solutions for faults, thereby achieving the effect of improving fault dispatching efficiency.

[0091] Alternatively, please refer to Figure 5The given module diagram, based on the above embodiment, shows that in this embodiment, the method for determining the fault profile is divided into a data preparation stage, a feature mining stage, and a result output stage. The data preparation stage includes an acquisition module and a feature selection module. The main function of the feature selection module is to extract a feature set for calculating fault correlations. The feature mining stage includes a representation learning module and a clustering module. The main function of the representation learning module is to mine the correlations between faults and represent each fault with a relation representation vector. The clustering module calculates similar fault groups using a clustering algorithm, placing similar faults in the same category. The result output module includes a label output module and a portrait output module. The main function of the label output module is to construct a label set for group fault profiles based on similar groups. The portrait output module displays historical single faults or current fault profiles based on the label set of group faults, including but not limited to text display, radar chart display, and bar chart display.

[0092] Alternatively, please refer to Figure 6 , Figure 6 This is a schematic diagram of the terminal structure of the hardware operating environment involved in the embodiments of this application.

[0093] like Figure 6 As shown, the terminal can be a server or a PC, or other processing device. The processing device includes: a processor 1001, such as a CPU; a network interface 1003; a memory 1004; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The network interface 1003 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1004 can be a high-speed RAM or a stable memory. Alternatively, the memory 1004 may be a storage device independent of the aforementioned processor 1001.

[0094] Those skilled in the art will understand that Figure 6 The terminal structure shown does not constitute a limitation on the terminal and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0095] like Figure 6 As shown, the memory 1004, which is a computer storage medium, may include an operating system, a network communication module, and a fault profile determination program.

[0096] exist Figure 6 In the terminal shown, the processor 1001 can be used to call the fault profile determination program stored in the memory 1004 and perform the following operations:

[0097] A heterogeneous network is constructed based on historical fault data, wherein faults are the nodes and the relationships between faults are the edges.

[0098] The correlation between the faults is determined based on the heterogeneous network;

[0099] The fault groups are determined based on the aforementioned relationships;

[0100] Upon receiving a fault to be processed, determine the distance between the fault to be processed and the fault group;

[0101] The fault profile of the fault to be processed is determined based on the distance.

[0102] Optionally, the processor 1001 may call the fault profile determination program stored in the memory 1004, and further perform the following operations:

[0103] Based on the heterogeneous network, node sequences are obtained through a random walk network representation learning method.

[0104] The association is determined based on the node sequence.

[0105] Optionally, the processor 1001 may call the fault profile determination program stored in the memory 1004, and further perform the following operations:

[0106] Each node in the heterogeneous network is taken as the starting point for the walk;

[0107] Based on the preset edge relationship combination and the preset walk length, determine the node sequence corresponding to each walk starting point.

[0108] Optionally, the processor 1001 may call the fault profile determination program stored in the memory 1004, and further perform the following operations:

[0109] Based on the preset edge relationship combination, determine the candidate nodes associated with the starting point of the walk;

[0110] Based on the relationship weight between the starting point and the candidate nodes, the next hop node of the starting point is selected from the candidate nodes;

[0111] Repeatedly execute the process of determining candidate nodes for the next hop node based on the preset edge relationship combination, and select the next hop node from the candidate nodes of the next hop node according to the relationship weight;

[0112] When the walking length is greater than or equal to the preset walking length, the node sequence is determined according to the walking path.

[0113] Optionally, the processor 1001 may call the fault profile determination program stored in the memory 1004, and further perform the following operations:

[0114] Determine the fault group profile;

[0115] The step of determining the fault profile of the fault to be processed based on the distance includes:

[0116] Determine the target fault group corresponding to the fault to be processed, wherein the target fault group is the fault group with the smallest distance to the fault to be processed;

[0117] Based on the fault group profile of the target fault group, the fault profile of the fault to be processed is determined.

[0118] Optionally, the processor 1001 may call the fault profile determination program stored in the memory 1004, and further perform the following operations:

[0119] When the fault database is updated, fault data within a preset time window is identified in the fault database.

[0120] The fault data is used as the historical data.

[0121] Alternatively, please refer to Figure 7 This application also proposes a processing device 100, which includes:

[0122] Construction module 101 is used to construct a heterogeneous network based on historical fault data, wherein the heterogeneous network uses faults as nodes and the correlation between faults as edges;

[0123] The determination module 102 is used to determine the correlation between the faults based on the heterogeneous network; and to determine a fault group based on the correlation.

[0124] The calculation module 103 is used to determine the distance between the fault to be processed and the fault group when a fault to be processed is received;

[0125] Analysis module 104 is used to determine the fault profile of the fault to be processed based on the distance.

[0126] The processing device described is the device used to implement the method of Embodiment 1 of this application. Therefore, based on the method described in Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the device, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0127] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0128] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. This application can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0132] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0133] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of the invention. Therefore, if these modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A method for determining a fault profile, characterized in that, The method includes the following steps: A heterogeneous network is constructed based on historical fault data. The heterogeneous network uses faults as nodes and the relationships between faults as edges. The edge types of the heterogeneous network include at least one of business impact, fault cause, and fault resolution time, which are used to represent different semantic information between faults. Based on the heterogeneous network, node sequences are obtained through a random walk network representation learning method. The correlation between the faults is determined based on the node sequence; The fault groups are determined based on the aforementioned relationships; Upon receiving a fault to be processed, determine the distance between the fault to be processed and the fault group; The fault profile of the fault to be processed is determined based on the distance. The step of obtaining the node sequence based on the heterogeneous network using the random walk network representation learning method includes: Each node in the heterogeneous network is taken as the starting point for the walk; Based on the preset edge relationship combination and the preset walk length, determine the node sequence corresponding to each walk starting point; The step of determining the node sequence corresponding to each starting point of the walk based on the preset edge relationship combination and the preset walk length includes: Based on the preset edge relationship combination, determine the candidate nodes associated with the starting point of the walk; Based on the relationship weight between the starting point and the candidate nodes, the next hop node of the starting point is selected from the candidate nodes; Repeatedly execute the process of determining candidate nodes for the next hop node based on the preset edge relationship combination, and select the next hop node from the candidate nodes of the next hop node according to the relationship weight; When the walking length is greater than or equal to the preset walking length, the node sequence is determined according to the walking path.

2. The method as described in claim 1, characterized in that, Before the step of determining the fault profile of the fault to be processed based on the distance, the method further includes: Determine the fault group profile; The step of determining the fault profile of the fault to be processed based on the distance includes: Determine the target fault group corresponding to the fault to be processed, wherein the target fault group is the fault group with the smallest distance to the fault to be processed; Based on the fault group profile of the target fault group, the fault profile of the fault to be processed is determined.

3. The method as described in claim 1, characterized in that, Before the step of constructing a heterogeneous network based on historical fault data, the following steps are also included: When the fault database is updated, fault data within a preset time window is identified in the fault database. The fault data is used as the historical data.

4. A processing device, characterized in that, The processing equipment includes: The construction module is used to build a heterogeneous network based on historical fault data. The heterogeneous network uses faults as nodes and the relationships between faults as edges. The edge types of the heterogeneous network include at least one of business impact, fault cause, and fault resolution time, which are used to represent different semantic information between faults. A determination module is configured to: obtain node sequences based on the heterogeneous network using a random walk network representation learning method; determine the association relationships between the faults based on the node sequences; and determine a fault group based on the association relationships. The step of obtaining node sequences based on the heterogeneous network using the random walk network representation learning method includes: taking each node in the heterogeneous network as a walk start point; determining the node sequence corresponding to each walk start point based on a preset edge relationship combination and a preset walk length. The step of determining the node sequence corresponding to each walk start point based on the preset edge relationship combination and the preset walk length includes: determining candidate nodes associated with the walk start point based on the preset edge relationship combination; selecting the next-hop node of the walk start point from the candidate nodes based on the relationship weight between the walk start point and the candidate nodes; repeatedly determining candidate nodes for the next-hop node based on the preset edge relationship combination, and selecting the next-hop node from the candidate nodes of the next-hop node based on the relationship weight; when the walk length is greater than or equal to the preset walk length, determining the node sequence based on the walk path. A calculation module is used to determine the distance between the fault to be processed and the fault group when a fault to be processed is received; The analysis module is used to determine the fault profile of the fault to be processed based on the distance.

5. A processing device, characterized in that, The method includes a memory, a processor, and a fault profile determination program stored in the memory and executable on the processor, wherein the processor, when executing the fault profile determination program, implements the method of any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, It stores a fault profile determination program, which, when executed by a processor, implements the method described in any one of claims 1-3.

Images