Tree model construction method, device, equipment and storage medium
The alarm data is preprocessed and normalized through the tree model construction method, and the tree model is constructed using the classification model, which solves the problem of degradation of the alarm association relationship in the 5G network scenario, and realizes efficient alarm association relationship management and dynamic growth .
Patent Information
- Application Number
- CN202010592638.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-06-24
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2040-06-24
AI Technical Summary
In the 5G network scenario, the traditional method of determining the root cause of alarms based on topology link assistance has decreased effectiveness, and the update method of knowledge graphs consumes a lot of resources or requires semi-automated and semi-manual intervention, which is a large workload.
The alarm data is preprocessed through the tree model construction method, a vectorized alarm association relationship set is obtained, and its characteristics are normalized. The classification model obtained by training determines the label, and a tree model is constructed to express the primary and secondary between alarms. Association relationships, dynamic generation and growth tree models.
It realizes efficient expression of the correlation between alarms, reduces resource consumption and manual intervention needs, and improves the dynamic management ability of alarm relationships.
Smart Images

Figure CN113839802B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data analysis and processing, and in particular to a tree model construction method, device, equipment and storage medium. Background Art
[0002] In the field of communications, communication equipment often generates a large number of alarms, some of which have correlations, which can be determined by topological links, business relationships, professional relationships, alarm levels, etc. In new network scenarios such as the fifth-generation mobile communication technology (5th-Generation, 5G), the topological relationships and business relationships have also changed, and the effectiveness of the traditional method of determining the root cause of alarms based on topological links will be greatly reduced.
[0003] The current solution is to express the alarm association relationship in the form of knowledge graph iteration and update. Common knowledge graph update methods are comprehensive update or incremental update, but the comprehensive update method consumes a lot of resources, and the incremental update method requires semi-automatic and semi-manual intervention, which is a lot of work. Summary of the invention
[0004] The main purpose of the embodiments of the present invention is to propose a tree model construction method, device, equipment and storage medium, aiming to express the primary and secondary correlation relationships between alarms through a tree model, and to learn and analyze the alarm data based on the trained classification model, and dynamically generate a corresponding tree model, thereby realizing the continuous growth of the tree.
[0005] To achieve the above object, an embodiment of the present invention provides a tree model construction method, which includes the following steps:
[0006] Preprocess the alarm data to obtain a vectorized alarm association relationship set;
[0007] Normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set;
[0008] Determine the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature and the classification model;
[0009] Among them, the classification model is obtained through training;
[0010] A tree model is constructed based on the labels of each alarm data in the collection.
[0011] To achieve the above purpose, an embodiment of the present invention further proposes a tree model construction device, which includes:
[0012] A processing module, used for pre-processing the alarm data to obtain a vectorized alarm association relationship set;
[0013] The processing module is further used to normalize the first feature of each alarm association relationship in the vectorized alarm association relationship set;
[0014] A determination module, used to determine the label of each alarm data in the set according to the vectorized alarm association relationship set, the first feature after normalization processing and the classification model;
[0015] Among them, the classification model is obtained through training;
[0016] The construction module is used to construct a tree model according to the labels of each alarm data in the collection.
[0017] To achieve the above-mentioned purpose, an embodiment of the present invention further proposes a device, which includes: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the above-mentioned method are implemented.
[0018] To achieve the above objectives, the present invention provides a readable and writable storage medium for computer storage, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the aforementioned method.
[0019] The embodiment of the present application provides a tree model construction method, device, equipment and storage medium, wherein the method includes preprocessing the alarm data to obtain a vectorized alarm association relationship set; normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set; determining the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature and the trained classification model; and constructing a tree model according to the label of each alarm data in the set. In this way, the primary and secondary association relationships between the alarms can be expressed through the tree model, and the alarm data can be learned and analyzed based on the trained classification model, and the corresponding tree model can be dynamically generated, thereby realizing the continuous growth of the tree. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 It is a flowchart of a tree model construction method provided by an embodiment of the present invention.
[0021] Figure 2 Schematic diagram of one-hot encoding provided by an embodiment of the present invention.
[0022] Figure 3 It is a schematic diagram of a storage method of an alarm association relationship after one-hot encoding provided by an embodiment of the present invention.
[0023] Figure 4 It is a flow chart of a method for determining each alarm data label provided by an embodiment of the present invention.
[0024] Figure 5 It is a schematic diagram of a subtree model provided by an embodiment of the present invention.
[0025] Figure 6 It is a schematic diagram of a subtree model provided by an embodiment of the present invention.
[0026] Figure 7 It is a schematic diagram of a subtree model provided by an embodiment of the present invention.
[0027] Figure 8 It is a schematic diagram of a tree model provided by an embodiment of the present invention.
[0028] Fig. 9 It is a structural schematic diagram of a tree model building device provided by an embodiment of the present invention.
[0029] Fig.10 It is a schematic diagram of a device structure provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0030] In order to make the purpose, technical solution and advantages of the present application more clear, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other arbitrarily without conflict.
[0031] In addition, in the embodiments of the present application, words such as "optionally" or "exemplarily" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "optionally" or "exemplarily" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "optionally" or "exemplarily" is intended to present related concepts in a specific way.
[0032] In order to facilitate understanding of the solutions provided by the embodiments of the present application, some descriptions of concepts related to the present application are given as examples for reference. As shown below:
[0033] Primary and secondary relationship: Under the same topology relationship, the failure of the optical port receiving link causes the remote radio unit (RRU) link to be disconnected. Under the same topology relationship and service relationship, the RRU link disconnection causes the Long Term Evolution (LTE) cell to exit service (abbreviated as: LTE cell service exit). Assuming that the LTE cell service exit and RRU link disconnection alarms occur within 10 minutes, the LTE cell service exit can be used as the primary alarm and the RRU link disconnection can be used as the sub-alarm. The correlation relationship between the two alarms is a primary and secondary relationship.
[0034] Homologous relationship: The same fault cause leads to multiple alarm associations generated by the same network element, or the same fault cause leads to multiple alarm associations generated by multiple network elements connected in the same profession, or the same fault cause leads to multiple alarm associations involving multiple professions. For example, if a network element goes down, the "destination signaling point unreachable" alarms from other network elements to the network element can be homologous.
[0035] Parent-child relationship: Indicates the primary and secondary relationship of the alarm association relationship in the tree model. In the primary and secondary relationship, the primary alarm can be used as the parent alarm, and the child alarm remains unchanged, that is, the primary and secondary relationship is converted into the parent-child relationship in the tree model. In the homology relationship, the source alarm can be used as the parent alarm, and other alarms can be used as child alarms, so that the homology relationship can be converted into the parent-child relationship in the tree model.
[0036] Based on the above concepts, the present application embodiment provides a tree model construction method, such as Figure 1 As shown, the method comprises the following steps:
[0037] S101. Preprocess the alarm data to obtain a vectorized alarm association relationship set.
[0038] There are a large number of alarms generated due to various reasons in wireless networks. Since individual alarms may not have a reference effect on the correlation relationship of other alarms, the alarm data can be pre-processed in a data cleaning manner.
[0039] Exemplarily, the embodiment of the present application provides an implementation method for preprocessing alarm data, which can be to mine the alarm data to obtain an alternative set of alarm association relationships, thereby realizing the cleaning process of the alarm data. Furthermore, in the embodiment of the present application, the alarm data in the alternative set of alarm association relationships obtained by mining can also be one-hot encoded, and then, the alarm vector neural network is trained using the one-hot encoded alarm data as input to obtain a vectorized alarm association relationship set.
[0040] Optionally, in the above preprocessing process of the alarm data, an Apriori algorithm, an FP Growth algorithm, an FPMAX algorithm, etc. may be used for alarm mining, and the alarm vector neural network may use a network model such as Skip-gram.
[0041] S102: Normalize the first feature of each alarm association relationship in the vectorized alarm association relationship set.
[0042] Exemplarily, the first feature in this step may include at least one of an alarm topology relationship, an alarm business relationship, an alarm level, an alarm professional relationship, an alarm cosine similarity, an alarm support degree, and an alarm certainty degree.
[0043] Optionally, when the first feature includes at least one of the alarm topology relationship, alarm business relationship, alarm level, and alarm professional relationship, at least one of the alarm topology relationship, alarm business relationship, alarm level, and alarm professional relationship of each alarm association relationship can be normalized according to the actual network structure. For example, in the actual network structure, assuming that alarm A and alarm B occur in the indoor baseband processing unit (Building Base Band Unit, BBU) and the base station respectively, the topology relationship feature is taken as an example for normalization. Assume that A is the parent alarm of B, and B is not the parent alarm of A, then the dimension that A is the parent alarm of B can be marked as 0, and the dimension that B is not the parent alarm of A can be marked as 1.
[0044] Similarly, the alarm professional relationship, alarm business relationship, and alarm level may also be normalized according to the actual network structure, wherein the normalized value of the alarm level may be the actual alarm level value.
[0045] In the case where the first feature includes at least one of the alarm support, alarm confidence, and alarm cosine similarity, at least one of the alarm support, alarm confidence, and alarm cosine similarity of each alarm association relationship can be normalized according to the normalization threshold. For example, taking the alarm cosine similarity as an example, assuming that the cosine similarity between alarm A and alarm B is 0.9, and the cosine similarity normalization threshold is 0.7, that is, the cosine similarity between alarm A and alarm B is greater than the corresponding cosine similarity normalization threshold, then the dimension can be set to 1. On the contrary, if the cosine similarity between alarm A and alarm B is less than the corresponding cosine similarity normalization threshold, then the dimension can be set to 0.
[0046] The cosine similarity between the alarm data A and the alarm data B can be obtained through the alarm vector neural network in step S101, and the alarm support and alarm confidence can be obtained in the process of alarm data mining.
[0047] It should be noted that the above-mentioned alarm support, alarm confidence, and alarm cosine similarity can correspond to corresponding normalized thresholds respectively. For example, taking the alarm confidence as an example, this feature can be normalized according to its corresponding normalized threshold. Of course, the normalized thresholds corresponding to different first features can be the same, and the embodiments of the present application are not limited to this.
[0048] It can be understood that when the first feature includes at least one of the alarm topology relationship, alarm business relationship, alarm level, alarm professional relationship, and at least one of the alarm support, alarm certainty, and alarm cosine similarity, the corresponding first feature can be normalized according to the above two methods respectively.
[0049] S103: Determine a label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature, and the classification model.
[0050] The above classification model can be obtained through machine learning training. For example, training and optimization are performed through training data and test data. The training data and test data can be selected according to the alarm time period that occurs continuously. Of course, when training the classification model, frequent set mining can be performed on the training data and test data in advance through a mining algorithm to narrow the data range.
[0051] Optionally, in the process of training the classification model based on training data and test data, the learning effect of the classification model can be evaluated. For example, after evaluating the network model based on test data, if the classification accuracy of the model is less than 80%, the network model can be optimized and re-learned by adjusting parameters.
[0052] Exemplarily, there are two ways to adjust parameters: 1. Redesign the training data, for example, if there is too little training data, it is easy to overfit; 2. Redefine regularization, for example, if the training data has better training results, but the accuracy of the test data test results is not high.
[0053] After the classification model is trained based on the training data and the test data, classification and learning can be performed based on the classification model in combination with the vectorized alarm association relationship set obtained through steps S101 and S102 and the first feature after normalization to determine the label of each alarm data in the vectorized alarm association relationship set.
[0054] The label of the alarm data may indicate that the alarm data is a parent node or a child node.
[0055] Note that the above classification model can adopt common machine learning classification models such as logistic regression or SVM.
[0056] S104: Construct a tree model according to the labels of each alarm data in the set.
[0057] Through the above steps, the label of each alarm data in the vectorized alarm association relationship set is obtained, that is, after determining that the alarm data is a parent node or a child node, a tree model can be constructed according to the label of the alarm data.
[0058] Optionally, in the case of an existing knowledge graph (hereinafter referred to as the original tree model), a subtree model can be constructed according to the labels of the alarm data in this step, and the constructed subtree model can be integrated into the original tree model to construct a new tree model. In this way, not only can the association relationship between alarms be conveniently and quickly found based on the tree model, but also the continuous growth of the tree model can be dynamically realized.
[0059] The embodiment of the present application provides a tree model construction method, specifically preprocessing the alarm data to obtain a vectorized alarm association relationship set; normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set; determining the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature and the trained classification model; and constructing a tree model according to the label of each alarm data in the set. In this way, the primary and secondary association relationships between the alarms can be expressed through the tree model, and the alarm data can be learned and analyzed based on the trained classification model, and the corresponding tree model can be dynamically generated, thereby realizing the continuous growth of the tree.
[0060] like Figure 2 As shown, in one embodiment, during the process of preprocessing the alarm data in the above step S101, when the alarm data is one-hot encoded, only the corresponding alarm dimension can be set to 1, and the remaining positions can be set to 0. That is, when one-hot encoding is used in all the n types of alarms that may be associated, only a certain alarm itself is set to 1, and the remaining associated alarms are set to 0. Accordingly, when saving data, for example, saving the alarm association relationship after one-hot encoding to the database, it can be designed according to the dimension of the position where the alarm is 1. Figure 3 As shown, Figure 2 The position where LTE cell decommissioning is 1 is 6, so the unique hot code corresponding to LTE cell decommissioning is 6. Figure 2 The position where the RRU link disconnection is 1 is 3, so the unique hot code corresponding to the RRU link disconnection is 3, and the position where the optical port receiving link failure is 1 is 2, so the unique hot code corresponding to the optical port receiving link failure is 2.
[0061] like Figure 4 As shown, in one embodiment, the implementation method of determining the label of each alarm data in the vectorized alarm association relationship set in the above step S103 may include but is not limited to the following steps:
[0062] S401 : Determine the probability and cosine similarity corresponding to each alarm association relationship according to the vectorized alarm association relationship set, the normalized first feature, and the classification model.
[0063] The probability in this step can be used to represent the probability that any alarm data is the parent node of another alarm data in the corresponding alarm association relationship, and the cosine similarity can be used to represent the correlation between any alarm data and another alarm data in the corresponding alarm association relationship. For example, vectorized training is performed through the alarm vector neural network, the cosine similarity between the two alarm data in the vectorized alarm association relationship set is calculated, and the probability that any alarm data is the parent node of another alarm data is calculated through the Sigmoid function in the logistic regression classification model.
[0064] S402: When the cosine similarity is greater than or equal to the similarity normalization threshold and the probability of the corresponding alarm association relationship is greater than a preset value, mark the label of any alarm data in the corresponding alarm association relationship as a parent node and mark the label of another alarm data as a child node.
[0065] Exemplarily, assuming that the similarity normalization threshold is 0.5, the calculated cosine similarity between alarm data A and alarm data B is 0.9, which is greater than the similarity normalization threshold, and it is determined that there is a strong correlation between alarm data A and alarm data B. Correspondingly, if the probability that alarm data A is the parent node of alarm data B is 0.83, which is greater than the preset value of 0.45, the label of alarm data A is marked as the parent node, and the label of alarm data B is marked as the child node.
[0066] S403: When the cosine similarity is less than the similarity normalization threshold, the labels of any alarm data and another alarm data in the corresponding alarm association relationship are marked as child nodes.
[0067] Similarly, assuming that the similarity normalization threshold is 0.5, the cosine similarity between alarm data A and alarm data B calculated based on the classification model is 0.35, which is less than the similarity normalization threshold, and it is determined that there is no strong correlation between alarm data A and alarm data B. Therefore, the labels of alarm data A and alarm data B can be marked as child nodes.
[0068] It can be understood that there is only one parent node at the top of the tree model. Since there is no strong correlation between alarm data A and alarm data B, these two data cannot be the parent node at the top of the tree model at the same time. Even if alarm data A and alarm data B may be the parent nodes of other nodes respectively, then alarm data A and alarm data B must be the child nodes of other nodes. Therefore, the labels of alarm data A and alarm data B that do not have a strong correlation can both be marked as child nodes.
[0069] In one embodiment, constructing a tree model according to labels of each alarm data in the set in the above step S104 may include the following different situations.
[0070] In the case where there is no conflict between the labels of the alarm data, a subtree model is constructed based on the labels of the alarm data. For example, assuming that in alarm data A and alarm data B, alarm data A is the parent node and alarm data B is the child node, and in alarm data B and alarm data C, alarm data B is the parent node and alarm data C is the child node, that is, alarm data A is the parent node of alarm data B, and alarm data B is the parent node of alarm data C. Then, based on the labels of alarm data A, B, and C, a subtree model can be constructed as follows: Figure 5 The subtree model shown.
[0071] In the case where there is a conflict between the labels of each alarm data, the labels of each alarm data are updated according to the probability corresponding to the alarm association relationship to which each alarm data belongs, and a subtree model is constructed according to the updated labels of each alarm data. For example, assuming that in alarm data A and alarm data B, alarm data A is the parent node and alarm data B is the child node, and in alarm data B and alarm data C, alarm data B is the parent node and alarm data C is the child node, and in alarm data A and alarm data C, alarm data C is the parent node and alarm data A is the child node, that is, the labels of alarm data A, B, and C conflict and cannot form a subtree model, then the labels of each alarm data can be updated based on the probability corresponding to the alarm association relationship to which each alarm data belongs. Assume that in the alarm association relationships A and B, the probability that alarm data A is the parent node of alarm data B is 53%, in the alarm association relationships B and C, the probability that alarm data B is the parent node of alarm data C is 85%, and in the alarm association relationships A and C, the probability that alarm data C is the parent node of alarm data A is 75%. Then, we can select the alarm association relationship with a probability close to 50%, that is, the alarm association relationships A and B, and update the probability that alarm data B is the parent node of alarm data A to 47%. Then, in alarm data A and alarm data B, we can update the label of alarm data A to a child node, and update the label of alarm data B to a parent node, that is, the updated labels of each alarm data are, alarm data B is the parent node of C, and alarm data C is the parent node of alarm data A. In this way, based on the updated labels of each alarm data, we can construct the following Figure 6 The subtree model shown.
[0072] Of course, the above examples all show that there is a strong correlation between the alarm data. If there is no strong correlation between the alarm data, that is, the cosine similarity is less than the similarity normalization threshold, then the label of each alarm data can be updated as follows. For example, assuming that in the alarm association relationships A and C, there is no strong correlation between alarm data A and alarm data C, the alarm association relationships A and B, and the alarm association relationships C and B are all strongly correlated, and the probability that alarm data A is the parent node of alarm data B is 53%, and the probability that alarm data B is the parent node of alarm data C is 85%, then according to the above method, the labels of alarm data A and C that do not have a strong correlation can be marked as child nodes. In alarm data A and alarm data B, alarm data A is the parent node and alarm data B is the child node. In alarm data B and alarm data C, alarm data B is the parent node and alarm data C is the child node. Because the label of alarm data A is updated to a child node, and the probability that it is the parent node of alarm data B is 53%, then the probability that alarm data B is the parent node of alarm data A is 47%. The label of alarm data B can be updated to a parent node, that is, alarm data B is used as a parent node, and alarm data A and C are used as child nodes. In this way, the following can be constructed: Figure 7 The subtree model shown.
[0073] After constructing subtree models for the above-mentioned different situations, a tree model can be constructed according to the constructed subtree models, that is, the growth of the tree model is realized.
[0074] Similarly, the following different situations may exist in the process of building a tree model. For example, if the parent node of the subtree model exists in the original tree model, the node of the original tree model that is the same as the parent node of the subtree model can be used as the target node, and the subtree model can be attached to the target node of the original tree model. Figure 8 As shown, assuming that the constructed subtree model is as shown in the dotted box part, its parent node is LTE cell decommissioning, and there is a node with the same parent node as "LTE cell decommissioning" in the original tree model, then the "LTE cell decommissioning" in the original tree model is used as the target node, and the subtree model in the dotted box part is attached to the "LTE cell decommissioning" node of the original tree model, thereby realizing the growth of the tree model.
[0075] In another case, when the parent node of the subtree model does not exist in the original tree model, resulting in the subtree model being unable to be attached to the original tree model, the labels of each alarm data in the subtree model can be updated according to the probability corresponding to the alarm association relationship to which each alarm data belongs in the subtree model. For example, the update can be performed according to the above-listed method for updating the labels of each alarm data, and the subtree model can be reconstructed according to the updated labels of each alarm data until the parent node of the reconstructed subtree model exists in the original tree model. In this way, the node in the original tree model that is the same as the parent node of the reconstructed subtree model can be used as the target node, and the reconstructed subtree model can be attached to the target node of the original tree model to achieve the growth of the tree model.
[0076] Fig. 9 A tree model construction device provided in an embodiment of the present application, such as Fig. 9 As shown, the device includes: a processing module 901, a determination module 902, and a construction module 903;
[0077] The processing module is used to pre-process the alarm data to obtain a vectorized alarm association relationship set;
[0078] The processing module is further used to normalize the first feature of each alarm association relationship in the vectorized alarm association relationship set;
[0079] A determination module, used to determine the label of each alarm data in the set according to the vectorized alarm association relationship set, the first feature after normalization processing and the classification model;
[0080] Among them, the classification model is obtained through training;
[0081] The first feature may include at least one of the following: alarm topology relationship, alarm business relationship, alarm level, alarm professional relationship, alarm cosine similarity, alarm support, and alarm certainty.
[0082] The construction module is used to construct a tree model according to the labels of each alarm data in the collection.
[0083] In one embodiment, the above-mentioned processing module is used to mine the alarm data to obtain a candidate set of alarm association relationships; to perform one-hot encoding on the alarm data in the candidate set of alarm association relationships; to perform vector alarm neural network training based on the one-hot encoded alarm data to obtain a vectorized alarm association relationship set.
[0084] In one embodiment, the normalization processing process of the above-mentioned processing module can be to normalize at least one of the alarm topology relationship, alarm business relationship, alarm level, and alarm professional relationship of each alarm association relationship according to the actual network structure when the first feature includes at least one of the alarm topology relationship, alarm business relationship, alarm level, and alarm professional relationship;
[0085] and / or, when the first feature includes at least one of alarm support, alarm confidence, and alarm cosine similarity, normalizing at least one of the alarm support, alarm confidence, and alarm cosine similarity of each alarm association relationship according to a normalization threshold;
[0086] Among them, the alarm support, alarm confidence, and alarm cosine similarity each correspond to a corresponding normalized threshold.
[0087] In one embodiment, the determination module may be used to determine the probability and cosine similarity corresponding to each alarm association relationship based on the vectorized alarm association relationship set, the normalized first feature, and the classification model;
[0088] Among them, the probability is used to represent the probability that any alarm data is the parent node of another alarm data in the corresponding alarm association relationship, and the cosine similarity is used to represent the correlation between any alarm data and another alarm data in the corresponding alarm association relationship;
[0089] When the cosine similarity is greater than or equal to the similarity normalization threshold, and the probability of the corresponding alarm association relationship is greater than a preset value, the determination module can be used to mark the label of any alarm data in the corresponding alarm association relationship as a parent node, and mark the label of another alarm data as a child node; when the cosine similarity is less than the similarity normalization threshold, the determination module can be used to mark the labels of any alarm data and another alarm data in the corresponding alarm association relationship as child nodes.
[0090] In one embodiment, the above-mentioned construction module can be used to construct a subtree model according to the labels of each alarm data when there is no conflict between the labels of each alarm data;
[0091] Alternatively, when there is a conflict between the labels of the alarm data, the labels of the alarm data are updated according to the probability corresponding to the alarm association relationship to which the alarm data belongs, and a subtree model is constructed according to the updated labels of the alarm data;
[0092] Construct a tree model based on the subtree model.
[0093] Exemplarily, the implementation of the above-mentioned tree model construction may include but is not limited to the following methods:
[0094] When the parent node of the subtree model exists in the original tree model, the node in the original tree model that is the same as the parent node of the subtree model is used as the target node, and the subtree model is attached to the target node of the original tree model.
[0095] Alternatively, when the parent node of the subtree model does not exist in the original tree model, the label of each alarm data in the subtree model is updated according to the probability corresponding to the alarm association relationship to which each alarm data in the subtree model belongs;
[0096] Reconstruct the subtree model according to the updated labels of each alarm data until the parent node of the reconstructed subtree model exists in the original tree model;
[0097] The node in the original tree model that is the same as the parent node of the reconstructed subtree model is taken as the target node, and the reconstructed subtree model is attached to the target node of the original tree model.
[0098] The tree model building device provided in this embodiment is used to implement Figure 1 The tree model construction method of the illustrated embodiment has similar implementation principles and technical effects, which will not be described in detail here.
[0099] Fig.10 A schematic diagram of a device provided in an embodiment is shown in FIG. Fig.10 As shown, the device includes a processor 1001 and a memory 1002; the number of processors 1001 in the device can be one or more. Fig.10 A processor 1001 is taken as an example; the processor 1001 and the memory 1002 in the device may be connected via a bus or other means. Fig.10 The example of connecting through bus is taken in the following.
[0100] The memory 1002 is a computer-readable storage medium that can be used to store software programs, computer executable programs, and modules. Figure 1The program instructions / modules corresponding to the tree model construction method in the embodiment (for example, the processing module 901, the determination module 902, and the construction module 903 in the tree model construction device). The processor 1001 implements the above-mentioned tree model construction method by running the software programs, instructions, and modules stored in the memory 1002.
[0101] The memory 1002 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the set-top box, etc. In addition, the memory 1002 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
[0102] The embodiment of the present application further provides a readable and writable storage medium for computer storage, the storage medium storing one or more programs, the one or more programs being executable by one or more processors to execute a tree model construction method, the method comprising:
[0103] Preprocess the alarm data to obtain a vectorized alarm association relationship set;
[0104] Normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set;
[0105] Determine the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature and the classification model;
[0106] Among them, the classification model is obtained through training;
[0107] A tree model is constructed based on the labels of each alarm data in the collection.
[0108] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above and the functional modules / units in the devices may be implemented as software, firmware, hardware, or a suitable combination thereof.
[0109] In hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0110] The above only illustrates the exemplary embodiments of the present application with reference to the accompanying drawings, and does not limit the scope of the present invention. Any modification, equivalent substitution and improvement made by those skilled in the art without departing from the scope and essence of the present invention shall be within the scope of the present invention.
Claims
1. A tree model construction method, characterized in that: The method comprises the following steps: Preprocess the alarm data to obtain a vectorized alarm association relationship set; Normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set; Determine a label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature, and the classification model; Wherein, the classification model is obtained through training; Building a tree model according to the labels of each alarm data in the set; The step of determining the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature, and the classification model includes: Determine the probability and cosine similarity corresponding to each alarm association relationship according to the vectorized alarm association relationship set, the normalized first feature and the classification model; The probability is used to represent the probability that any alarm data is the parent node of another alarm data in the corresponding alarm association relationship, and the cosine similarity is used to represent the magnitude of the correlation between any alarm data and another alarm data in the corresponding alarm association relationship; When the cosine similarity is greater than or equal to the similarity normalization threshold and the probability of the corresponding alarm association relationship is greater than a preset value, mark the label of any alarm data in the corresponding alarm association relationship as a parent node, and mark the label of another alarm data as a child node; When the cosine similarity is less than a similarity normalization threshold, labels of any alarm data and another alarm data in the corresponding alarm association relationship are marked as child nodes.
2. The method according to claim 1, characterized in that The preprocessing of the alarm data to obtain a vectorized alarm association relationship set includes: Mining the alarm data to obtain an alternative set of alarm association relationships; Performing one-hot encoding on the alarm data in the alarm association candidate set; The alarm vector neural network is trained based on the one-hot encoded alarm data to obtain a vectorized alarm association relationship set.
3. The method according to claim 1, characterized in that The first feature includes at least one of the following: Alarm topology relationship, alarm business relationship, alarm level, alarm professional relationship, alarm cosine similarity, alarm support, and alarm certainty.
4. The method according to claim 3, characterized in that Normalizing the first feature of each alarm association relationship in the vectorized alarm association relationship set includes: In the case where the first feature includes at least one of an alarm topology relationship, an alarm business relationship, an alarm level, and an alarm professional relationship, normalizing at least one of the alarm topology relationship, the alarm business relationship, the alarm level, and the alarm professional relationship of each alarm association relationship according to the actual network structure; And / or, when the first feature includes at least one of alarm support, alarm confidence, and alarm cosine similarity, at least one of the alarm support, alarm confidence, and alarm cosine similarity of each alarm association relationship is normalized according to a normalization threshold; Among them, the alarm support, alarm confidence, and alarm cosine similarity each correspond to a corresponding normalized threshold.
5. The method according to claim 1, characterized in that The step of constructing a tree model according to the labels of each alarm data in the set includes: In the case where there is no conflict between the labels of the alarm data, constructing a subtree model according to the labels of the alarm data; Alternatively, in the case where there is a conflict between the labels of the alarm data, the labels of the alarm data are updated according to the probabilities corresponding to the alarm association relationships to which the alarm data belong, and a subtree model is constructed according to the updated labels of the alarm data; The tree model is constructed according to the subtree model.
6. The method according to claim 5, characterized in that The step of constructing the tree model according to the subtree model comprises: In the case where the parent node of the subtree model exists in the original tree model, the node in the original tree model that is the same as the parent node of the subtree model is used as the target node, and the subtree model is attached to the target node of the original tree model; Alternatively, when the parent node of the subtree model does not exist in the original tree model, the label of each alarm data in the subtree model is updated according to the probability corresponding to the alarm association relationship to which each alarm data in the subtree model belongs; Reconstructing the subtree model according to the updated labels of each alarm data until the parent node of the reconstructed subtree model exists in the original tree model; The node in the original tree model that is the same as the parent node of the reconstructed subtree model is taken as the target node, and the reconstructed subtree model is attached to the target node of the original tree model.
7. A tree model construction device, characterized in that: include: A processing module, used for pre-processing the alarm data to obtain a vectorized alarm association relationship set; The processing module is further used to normalize the first feature of each alarm association relationship in the vectorized alarm association relationship set; A determination module, used to determine the label of each alarm data in the set according to the vectorized alarm association relationship set, the normalized first feature and the classification model; Wherein, the classification model is obtained through training; A construction module, used to construct a tree model according to the label of each alarm data in the set; The determination module is specifically used to: determine the probability and cosine similarity corresponding to each alarm association relationship according to the vectorized alarm association relationship set, the normalized first feature and the classification model; The probability is used to represent the probability that any alarm data is the parent node of another alarm data in the corresponding alarm association relationship, and the cosine similarity is used to represent the magnitude of the correlation between any alarm data and another alarm data in the corresponding alarm association relationship; When the cosine similarity is greater than or equal to the similarity normalization threshold and the probability of the corresponding alarm association relationship is greater than a preset value, mark the label of any alarm data in the corresponding alarm association relationship as a parent node, and mark the label of another alarm data as a child node; When the cosine similarity is less than a similarity normalization threshold, labels of any alarm data and another alarm data in the corresponding alarm association relationship are marked as child nodes.
8. A device, characterized in that: include: A memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, wherein the program, when executed by the processor, realizes the tree model construction method as described in any one of claims 1 to 6.
9. A readable and writable storage medium for computer storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the tree model construction method according to any one of claims 1 to 6.
Citation Information
Patent Citations
Rapid diagnosis method for satellite channel link faults
CN108880706A
Alarm processing method and device, equipment and medium
CN109951306A
Fault root cause analysis method and device
CN110609759A
Correlation analysis and prediction method for power grid alarm signal
CN110929951A
Network alarm analysis model creation method, alarm analysis method and device
CN111125268A
