CN116633758A - Network fault prediction method and system based on full-heterogeneous element comparison learning model - Google Patents
Network fault prediction method and system based on full-heterogeneous element comparison learning model Download PDFInfo
- Publication number
- CN116633758A CN116633758A CN202310276842.9A CN202310276842A CN116633758A CN 116633758 A CN116633758 A CN 116633758A CN 202310276842 A CN202310276842 A CN 202310276842A CN 116633758 A CN116633758 A CN 116633758A
- Authority
- CN
- China
- Prior art keywords
- fault
- network
- heterogeneous
- node
- occurrence time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 230000006870 function Effects 0.000 claims abstract description 14
- 239000013598 vector Substances 0.000 claims abstract description 14
- 238000010276 construction Methods 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 5
- 238000005295 random walk Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 241001334134 Rugopharynx epsilon Species 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0677—Localisation of faults
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The application discloses a network fault prediction method and a system based on a full heterogeneous element comparison learning model, wherein the method comprises the following steps: s100: converting network fault data into time series data, wherein the network fault data at least comprises a fault type, fault occurrence time and fault occurrence position; s200: constructing a network fault prediction model, wherein the network fault prediction model comprises an embedding module, a heterogeneous path network, a self-supervision comparison learning module and a prediction module; the embedding module is used for receiving the time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network is used for fusing fault node information; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting faults by adopting the loss function. The application can improve the accuracy of network fault prediction.
Description
Technical Field
The application belongs to the technical field of network fault prediction, and particularly relates to a network fault prediction method and system based on a full-heterogeneous element comparison learning model.
Background
With the increase of the types of electronic devices and the development of various software functions, network faults gradually appear in new situations while facilitating life of people. How to efficiently solve network faults is also a problem of wide concern for large operators. These network failures include service interruption, low network rates, and network noise. However, the complexity and randomness of network failures makes it difficult to make efficient predictions of network failures using conventional methods. The traditional statistical method is used for predicting network faults to be too mechanized, and after the machine learning network faults are used, accurate prediction of the network faults is expected to be realized, and as the characteristics are increased along with fault data information, the more accurate the prediction result of the machine learning method is.
Disclosure of Invention
The application aims to provide a network fault prediction method and a system based on a full heterogeneous element comparison learning model, and the method and the system can improve the accuracy of network fault prediction.
In order to achieve the above objective, an aspect of the present application provides a network failure prediction method based on a full heterogeneous element comparison learning model, including:
s100: converting network fault data into time series data, wherein the network fault data at least comprises a fault type, fault occurrence time and fault occurrence position;
s200: constructing a network fault prediction model, wherein the network fault prediction model comprises an embedding module, a heterogeneous path network, a self-supervision comparison learning module and a prediction module; the embedding module is used for receiving the time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network comprises a first heterogeneous path network and a second heterogeneous path network which are respectively constructed based on different element paths; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting the fault probability by adopting a loss function;
the first heterogeneous path network is constructed by the following steps: obtaining fault type and fault occurrence time data from the time sequence data, taking the fault type and the fault occurrence time as node objects, and constructing a first heterogeneous path network by taking the relation between the faults and the fault occurrence time and the relation between the faults and the relation between the fault occurrence time and the fault occurrence time as edges; adding random walk so as to respectively calculate the association degree between nodes in each instance path and output;
the construction of the second heterogeneous path network is as follows: obtaining fault type, fault occurrence time and fault occurrence position data from the time sequence data, taking the fault type, the fault occurrence time and the fault occurrence position as node objects, and constructing a second heterogeneous path network by taking the relation between the fault type and the fault occurrence time, the relation between the fault occurrence time and the fault occurrence time as edges; and respectively calculating projection characteristics of each node object in the second heterogeneous meta-path network, designing two meta-paths using different node objects, and fusing the embedding and output of the two meta-paths by using an attention mechanism.
Further, the degree of association between nodes in each instance pathWherein Pro (a) f ,a f+1 ) A random walk-in node a represented in example path p f To a f+1 Probability of a) f+1 To at node a f Surrounding randomly selected nodes, i.e. a f+1 For node a f Is a random neighbor node of the network.
Further, embedding of second heterogeneous path network outputWherein:
βp n is an important weight for measuring the meta-path,
ωp n for balancing p n Is of importance in terms of (a) the importance of (c),
t is the transpose, tanh represents the activation function,for the characteristic of convolutional neural network coding, W E R d×d And r.epsilon.R d×1 Are all learnable parameters, < >>Representing a semantic attention vector, V represents a node set.
Further, the self-supervision and contrast learning module is configured to construct a view fault network of the fault node to perform self-supervision learning, and further includes:
the constructed view fault network comprises a positive pair and a negative pair, wherein the view fault network of the same fault node is used as the positive pair, and the two view fault networks of the same fault node are respectively the output of a first heterogeneous path network and a second heterogeneous path network; and taking view fault networks of any two different fault nodes as negative pairs, wherein the view fault networks of the two different fault nodes are respectively the output of the first heterogeneous path network and the second heterogeneous path network.
In another aspect, the embodiment of the application provides a network fault prediction system based on a full heterogeneous element comparison learning model, which comprises:
the first module is used for converting network fault data into time series data, wherein the network fault data at least comprises a fault type, a fault occurrence time and a fault occurrence position;
the second module further comprises an embedding module, a heterogeneous path network, a self-supervision and contrast learning module and a prediction module; the embedding module is used for receiving the time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network comprises a first heterogeneous path network and a second heterogeneous path network which are respectively constructed based on different element paths; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting the fault probability by adopting a loss function;
the first heterogeneous path network is constructed by the following steps: obtaining fault type and fault occurrence time data from the time sequence data, taking the fault type and the fault occurrence time as node objects, and constructing a first heterogeneous path network by taking the relation between the faults and the fault occurrence time and the relation between the faults and the relation between the fault occurrence time and the fault occurrence time as edges; adding random walk so as to respectively calculate the association degree between nodes in each instance path and output;
the construction of the second heterogeneous path network is as follows: obtaining fault type, fault occurrence time and fault occurrence position data from the time sequence data, taking the fault type, the fault occurrence time and the fault occurrence position as node objects, and constructing a second heterogeneous path network by taking the relation between the fault type and the fault occurrence time, the relation between the fault occurrence time and the fault occurrence time as edges; and respectively calculating projection characteristics of each node object in the second heterogeneous meta-path network, designing two meta-paths using different node objects, and fusing the embedding and output of the two meta-paths by using an attention mechanism.
Compared with the prior art, the application has the following advantages and beneficial effects:
the meta path is constructed by two different meta learning methods, so that the fault node information is expressed more abundantly and the characteristics are extracted, thereby better understanding the influence relation of different faults; all fault information interaction in the two views is maximized through a self-supervision comparison learning framework, global learning is conducted, and prediction accuracy of the model is further improved. The application can improve the accuracy of network fault prediction.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of the present application;
FIG. 2 is a block diagram of a heterogeneous comparative learning model in an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made in detail and with reference to the accompanying drawings, wherein it is apparent that the embodiments described are only some, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In general, the prediction operation of faults only accords with part of the characteristics, namely, the network has a certain potential characteristic representation of certain faults, aiming at the problems, the application provides a heterogeneous element comparison learning model and a prediction module, and captures fault advanced node information to generate different prediction modes through self-supervision comparison learning so that the model has more generalization capability, thereby realizing the accurate prediction of network faults.
For the purpose of making the technical scheme and the advantageous effects of the present application more clear, the following embodiments of the present application will be further described with reference to the embodiments. It should be understood that the embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The embodiment of the application discloses a network fault prediction method based on a full heterogeneous element comparison learning model, which comprises the following specific steps of:
s100: the network failure data is converted into time series data.
In the embodiment of the application, the network fault data at least comprises a fault type, a fault occurrence time and a fault occurrence position. Converting network failure data into time series data, comprising: and converting the time node and the time-series reference according to the fault occurrence time in the network fault data into time-series data.
S200: the method comprises the steps of constructing a network fault prediction model, wherein the network fault prediction model comprises an embedding module, a heterogeneous path network, a self-supervision and comparison learning module and a prediction module, and the embedding module is used for receiving time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network comprises a first heterogeneous path network and a second heterogeneous path network which are respectively constructed based on different element paths; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting faults by adopting the loss function.
The construction method of the network failure prediction model will be described in detail below.
S210: an embedding module is constructed, and the embedding module is used for receiving the time series data and converting the time series data into an embedding vector x= (x) 1 ,x 2 ,x 3 ,...,x t ) T represents time.
S220: a first heterogeneous path network and a second heterogeneous path network are constructed.
The heterogeneous path network may be represented as g= (V, E, a, R), where V and E represent a set of node objects and edges, respectively; a represents a node object, and in the embodiment of the application, the node object comprises a fault node F, a fault occurrence time node T and a fault occurrence position node L; r represents the relationship type among nodes, and the relationship type among nodes refers to the relationship among node objects, and the relationship among nodes is taken as an edge in the heterogeneous path network, wherein the I A+R I is more than 2. The relationship types between nodes include four classes: the relationship between the fault and the fault occurrence time, the relationship between the fault and the fault, the relationship between the fault occurrence time and the fault occurrence time, and the relationship between the fault and the fault occurrence position. For example, a fault occurs at a certain place at a certain time, which is the relationship among the corresponding fault node, the fault occurrence time node, and the fault occurrence position node; as another example, two faults are associated or similar, which is the relationship between the corresponding faulty nodes. Definition of meta-paths in heterogeneous meta-path networksWherein A is 1 、A 2 、...A n Representing different node objects, R 1 、R 2 、...R n Representing relationships between node objects.
The embodiment of the application constructs two element paths with different connection modes, and codes the two element paths to respectively obtain a first heterogeneous element path network and a second heterogeneous element path network. The construction of the first heterogeneous path network and the second heterogeneous path network will be described below, respectively.
(1) Construction of first heterogeneous path network
In the construction of the first meta-path, the node object considers only the failure node F and the failure occurrence time node T, and considers only the meta-path having a path length of 2 or 3, based on which the meta-path is constructed. For the constructed multiple element paths, the characteristic value Eig (P) of the element path P is calculated by adopting the following formula:
in the formula (1), the characteristic value Eig (P) is the sum of the association degrees of the constructed multiple element paths, P represents an example path of P, and cor (P) represents the association degree between nodes under the example path.
For example path p= (a) 1 ,a 2 ,…,a k ) Wherein a is 1 ∈F，a k ∈F，a 1 And a k All represent fault nodes, the others are intermediate nodes of example paths, and the object types of the intermediate nodes are not limited. Let cor (p) be the particle following the example path p from a 1 Walk to a k Then the estimation formula for cor (p) is as follows:
in the formula (2), pro (a) f ,a f+1 ) A random walk-in node a represented in example path p f To a f+1 Probability of a) f+1 To at node a f Surrounding randomly selected nodes, i.e. a f+1 For node a f Is a random neighbor node of the network.
Pro(a f ,a f+1 ) The calculation of (2) is as follows:
in the formula (2), ω (a) f ,a f+1 ) Representing node a f And a f+1 The weight of the space; instance Path p= (v) 1 ,v 2 ,…,v n+1 )，v i ∈V，ω(a f ,v i ) Representing node a f And v i The weight of the space; n (a) f ) Representing node a in instance path p f And a in a neighbor node of (a) f+1 A set of nodes of a consistent type.
Based on the above, the method for constructing the first heterogeneous path network in the embodiment of the present application includes:
and obtaining fault type and fault occurrence time data from the time sequence data, taking the fault type and the fault occurrence time as node objects, and constructing a first heterogeneous path network by taking the relation between the faults and the fault occurrence time and the relation between the faults and the relation between the fault occurrence time and the fault occurrence time as edges.
And calculating the characteristic value Eig (P) of all element paths in the first heterogeneous element path network, adding random walk to calculate the relevance cor (P) among nodes in each example path respectively, and outputting the relevance cor (P). The calculation of the degree of association cor (p) is shown in equation (2).
(2) Construction of a second heterogeneous path network
In the construction of the second meta-path, the node object also joins the failure occurrence location node L but only considers the meta-path with path 2, based on which M meta-paths { P } are constructed 1 ,P 2 ,P 3 ,...,P M }. Considering that there are different types of node objects and features exist in different spaces, all types of node objects are projected into a common potential vector space, the same type of node object is placed into a type-specific mapping matrix, thereby feature x of the node object b Transforming into common space, feature x b Extracting from the embedded vector; node embedding is learned from the perspective of a higher-order meta-path structure. Specifically, from M meta-paths { P ] 1 ,P 2 ,P 3 ,...,P M Randomly selecting two different element paths, respectively denoted as P n P m Node b selected from random 1 Initially, node b is obtained separately 1 Based on meta-path P n P m Is a set of neighbor nodes of (a)
Meta path P n Specific coding features are given in the following formula:
in the formulas (4) to (5):representing node b 1 Is a projection feature of (2); sigma (·) represents the activation function; />Representing node b 1 Is characterized by (2); />Representing the vector deviation; phi (phi) b1 Representative and node b 1 Nodes of the same type->Represents a mapping matrix of a particular type, referred to as node b 1 The type of (2); d, d b And d b' Representing the degree of nodes b and b', h b And h b' The projected features representing nodes b and b', respectively, can be obtained by calculation of equation (4).
For node b 1 M-ary paths { P ] 1 ,P 2 ,P 3 ,...,P M Embedding in }
Using the attention mechanismBlend into the final embedding:
in the above, z b Representing node information fused under the current construction path structure, namely an output value of the second heterogeneous path network; beta p n Is a measure of meta-path P n Importance weight of W.epsilon.R d×d And r.epsilon.R d×1 Are all parameters that can be learned and are,representing semantic attention vectors, V represents a set of nodes, ωp n Is used to balance p n T is the transpose, tanh represents the activation function,the characteristics of the code for convolutional neural networks.
Based on the above, in the embodiment of the present application, the second heterogeneous path network is constructed as follows:
and acquiring fault type, fault occurrence time and fault occurrence position data from the time sequence data, taking the fault type, the fault occurrence time and the fault occurrence position as node objects, and constructing a second heterogeneous path network by taking the relation between the fault type and the fault occurrence time, the relation between the fault occurrence time and the fault occurrence time as edges.
Respectively calculating projection characteristics of each node object in the second heterogeneous meta-path network, designing two meta-paths using different node objects, fusing embedding of the two meta-paths by using an attention mechanism, and outputting
S230: and constructing a self-supervision and comparison learning module, wherein the self-supervision and comparison learning module is used for constructing a fault network view of the fault node to perform self-supervision learning.
Taking a failure network view of the same failure node as a dead sideThe two fault network views of the same fault node are respectively the output of the first heterogeneous path network and the second heterogeneous path network; taking the fault network view of any two different fault nodes as a negative pair +.>Likewise, the failure network views of two different failure nodes are the outputs of the first and second heterogeneous path networks, respectively.
Follow SimCLR and model with contrast loss InfoNCE to obtain final network failure prediction loss function value L loss The results were as follows:
in equations (9) - (11), s (·) is used to measure the similarity between the two vectors, in the embodiment of the application, a cosine similarity function is used, τ represents a temperature parameter, typically set to 0.1 or 0.2, v represents a node set,and->Loss function values generated under two-element path view angles respectively, η is balance +.>And->The coefficient between them is generally set to 0.5.
The smaller the prediction loss value, the higher the prediction accuracy. The self-supervision contrast learning aims at maximizing the mutual information of two views, thereby realizing global learning.
S240: the prediction module is configured to receive a given candidate node sequence and predict a probability that the candidate node will be the next failure.
Note that the above is only a preferred embodiment of the present application and the technical principle applied. It will be understood by those skilled in the art that the present application is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the application. Therefore, while the application has been described in connection with the above embodiments, the application is not limited to the above embodiments, but may include many other equivalent embodiments without departing from the spirit of the application, which fall within the scope of the application.
Claims (5)
1. The network fault prediction method based on the full-heterogeneous element comparison learning model is characterized by comprising the following steps of:
s100: converting network fault data into time series data, wherein the network fault data at least comprises a fault type, fault occurrence time and fault occurrence position;
s200: constructing a network fault prediction model, wherein the network fault prediction model comprises an embedding module, a heterogeneous path network, a self-supervision comparison learning module and a prediction module; the embedding module is used for receiving the time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network comprises a first heterogeneous path network and a second heterogeneous path network which are respectively constructed based on different element paths; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting the fault probability by adopting a loss function;
the first heterogeneous path network is constructed by the following steps: obtaining fault type and fault occurrence time data from the time sequence data, taking the fault type and the fault occurrence time as node objects, and constructing a first heterogeneous path network by taking the relation between the faults and the fault occurrence time and the relation between the faults and the relation between the fault occurrence time and the fault occurrence time as edges; adding random walk so as to respectively calculate the association degree between nodes in each instance path and output;
the construction of the second heterogeneous path network is as follows: obtaining fault type, fault occurrence time and fault occurrence position data from the time sequence data, taking the fault type, the fault occurrence time and the fault occurrence position as node objects, and constructing a second heterogeneous path network by taking the relation between the fault type and the fault occurrence time, the relation between the fault occurrence time and the fault occurrence time as edges; and respectively calculating projection characteristics of each node object in the second heterogeneous meta-path network, designing two meta-paths using different node objects, and fusing the embedding and output of the two meta-paths by using an attention mechanism.
2. The network failure prediction method based on the full heterogeneous element contrast learning model as claimed in claim 1, wherein the method is characterized in that:
the association degree between nodes in each instance pathWherein Pro (a) f ,a f+1 ) A random walk-in node a represented in example path p f To a f+1 Probability of a) f+1 To at node a f Surrounding randomly selected nodes, i.e. a f+1 For node a f Is a random neighbor node of the network.
3. The network failure prediction method based on the full heterogeneous element contrast learning model as claimed in claim 1, wherein the method is characterized in that:
embedding of the second heterogeneous path network outputWherein:
βp n is an important weight for measuring the meta-path,
ωp n for balancing p n Is of importance in terms of (a) the importance of (c),
t represents the transpose, tanh represents the activation function,for the characteristic of convolutional neural network coding, W E R d×d And r.epsilon.R d×1 Are all learnable parameters, < >>Representing a semantic attention vector, V represents a node set.
4. The network failure prediction method based on the full heterogeneous element contrast learning model as claimed in claim 1, wherein the method is characterized in that:
the self-supervision and contrast learning module is used for constructing a view fault network of a fault node to perform self-supervision learning, and further comprises:
the constructed fault network view comprises a positive pair and a negative pair, wherein the fault network view of the same fault node is used as the positive pair, and the two fault network views of the same fault node are respectively the output of a first heterogeneous path network and a second heterogeneous path network; and taking the fault network views of any two different fault nodes as negative pairs, wherein the fault network views of the two different fault nodes are respectively the output of the first heterogeneous path network and the second heterogeneous path network.
5. The network fault prediction system based on the full heterogeneous element comparison learning model is characterized by comprising the following components:
the first module is used for converting network fault data into time series data, wherein the network fault data at least comprises a fault type, a fault occurrence time and a fault occurrence position;
the second module further comprises an embedding module, a heterogeneous path network, a self-supervision and contrast learning module and a prediction module; the embedding module is used for receiving the time sequence data and converting the time sequence data into an embedding vector; the heterogeneous path network comprises a first heterogeneous path network and a second heterogeneous path network which are respectively constructed based on different element paths; the self-supervision comparison learning module is used for comparing and learning the fault network view of the fault node; the prediction module is used for predicting the fault probability by adopting a loss function;
the first heterogeneous path network is constructed by the following steps: obtaining fault type and fault occurrence time data from the time sequence data, taking the fault type and the fault occurrence time as node objects, and constructing a first heterogeneous path network by taking the relation between the faults and the fault occurrence time and the relation between the faults and the relation between the fault occurrence time and the fault occurrence time as edges; adding random walk so as to respectively calculate the association degree between nodes in each instance path and output;
the construction of the second heterogeneous path network is as follows: obtaining fault type, fault occurrence time and fault occurrence position data from the time sequence data, taking the fault type, the fault occurrence time and the fault occurrence position as node objects, and constructing a second heterogeneous path network by taking the relation between the fault type and the fault occurrence time, the relation between the fault occurrence time and the fault occurrence time as edges; and respectively calculating projection characteristics of each node object in the second heterogeneous meta-path network, designing two meta-paths using different node objects, and fusing the embedding and output of the two meta-paths by using an attention mechanism.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310276842.9A CN116633758A (en) | 2023-03-21 | 2023-03-21 | Network fault prediction method and system based on full-heterogeneous element comparison learning model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310276842.9A CN116633758A (en) | 2023-03-21 | 2023-03-21 | Network fault prediction method and system based on full-heterogeneous element comparison learning model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116633758A true CN116633758A (en) | 2023-08-22 |
Family
ID=87612269
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310276842.9A Pending CN116633758A (en) | 2023-03-21 | 2023-03-21 | Network fault prediction method and system based on full-heterogeneous element comparison learning model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116633758A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117560275A (en) * | 2023-12-29 | 2024-02-13 | 安徽思高智能科技有限公司 | Root cause positioning method and device for micro-service system based on graphic neural network model |
-
2023
- 2023-03-21 CN CN202310276842.9A patent/CN116633758A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117560275A (en) * | 2023-12-29 | 2024-02-13 | 安徽思高智能科技有限公司 | Root cause positioning method and device for micro-service system based on graphic neural network model |
| CN117560275B (en) * | 2023-12-29 | 2024-03-12 | 安徽思高智能科技有限公司 | Root cause positioning method and device for micro-service system based on graphic neural network model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109596913B (en) | Charging pile fault cause diagnosis method and device | |
| CN116505665B (en) | Fault monitoring method and system for power grid distribution line | |
| CN114169091B (en) | Method for establishing prediction model of residual life of engineering machinery part and prediction method | |
| CN115168443A (en) | Anomaly detection method and system based on GCN-LSTM and attention mechanism | |
| CN112822052B (en) | Network fault root cause positioning method based on network topology and alarm | |
| Tang et al. | Reputation-aware data fusion and malicious participant detection in mobile crowdsensing | |
| CN113821418B (en) | Fault root cause analysis method and device, storage medium and electronic equipment | |
| Eliassi et al. | Application of Bayesian networks in composite power system reliability assessment and reliability‐based analysis | |
| CN113761073A (en) | Method, apparatus, device and storage medium for information processing | |
| CN113191530B (en) | Block link point reliability prediction method and system with privacy protection function | |
| CN113191062B (en) | Power distribution network fault section positioning method and system based on multisource incomplete information | |
| CN116894096A (en) | News event prediction method based on recursive double hypergraph neural network | |
| CN116633758A (en) | Network fault prediction method and system based on full-heterogeneous element comparison learning model | |
| CN115718846A (en) | Big data mining method and system for intelligent interactive network | |
| CN104657434A (en) | Construction method for social network structure | |
| CN110659394A (en) | Recommendation method based on two-way proximity | |
| CN114757391B (en) | Network data space design and application method oriented to service quality prediction | |
| CN116862021B (en) | Anti-Bayesian-busy attack decentralization learning method and system based on reputation evaluation | |
| CN118037048A (en) | Urban storm waterlogging disaster chain risk assessment system based on Bayesian network | |
| CN113541986A (en) | Fault prediction method and device for 5G slice and computing equipment | |
| CN112396166A (en) | Graph convolution neural network training method and device based on mixed granularity aggregator | |
| CN117290786A (en) | Fault detection method based on iterative depth time sequence causal discovery | |
| Yeh | The extension of universal generating function method to search for all one-to-many $ d $-minimal paths of acyclic multi-state-arc flow-conservation networks | |
| CN114036723B (en) | Method, device, equipment and storage medium for predicting running state of comprehensive energy system | |
| CN115455302A (en) | Knowledge graph recommendation method based on optimized graph attention network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |