CN115567406A - Method, device and system for managing network nodes - Google Patents

Method, device and system for managing network nodes Download PDF

Info

Publication number
CN115567406A
CN115567406A CN202211142185.0A CN202211142185A CN115567406A CN 115567406 A CN115567406 A CN 115567406A CN 202211142185 A CN202211142185 A CN 202211142185A CN 115567406 A CN115567406 A CN 115567406A
Authority
CN
China
Prior art keywords
model
network
network node
indexes
index
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
Application number
CN202211142185.0A
Other languages
Chinese (zh)
Inventor
张静
张宪波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jingdong Technology Information Technology Co Ltd
Original Assignee
Jingdong Technology Information Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jingdong Technology Information Technology Co Ltd filed Critical Jingdong Technology Information Technology Co Ltd
Priority to CN202211142185.0A priority Critical patent/CN115567406A/en
Publication of CN115567406A publication Critical patent/CN115567406A/en
Pending legal-status Critical Current

Links

Images

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
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • H04L41/142—Network analysis or design using statistical or mathematical methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Evolutionary Computation (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Algebra (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Medical Informatics (AREA)
  • Probability & Statistics with Applications (AREA)

Abstract

The invention discloses a method, a device and a system for managing network nodes, and relates to the technical field of computers. One embodiment of the method comprises: determining a first operational value of the network node based on the first operational indicator of the network node; determining a second operation value of the network node according to the first operation index and a second operation index of the associated network node; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node. The embodiment of the invention improves the universality and flexibility of the management network node and improves the accuracy rate of judging the running condition of the network node; the automation degree of the management network node is improved to a great extent, and the labor cost and the time cost of operation and maintenance are reduced.

Description

Method, device and system for managing network nodes
Technical Field
The present invention relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, and a system for managing network nodes.
Background
In the internet application system, a large number of background network nodes are included, and the running of the application is jointly completed by running multiple types of software (such as middleware and other software) on each network node.
At present, a method for managing network nodes generally includes that an operation and maintenance worker monitors a value of an operation index set on a network node, and judges whether an operation condition of the network node is abnormal according to a set judgment strategy after analyzing the value of the operation index; the existing method for managing the operation condition of the network node depends on the experience of operation and maintenance personnel, and has the problems of poor flexibility and poor universality, so that the accuracy rate of judging the operation condition of the network node is low, and the waste of labor cost and time cost of the operation and maintenance personnel is caused.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method, an apparatus, and a system for managing a network node, which can determine a first operation value of the network node according to a first operation index of the network node; determining a second operation value of the network node according to the first operation index and a second operation index of the associated network node; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node. The embodiment of the invention improves the universality and flexibility of the management network node and improves the accuracy rate of judging the running condition of the network node; the automation degree of the management network node is improved to a great extent, and the labor cost and the time cost of operation and maintenance are reduced.
To achieve the above object, according to an aspect of an embodiment of the present invention, there is provided a method for managing a network node, including: respectively obtaining a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index both comprise multidimensional indexes; determining a first operation numerical value of the network node according to the multidimensional index included by the first operation index; determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node.
Optionally, the determining a first operation value of the network node includes:
inputting multidimensional indexes included by the first operation indexes into a neural network model, and outputting a first operation numerical value of the network node by using the neural network model; wherein the neural network model comprises a long-short term memory network and a full convolution network; and integrating the output of the long-short term memory network aiming at the first operation index and the output of the full convolution network aiming at the first operation index to obtain the first operation numerical value.
Optionally, the determining a second operation value of the network node includes:
inputting a detection model for the multidimensional indexes included by the first operation indexes of the network nodes and the multidimensional indexes included by the second operation indexes of the other network nodes, and outputting second operation values of the network nodes relative to the other nodes by using the detection model, wherein the detection model comprises a preset dimensionality reduction model and an adaptive clustering model.
Optionally, the outputting, by the detection model, a second operation value of the network node relative to the plurality of other nodes includes: performing dimensionality reduction operation on the multidimensional index included by the first operation index and the multidimensional index included by the second operation index by using the preset dimensionality reduction model to obtain a first characteristic index of the network node and a plurality of second characteristic indexes corresponding to other network nodes; and inputting the first characteristic index and a plurality of second characteristic indexes into the self-adaptive clustering model, and calculating a second operation numerical value of the network node.
Optionally, the method for managing a network node further includes: performing iterative training on a preset dimensionality reduction model and a self-adaptive clustering model by using the obtained running indexes of a plurality of network nodes in a historical time range, judging whether the training result meets an iteration stop condition or not aiming at the training result of each iteration, and if so, determining that dimensionality reduction core parameters included in the training result are dimensionality reduction core parameters of the preset dimensionality reduction model; and combining a preset dimension reduction model with the dimension reduction core parameters and the self-adaptive clustering model to obtain the detection model.
Optionally, the determining whether the training result satisfies an iteration stop condition includes: comparing each first model evaluation value included in the training result with a second model evaluation value obtained by training the acquired operation index of the network node in the historical time range through a neural network model; and determining that the training result meets an iteration stop condition under the condition that the comparison result is that the difference between each first model evaluation value and each second model evaluation value meets a preset error range.
Optionally, the performing iterative training on the preset dimensionality reduction model and the adaptive clustering model by using the obtained operation indexes of the plurality of network nodes in the historical time range includes: circularly executing the following steps from N1 to N3 until an iteration stop condition is met; n1: performing dimension reduction operation on the operation index in the historical time range by using the preset dimension reduction model and the dimension reduction core parameter of the current cycle period to obtain a characteristic index for training of the current cycle period; n2: inputting the characteristic indexes for training of the current cycle period into the self-adaptive clustering model, and determining a first model evaluation value according to the output of the self-adaptive clustering model; n3: and under the condition that the iteration stopping condition is not met, adjusting the dimensionality reduction core parameter of the current cycle period, taking the adjusted dimensionality reduction core parameter as the dimensionality reduction core parameter of the next cycle period, and executing the step N1.
Optionally, the adaptive clustering model comprises an iterative computational model of processing features;
for each iterative training, further comprising: and utilizing the iterative computation model to adaptively compute the clustering radius of the characteristic data output by the preset dimensionality reduction model so as to output a clustering result based on the clustering radius, and enabling the clustering result to be optimal after iteration is stopped.
Optionally, the method for managing a network node further includes:
iteratively performing the following operations to train a neural network model: respectively inputting the obtained operation indexes of the plurality of network nodes in a historical time range into the long-short term memory network and the full convolution network, and combining the output of the long-short term memory network aiming at the operation indexes in the historical time range with the output of the full convolution network aiming at the operation indexes in the historical time range to obtain the output of the neural network model; evaluating a training result of the neural network model by using a preset loss function and the output of the neural network model; and adjusting the neural network model according to the evaluation result.
Optionally, the method for managing a network node further includes:
acquiring characteristic data of the operation indexes in the historical time range, inputting the characteristic data into an attention mechanism model, and increasing the weight value of the characteristic data close to the current time by using the attention mechanism model; the inputting the obtained operation indexes of the plurality of network nodes in the historical time range into the long-short term memory network comprises: inputting the feature data processed by the attention mechanism model into the long-short term memory network to train the long-short term memory network.
Optionally, the method for managing a network node further includes: obtaining the first operation index of the network node; inputting the first operation index into a time sequence prediction model, predicting the operation condition of the network node in a future time range by using the time sequence prediction model, wherein the time sequence prediction model is obtained by performing periodic characteristic training based on the operation index in a historical time range and performing periodic adjustment on a prediction result predicted by the time sequence prediction model.
To achieve the above object, according to a second aspect of an embodiment of the present invention, there is provided an apparatus for managing a network node, including: the system comprises a node operation index acquisition module, a node operation numerical value acquisition module and a node operation condition determination module; wherein, the first and the second end of the pipe are connected with each other,
the node operation index acquisition module is used for respectively acquiring a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index both comprise multidimensional indexes;
the node operation value obtaining module is used for determining a first operation value of the network node according to the multidimensional indexes included by the first operation indexes; determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index;
and the node operation condition determining module is used for determining the operation condition of the network node based on the first operation numerical value and the second operation numerical value so as to manage the network node based on the operation condition of the network node.
Optionally, the apparatus for managing a network node, configured to determine a first operation value of the network node, includes: inputting the multidimensional indexes included by the first operation indexes into a neural network model, and outputting first operation numerical values of the network nodes by using the neural network model; wherein the neural network model comprises a long-short term memory network and a full convolution network; and integrating the output of the long-short term memory network aiming at the first operation index and the output of the full convolution network aiming at the first operation index to obtain the first operation numerical value.
Optionally, the apparatus for managing a network node, configured to determine a second operation value of the network node, includes: inputting a detection model for the multidimensional indexes included by the first operation indexes of the network nodes and the multidimensional indexes included by the second operation indexes of the other network nodes, and outputting second operation values of the network nodes relative to the other nodes by using the detection model, wherein the detection model comprises a preset dimensionality reduction model and an adaptive clustering model.
Optionally, the means for managing the network node is configured to output a second operation value of the network node relative to the plurality of other nodes by using the detection model, and includes: performing dimensionality reduction operation on the multidimensional indexes included by the first operation indexes and the multidimensional indexes included by the second operation indexes by using the preset dimensionality reduction model to obtain first characteristic indexes of the network nodes and second characteristic indexes corresponding to the other network nodes; and inputting the first characteristic index and a plurality of second characteristic indexes into the self-adaptive clustering model, and calculating a second operation numerical value of the network node.
Optionally, the apparatus for managing a network node is a method for managing a network node, and further includes: performing iterative training on a preset dimensionality reduction model and a self-adaptive clustering model by using the obtained running indexes of a plurality of network nodes in a historical time range, judging whether the training result meets an iteration stop condition or not aiming at the training result of each iteration, and if so, determining that dimensionality reduction core parameters included in the training result are dimensionality reduction core parameters of the preset dimensionality reduction model; and combining a preset dimension reduction model with the dimension reduction core parameters and the self-adaptive clustering model to obtain the detection model.
Optionally, the device for managing a network node, configured to determine whether the training result satisfies an iteration stop condition, includes: comparing each first model evaluation value included in the training result with a second model evaluation value obtained by training the acquired operation index of the network node in the historical time range through the neural network model; and determining that the training result meets an iteration stop condition under the condition that the comparison result is that the difference value between each first model evaluation value and each second model evaluation value meets a preset error range.
Optionally, the apparatus for managing network nodes is configured to perform iterative training on a preset dimension reduction model and an adaptive clustering model by using the obtained operation indexes of the plurality of network nodes in a historical time range, and includes: circularly executing the following steps from N1 to N3 until an iteration stop condition is met; n1: performing dimension reduction operation on the operation index in the historical time range by using the preset dimension reduction model and the dimension reduction core parameter of the current cycle period to obtain a characteristic index for training of the current cycle period; n2: inputting the characteristic indexes for training of the current cycle period into the self-adaptive clustering model, and determining a first model evaluation value according to the output of the self-adaptive clustering model; n3: and under the condition that the iteration stopping condition is not met, adjusting the dimensionality reduction core parameter of the current cycle period, taking the adjusted dimensionality reduction core parameter as the dimensionality reduction core parameter of the next cycle period, and executing the step N1.
Optionally, the apparatus for managing a network node includes: the adaptive clustering model comprises an iterative computation model for processing the features; for each iterative training, further comprising: and utilizing the iterative computation model to adaptively compute the clustering radius of the characteristic data output by the preset dimension reduction model so as to output a clustering result based on the clustering radius, and enabling the clustering result to be optimal after iteration is stopped.
Optionally, the apparatus for managing network nodes is further configured to iteratively perform the following operations to train the neural network model: respectively inputting the obtained operation indexes of the plurality of network nodes in a historical time range into the long-short term memory network and the full convolution network, and combining the output of the long-short term memory network aiming at the operation indexes in the historical time range with the output of the full convolution network aiming at the operation indexes in the historical time range to obtain the output of the neural network model; evaluating a training result of the neural network model by using a preset loss function and the output of the neural network model; and adjusting the neural network model according to the evaluation result.
Optionally, the device for managing a network node is further configured to obtain feature data of the operation index in the historical time range, input the feature data into an attention mechanism model, and increase, by using the attention mechanism model, a weight value of the feature data closer to the current time; the step of inputting the acquired operation indexes of the plurality of network nodes in the historical time range into the long-short term memory network comprises the following steps: and inputting the characteristic data processed by the attention mechanism model into the long-short term memory network to train the long-short term memory network.
Optionally, the means for managing a network node is further configured to obtain the first operation index of the network node; inputting the first operation index into a time series prediction model, predicting the operation condition of the network node in a future time range by using the time series prediction model, wherein the time series prediction model is obtained by performing periodic characteristic training based on the operation index in a historical time range and performing periodic adjustment on a prediction result predicted by the time series prediction model.
To achieve the above object, according to a third aspect of an embodiment of the present invention, there is provided a system for managing a network node, including: a plurality of network nodes, at least one network node having the apparatus for managing network nodes of the second aspect; at least any two of the plurality of network nodes are communicatively coupled.
To achieve the above object, according to a fourth aspect of embodiments of the present invention, there is provided an electronic device for managing a network node, comprising: one or more processors; storage means for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement a method as in any one of the methods of managing a network node described above.
To achieve the above object, according to a fifth aspect of embodiments of the present invention, there is provided a computer readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method as in any one of the methods of managing network nodes described above.
One embodiment of the above invention has the following advantages or benefits: determining a first operational value of the network node based on the first operational indicator of the network node; determining a second operation value of the network node according to the first operation index and a second operation index of the associated network node; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node. The universality and flexibility of the management network node are improved, and the accuracy of judging the running condition of the network node is improved; the automation degree of the management network node is improved to a great extent, and the labor cost and the time cost of operation and maintenance are reduced.
Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
fig. 1 is a flowchart illustrating a method for managing network nodes according to an embodiment of the present invention;
fig. 2 is a flow diagram of a management network node according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of training a detection model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an apparatus for managing a network node according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a system for managing network nodes according to an embodiment of the present invention;
FIG. 6 is an exemplary system architecture diagram in which embodiments of the present invention may be employed;
fig. 7 is a schematic block diagram of a computer system suitable for use in implementing a terminal device or server according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, in which various details of embodiments of the invention are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
As shown in fig. 1, an embodiment of the present invention provides a method for managing a network node, where the method may include the following steps:
step S101: the method comprises the steps of respectively obtaining a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index both comprise multidimensional indexes.
Specifically, in a more complex internet application system, the backend includes a plurality of network nodes, and one or more types of software are run on the network nodes to implement the service functions of the application system together, for example: the network node runs middleware; the middleware is a type of software between the software, and can be connected with each part of the application system or other application systems so as to achieve the purposes of resource sharing and function sharing.
Further, in order to determine whether the network node is abnormal, a first operation index of the network node is obtained, where the first operation index includes a multidimensional index, for example, the obtained multidimensional index includes: transaction throughput per second, used memory, total number of keywords, traffic (in-and/or out-going), connection client data, expiration/culling amount of keywords per second, client newly-built connections per second, query hit rate, etc.
Further, a second operational metric of a plurality of other network nodes associated with the network node, wherein, similar to the first operational metric, the second operational metric also includes a multidimensional metric; wherein, it is assumed that there are 100 network nodes in one application system or multiple associated application systems, and the following are: network node 0\8230, network node 99; if the network node 0 is a managed (e.g., monitored) network node, the network nodes 1 to 99 are all other network nodes associated with the network node 0, where the association may be direct association (e.g., directly perform network connection to transmit data), indirect connection (e.g., not directly connect but provide setting services together, etc.); it is to be understood that evaluating the operation of a network node (associated with the first operation index) includes evaluating the operation of the network node itself, and also includes evaluating the operation of the network node in the population(s) (associated with the first operation index and the second operation index).
Step S102: determining a first operation numerical value of the network node according to the multidimensional indexes included in the first operation indexes; and determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index.
Specifically, a first operation numerical value of the network node is determined according to a multidimensional index included in a first operation index acquired from the network node in real time. Specifically, one embodiment of determining the first running value is: inputting the multidimensional indexes included by the first operation indexes into a neural network model, and outputting first operation numerical values of the network nodes by using the neural network model; one embodiment of determining the second running value of the network node is: and inputting the multidimensional indexes included by the first operation indexes of the network nodes and the multidimensional indexes included by the second operation indexes of the other network nodes into a detection model, and outputting second operation values of the network nodes relative to the other nodes by using the detection model. Wherein the first operational value or the second operational data may be calculated based on a probability value output by the model.
Further, the neural network model, the structure of the detection model and the calculation flow are consistent with the descriptions of step S201 to step S205, and are not described herein again.
Step S103: determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node.
Specifically, based on the first operation value (obtained by using a neural network model or the like, for example) and the second operation value (obtained by using a detection model or the like, for example), the operation condition of the network node is determined, for example: the operation condition of the network node is determined by using a formula a × score1+ b × score2, where score1 is a first operation numerical value, score2 is a second operation numerical value, exemplarily, a =0.5, and b =0.5, and the result numerical values may be uniformly mapped to 0 to 100 points according to the calculation result, it is understood that the more the points are, the more the network node is abnormal, and the specific value ranges and numerical values of a and b are not limited in the present invention.
Further, for example, the result calculated based on the first operation value and the second operation value is 0 to indicate that the network node is not abnormal and 1 to indicate that the network node is abnormal, for example, (0, 1) indicates that the first operation value is 0 and the second operation value is 1; (0, 0) indicates that the first operating value is 0 and the second operating value is also 0; and so on; then (0, 0) represents that the output indication by the neural network model is normal and the output indication by the detection model is normal, (0, 1) represents that the output indication by the neural network model is normal and the output indication by the detection model is abnormal, (1, 0) represents that the output indication by the neural network model is abnormal and the output indication by the detection model is normal, and (1, 1) represents that the outputs of the neural network model and the detection model are both indicated as abnormal. When the output of any model is abnormal, the network node can be judged to have abnormality, namely not only the network node which has no abnormality but has abnormality in the middle of a group (a plurality of other related network nodes) can be detected, but also the network node which has no abnormality in the group but has abnormality can be detected, and the specific index causing the abnormality can be positioned for the network node with the abnormality, so that the efficiency of positioning the abnormality by an operation and maintenance engineer is improved. That is, the network node is managed based on the operation condition of the network node.
Managing the network nodes based on the operational conditions of the network nodes, such as: according to the calculated scores, the identifiers of the network nodes in the set number and the specific index information corresponding to the abnormity can be sent to network operation and maintenance personnel from high to low in sequence, so that the corresponding processing can be performed in real time aiming at the abnormity condition.
As shown in fig. 2, an embodiment of the present invention provides a process for managing a network node, where the process may include the following steps:
step S201: and acquiring a multi-dimensional index included by the first operation index.
Specifically, the description of obtaining the multidimensional index included in the first operation index is consistent with the description of step S101, and is not repeated here.
Step S202: and acquiring the multidimensional indexes of the second operation indexes of a plurality of other network nodes related to the network node.
Specifically, the description about obtaining the multidimensional index included in the second operation index is consistent with the description in step S101, and is not repeated here.
Step S203: a first running value is output using the neural network model.
Specifically, in one embodiment of the present invention, the neural Network model includes a Long Short-Term Memory Network (LSTM) and a full volume Network (FCN).
Further, the determining a first running value of the network node comprises: inputting the multidimensional indexes included by the first operation indexes into a neural network model, and outputting first operation numerical values of the network nodes by using the neural network model; wherein the neural network model comprises a long-short term memory network and a full convolution network; and integrating the output of the long-short term memory network aiming at the first operation index and the output of the full convolution network aiming at the first operation index to obtain the first operation numerical value.
Specifically, the flow of one embodiment of training the neural network model including the long-short term memory network and the full convolution network is as follows:
1) Acquiring a multi-dimensional index (namely training data) of an operation index in a historical time range, then performing data reconstruction, processing screened index data into a set dimension (such as three dimensions), and setting the sample size of the index transmitted each time as the length of the data; setting the number of indexes reserved after the characteristic screening to be high; in order to be able to handle the anomaly persistence phenomenon, the average anomaly duration is set to be wide, and then the reconstructed data is input: LSTM model and FCN model.
2) The training data input into the LSTM block is set to 128 × 10 × 5, taking a sample as an example, the input data at each time step is 10 × 1, and an output of 10 × 1 is obtained, that is, the data structure output by the LSTM before the connection step is a matrix of 30 × 1, and 128 is a batch size, that is, the number of samples selected in one training, and the full data can be traversed after multiple training. Before inputting the LSTM model, distributing different weights to the training data through an attention mechanism model according to the distance between the time sequence and the current time (namely the time range of detecting the network node) to obtain new data, wherein the data related to the time sequence in the new data obtains a larger weight value; inputting the feature data processed by the attention mechanism model into the long-short term memory network to train the long-short term memory network; by the steps, the weights of the detection indexes close to the detection time point can be increased, the weights of the detection indexes far away from the detection time point are reduced, and the accuracy of the neural network model in detecting the operation condition of the network node is improved.
3) The training data input to the FCN block is set to 1 × 5 × 10, the convolution kernel is a matrix of 3 × 30, and the feature data is 128, the input data is scanned step by step using the convolution kernel matrix, the corresponding positions are multiplied and added, and meanwhile, 0 is used to fill the input data, so that 128 feature data of 5 × 10 are obtained.
4) Splicing the data matrix of 10 x 1 output by the LSTM and the data matrix of 128 x 1 output by the FCN to obtain a matrix of (10 + 128) 1, namely integrating the output results of the two models, and outputting prediction probability (namely the probability that the prediction state is abnormal) through a classifier (such as a Softmax function); that is, the output of the neural network model is obtained by combining the output of the long-short term memory network for the operation index in the historical time range with the output of the full convolution network for the operation index in the historical time range.
5) Evaluating the training result by using a preset Loss function, wherein the preset Loss function is Loss = cross control (y', y); and minimizing the pre-set loss function using an Adam optimizer while updating the parameters to be trained of LSTM and FCN by back propagation (i.e., adjusting the neural network model based on the results of the evaluation; wherein the attention mechanism model does not involve parameter updating).
The description of steps 1) -5) is: acquiring characteristic data of the operation index in the historical time range, inputting the characteristic data into an attention mechanism model, and increasing the weight value of the characteristic data closer to the current time by using the attention mechanism model; the step of inputting the acquired operation indexes of the plurality of network nodes in the historical time range into the long-short term memory network comprises the following steps: inputting the feature data processed by the attention mechanism model into the long-short term memory network to train the long-short term memory network.
Further, the following operations are iteratively performed to train the neural network model: respectively inputting the obtained operation indexes of the plurality of network nodes in a historical time range into the long-short term memory network and the full convolution network, and combining the output of the long-short term memory network aiming at the operation indexes in the historical time range with the output of the full convolution network aiming at the operation indexes in the historical time range to obtain the output of the neural network model; evaluating a training result of the neural network model by using a preset loss function and the output of the neural network model; and adjusting the neural network model according to the evaluation result.
Step S204: and outputting a second running numerical value by using the detection model.
Specifically, the detection model comprises a preset dimensionality reduction model and an adaptive clustering model.
In an embodiment of the present invention, the preset dimension reduction model is, for example: an optimized PCA (Principal Component Analysis) model, and an adaptive clustering model is, for example: optimized DBSCAN (sensitivity Based Spatial Clustering of Applications with Noise) model.
That is, the determining a second operational value of the network node comprises: inputting a detection model for the multidimensional indexes included by the first operation indexes of the network nodes and the multidimensional indexes included by the second operation indexes of the other network nodes, and outputting second operation values of the network nodes relative to the other nodes by using the detection model, wherein the detection model comprises a preset dimensionality reduction model and an adaptive clustering model.
Further, outputting, using the detection model, second running values of the network node relative to the plurality of other nodes, comprising: performing dimensionality reduction operation on the multidimensional indexes included by the first operation indexes and the multidimensional indexes included by the second operation indexes by using the preset dimensionality reduction model to obtain first characteristic indexes of the network nodes and second characteristic indexes corresponding to the other nodes; and inputting the first characteristic index and a plurality of second characteristic indexes into the self-adaptive clustering model, and calculating a second operation numerical value of the network node.
Preferably, the optimized preset dimension reduction model is a PCA (principal component analysis) model containing a kernel function, most information of the kernel function, which can keep network node operation index data, is introduced, and potential hidden variable data can be better estimated, so that the operation amount is reduced to enable the data to be linearly separable; the value discrimination of the PCA model aiming at the multi-dimensional indexes of part of network nodes can be improved, further, the optimal kernel method and the model hyper-parameters are searched by using a grid search method, wherein the optimal kernel method can be a polynomial kernel function selected from a plurality of kernel function methods (linear kernel, polynomial kernel function, gaussian kernel, exponential kernel and the like), and the universality of the whole model is improved.
Further, after the index data is subjected to dimensionality reduction by using a preset dimensionality reduction model, inputting feature data (including a first feature index and a second feature index) obtained after dimensionality reduction into a self-adaptive clustering model memorial clustering operation, so as to obtain a second operation numerical value of the network node (namely, an operation numerical value in a plurality of associated network nodes).
Preferably, two parameters may be determined first in the adaptive clustering model: epsilon: a radius of a neighboring area around a point; minPts: the number of points contained in the adjacent area; based on the above two parameters, in combination with the characteristics of epsilon-neighborwood, the points in the sample can be classified into three categories: core point (core point): if NBHD (p, epsilon) > = minPts is satisfied, the point is a nuclear sample point; edge point (border point): NBHD (p, epsilon) < minPts, but this point can be obtained by some core point (dense-reachable or directly-reachable); outlier (Outlier): if the point is neither a core point nor an edge point, the point is not in the category; under the conditions that the system is complex and the number of indexes corresponding to network nodes is large, the problem that the calculated amount of the existing DBSCAN model is large exists, the value distribution of multi-dimensional indexes of the network nodes is usually uneven, for example, the number of the corresponding network nodes in the same application system has a large difference, and therefore the problem that the accuracy of the existing density-based DBSCAN algorithm for managing the operation condition of the network nodes is low exists; preferably, an embodiment of the present invention optimizes the existing DBSCAN algorithm to obtain an adaptive clustering model. Specifically, the optimization method comprises the following steps: 1) An optimization method (L-BFGS), namely an iterative computation model, is introduced, each iterative computation in the existing BFGS algorithm needs a Hesse matrix obtained by the previous iteration, the storage space of the Hesse matrix is at least N (N + 1)/2, N is a characteristic dimension, and for an application scene with a higher dimension, the required storage space is huge. The present invention, however, utilizes L-BFGS to replace the previous Hesse matrix by storing a small amount of data from the previous m iterations. Therefore, the L-BFGS optimization method (namely the iterative computation model of the processing characteristics) can be used for greatly reducing the operation time and the computation cost. The optimal radius of the DBSCAN is automatically searched in a small range according to the input data through the optimization method, namely for the standardized data, the optimal radius is searched through an L-BFGS optimal search algorithm (namely, the clustering radius of the characteristic data output by the preset dimension reduction model is calculated in a self-adaptive mode), so that the CH coefficient is optimal (even if the clustering result is optimal); 2) And setting calinski _ harabasz _ score as a clustering level target function, setting the calinski _ harabasz _ score as a clustering index, also called CH score, and correspondingly representing that the clustering effect is better when the CH score value obtained by a clustering model is larger. In the DBSCAN iterative calculation process, the higher CH score indicates that the covariance of data in the classes is smaller, and the covariance between the classes is larger, so that the clustering effect and the operation efficiency are improved.
In summary, the multidimensional indexes included in the first operation indexes of the network nodes and the multidimensional indexes included in the second operation indexes of the other network nodes are input into a detection model (firstly, a preset dimensionality reduction model is used for dimensionality reduction, and then, an adaptive clustering model is used for clustering), and the detection model is used for outputting second operation values of the network nodes relative to the other nodes.
Step S205: determining an operational condition of the network node based on the first operational value and the second operational value.
Specifically, the description about determining the operation condition of the network node based on the first operation value and the second operation value is consistent with the description of step S103, and is not repeated here.
Step S206: and predicting the operation condition of the network node in a future time range by using the time series prediction model.
Specifically, based on the current multi-dimensional index data of the network nodes, the time series prediction model is utilized to predict the operation condition of the network nodes in the future time range; specifically, the first operation index of the network node is obtained; inputting the first operation index into a time series prediction model, predicting the operation condition of the network node in a future time range by using the time series prediction model, wherein the time series prediction model is obtained by performing periodic characteristic training based on the operation index in a historical time range and performing periodic adjustment on a prediction result predicted by the time series prediction model.
In an embodiment of the present invention, the time series prediction model may be a deep ar model; the DeepAR model is an upgraded Autoregressive model (Autoregressive model) which can output a probability distribution of future data based on data in a set historical time range; preferably, in order to improve the prediction effect of the DeepAR, feature processing is added to sequence data of an input model, noise data are removed, periodic features are transmitted to the DeepAR model, namely, the periodic features of the data are introduced in the training process of the DeepAR model, the prediction result is periodically corrected, the standard error of the time sequence prediction model is minimized by controlling the step length of model fitting, and the prediction accuracy of the time sequence prediction model is improved.
As shown in fig. 3, an embodiment of the present invention provides a method for training a detection model, which may include the following steps;
step S301: and obtaining the operation indexes of the plurality of network nodes in the historical time range.
Step S302: and performing dimensionality reduction operation on the operation index in the historical time range by using a preset dimensionality reduction model and the dimensionality reduction core parameter of the current cycle period to obtain the characteristic index for training of the current cycle period.
Step S303: and outputting the second model evaluation value by using the neural network model.
Step S304: inputting the characteristic indexes for training of the current cycle period into the self-adaptive clustering model, and determining a first model evaluation value according to the output of the self-adaptive clustering model.
Step S305: and judging whether an iteration stop condition is met, if so, executing step S307, otherwise, executing step S306.
Step S306: and adjusting the dimensionality reduction core parameter of the current cycle period, and taking the adjusted dimensionality reduction core parameter as the dimensionality reduction core parameter of the next cycle period.
Step S307: and determining a detection model comprising a preset dimension reduction model and an adaptive clustering model.
The description of step S301 to step S307 is the process of training the detection model, that is: performing iterative training on a preset dimensionality reduction model and a self-adaptive clustering model by using the obtained running indexes of a plurality of network nodes in a historical time range, judging whether the training result meets an iteration stop condition or not aiming at the training result of each iteration, and if so, determining dimensionality reduction core parameters included in the training result as dimensionality reduction core parameters of the preset dimensionality reduction model; and combining a preset dimension reduction model with the dimension reduction core parameters and the self-adaptive clustering model to obtain the detection model. The dimension reduction core parameter is K, for example.
Further, the determining whether the training result satisfies an iteration stop condition includes: comparing each first model evaluation value included in the training result with a second model evaluation value obtained by training the acquired operation index of the network node in the historical time range through a neural network model; and determining that the training result meets an iteration stop condition under the condition that the comparison result is that the difference value between each first model evaluation value and each second model evaluation value meets a preset error range.
Specifically, the first model evaluation value, or the second model evaluation value, may be an F1-score, where the F1-score represents, for example, a harmonic mean of precision (precision) and recall (recall) of the model, and the calculation method may be as shown in formula (1): f1 represents F1-score.
Figure BDA0003854087900000171
Wherein precision = TP/(TP + FP); recall = TP/(TP + FN); TP (True Positive) represents the prediction of a Positive class as a Positive class number; FN (False Negative) represents predicting a positive class as a Negative class number; FP (False Positive) represents the prediction of a negative class as a Positive class number; TN (True Negative) represents the prediction of a Negative class as a Negative class number.
It can be understood that, firstly, a second model evaluation value obtained by training the operation index of the acquired network node in the historical time range is used as a comparison basic value of a first model evaluation value of a detection model in each iteration training; and under the condition that the difference value between the second model evaluation value and the first model evaluation value is within a preset error range (for example, an arbitrary value within a range of 0 to 1), determining that the training result meets an iteration stop condition, and determining the detection model after the iteration as the trained detection model.
Further, the iterative training detection model comprises a preset dimension reduction model and a self-adaptive clustering model; specifically, historical operation index data is used as training data, iterative training is carried out on a preset dimensionality reduction model and a self-adaptive clustering model, and the method for one-time iterative training comprises the following steps: firstly inputting training data into a preset dimension reduction model to obtain a feature index after dimension reduction, then inputting the feature index after dimension reduction into an adaptive clustering model, calculating a first model evaluation value through the output of the adaptive clustering model, and comparing the first model evaluation value with a second model evaluation value to determine whether an iteration stop condition is met. The description of the process is the following steps N1 to N3, namely, the iterative training of the preset dimensionality reduction model and the adaptive clustering model is performed by using the obtained operation indexes of the plurality of network nodes in the historical time range, and the iterative training comprises the following steps: circularly executing the following steps from N1 to N3 until an iteration stop condition is met; n1: performing dimensionality reduction operation on the operation index in the historical time range by using the preset dimensionality reduction model and the dimensionality reduction core parameter of the current cycle period to obtain a characteristic index for training of the current cycle period; n2: inputting the characteristic indexes for training of the current cycle period into the self-adaptive clustering model, and determining a first model evaluation value according to the output of the self-adaptive clustering model; n3: and under the condition that the iteration stopping condition is not met, adjusting the dimensionality reduction core parameter of the current cycle period, taking the adjusted dimensionality reduction core parameter as the dimensionality reduction core parameter of the next cycle period, and executing the step N1.
Further, training for the adaptive clustering model includes training an iterative computation model included in the adaptive clustering model (e.g., an L-BFGS model); specifically, the method for training the adaptive clustering model comprises the following steps: the adaptive clustering model includes an iterative computation model that processes features; for each iterative training, further comprising: and utilizing the iterative computation model to adaptively compute the clustering radius of the characteristic data output by the preset dimensionality reduction model so as to output a clustering result based on the clustering radius, and enabling the clustering result to be optimal after iteration is stopped. The optimization of the adaptive clustering model is consistent with the description of step S204, and is not described herein again; it is to be understood that training the adaptive clustering model includes training all of the components comprised by the adaptive clustering model.
As shown in fig. 4, an apparatus 400 for managing a network node according to an embodiment of the present invention includes: a module 401 for obtaining node operation indexes, a module 402 for obtaining node operation numerical values, and a module 403 for determining node operation conditions; wherein the content of the first and second substances,
the node operation index obtaining module 401 is configured to obtain a first operation index of a network node and a second operation index of a plurality of other network nodes associated with the network node, where the first operation index and the second operation index both include multidimensional indexes;
the node operation value obtaining module 402 is configured to determine a first operation value of the network node according to a multidimensional index included in the first operation index; determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index;
the node operation condition determining module 403 is configured to determine an operation condition of the network node based on the first operation value and the second operation value, so as to manage the network node based on the operation condition of the network node.
As shown in fig. 5, an embodiment of the present invention provides a system 500 for managing a network node, including: a plurality of network nodes, at least one of which has said means 400 for managing network nodes; at least any two of the plurality of network nodes are communicatively coupled.
It is understood that the means for managing the network node may belong to the network node device to be detected, or to one or more other devices.
An embodiment of the present invention further provides an electronic device for managing a network node, including: one or more processors; the storage device is used for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors are enabled to realize the method provided by any one of the above embodiments.
Embodiments of the present invention further provide a computer-readable medium, on which a computer program is stored, where the computer program is executed by a processor to implement the method provided in any of the above embodiments.
Fig. 6 illustrates an exemplary system architecture 600 of a method of managing a network node or an apparatus for managing a network node to which embodiments of the present invention may be applied.
As shown in fig. 6, the system architecture 600 may include terminal devices 601, 602, 603, a network 604, and a server 605. The network 604 serves to provide a medium for communication links between the terminal devices 601, 602, 603 and the server 605. Network 604 may include various types of connections, such as wire, wireless communication links, or fiber optic cables, to name a few.
A user may use the terminal devices 601, 602, 603 to interact with the server 605 via the network 604 to receive or send messages or the like. Various client applications, such as an e-mall client application, a web browser application, a search application, an instant messaging tool, a mailbox client, etc., may be installed on the terminal devices 601, 602, 603.
The terminal devices 601, 602, 603 may be a variety of electronic devices having a display screen and supporting a variety of client applications, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.
The server 605 may be a server providing a service, e.g. a background management server providing support for client applications used by users with the terminal devices 601, 602, 603. The background management server can process the received request for obtaining the running condition of the network node and feed back the running condition of the network node corresponding to the request to the terminal equipment.
It should be noted that the method for managing network nodes provided by the embodiment of the present invention is generally executed by the server 605, and accordingly, the apparatus for managing network nodes is generally disposed in the server 605.
It should be understood that the number of terminal devices, networks, and servers in fig. 6 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for an implementation.
Referring now to FIG. 7, shown is a block diagram of a computer system 700 suitable for use with a terminal device implementing embodiments of the present invention. The terminal device shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.
As shown in fig. 7, the computer system 700 includes a Central Processing Unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data necessary for the operation of the system 700 are also stored. The CPU 701, ROM 702, and RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
The following components are connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.
In particular, according to embodiments of the present disclosure, the processes described above with reference to the flow diagrams may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709, and/or installed from the removable medium 711. The computer program performs the above-described functions defined in the system of the present invention when executed by the Central Processing Unit (CPU) 701.
It should be noted that the computer readable medium shown in the present invention can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present invention, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the present invention, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules and/or units described in the embodiments of the present invention may be implemented by software, and may also be implemented by hardware. The described modules and/or units may also be provided in a processor, and may be described as: a processor comprises a node operation index acquisition module, a node operation value acquisition module and a node operation condition determination module. The name of these modules does not in some cases form a limitation on the module itself, for example, the obtain node operation index module may also be described as a "module for obtaining a first operation index of a network node and second operation indexes of a plurality of other network nodes associated with the network node".
As another aspect, the present invention also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise: respectively obtaining a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index both comprise multidimensional indexes; determining a first operation numerical value of the network node according to the multidimensional index included by the first operation index; determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node.
According to the embodiment of the invention, the first operation numerical value of the network node can be determined according to the first operation index of the network node; and determining a second operational value of the network node based on the first operational indicator and a second operational indicator associated with the network node; determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node. The universality and flexibility of the management network node are improved, and the accuracy of judging the running condition of the network node is improved; the automation degree of the management network node is improved to a great extent, and the consumption of labor cost and time cost is reduced.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (15)

1. A method of managing a network node, comprising:
respectively obtaining a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index comprise multidimensional indexes;
determining a first operation numerical value of the network node according to the multidimensional index included by the first operation index;
determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index;
determining an operational condition of the network node based on the first operational value and the second operational value to manage the network node based on the operational condition of the network node.
2. The method of managing network nodes of claim 1,
the determining a first running value of the network node comprises:
inputting the multidimensional indexes included by the first operation indexes into a neural network model, and outputting first operation numerical values of the network nodes by using the neural network model; wherein the neural network model comprises a long-short term memory network and a full convolution network; and integrating the output of the long-short term memory network aiming at the first operation index and the output of the full convolution network aiming at the first operation index to obtain the first operation numerical value.
3. The method of managing network nodes of claim 1,
the determining a second running value of the network node comprises:
inputting a detection model for the multidimensional indexes included by the first operation indexes of the network nodes and the multidimensional indexes included by the second operation indexes of the other network nodes, and outputting second operation values of the network nodes relative to the other nodes by using the detection model, wherein the detection model comprises a preset dimensionality reduction model and an adaptive clustering model.
4. The method of claim 3, wherein said outputting, using the detection model, a second operational value of the network node relative to the plurality of other nodes comprises:
performing dimensionality reduction operation on the multidimensional indexes included by the first operation indexes and the multidimensional indexes included by the second operation indexes by using the preset dimensionality reduction model to obtain first characteristic indexes of the network nodes and second characteristic indexes corresponding to the other network nodes;
and inputting the first characteristic index and a plurality of second characteristic indexes into the self-adaptive clustering model, and calculating a second operation numerical value of the network node.
5. The method of claim 3, further comprising:
performing iterative training on a preset dimensionality reduction model and a self-adaptive clustering model by using the obtained running indexes of a plurality of network nodes in a historical time range, judging whether the training result meets an iteration stop condition or not aiming at the training result of each iteration, and if so, determining that dimensionality reduction core parameters included in the training result are dimensionality reduction core parameters of the preset dimensionality reduction model;
and combining a preset dimension reduction model with the dimension reduction core parameters and the self-adaptive clustering model to obtain the detection model.
6. The method of claim 5, wherein the determining whether the training result satisfies an iteration stop condition comprises:
comparing each first model evaluation value included in the training result with a second model evaluation value obtained by training the acquired operation index of the network node in the historical time range through a neural network model;
and determining that the training result meets an iteration stop condition under the condition that the comparison result is that the difference value between each first model evaluation value and each second model evaluation value meets a preset error range.
7. The method of claim 5,
the iterative training of the preset dimensionality reduction model and the self-adaptive clustering model by using the acquired running indexes of the plurality of network nodes in the historical time range comprises the following steps:
circularly executing the following steps from N1 to N3 until an iteration stop condition is met;
n1: performing dimensionality reduction operation on the operation index in the historical time range by using the preset dimensionality reduction model and the dimensionality reduction core parameter of the current cycle period to obtain a characteristic index for training of the current cycle period;
n2: inputting the characteristic indexes for training of the current cycle period into the self-adaptive clustering model, and determining a first model evaluation value according to the output of the self-adaptive clustering model;
n3: and under the condition that the iteration stop condition is not met, adjusting the dimension reduction core parameter of the current cycle period, taking the adjusted dimension reduction core parameter as the dimension reduction core parameter of the next cycle period, and executing the step N1.
8. The method of claim 5,
the adaptive clustering model includes an iterative computation model of processing features;
for each iterative training, further comprising:
and utilizing the iterative computation model to adaptively compute the clustering radius of the characteristic data output by the preset dimension reduction model so as to output a clustering result based on the clustering radius, and enabling the clustering result to be optimal after iteration is stopped.
9. The method of claim 2, further comprising:
iteratively performing the following operations to train the neural network model:
respectively inputting the obtained operation indexes of the plurality of network nodes in a historical time range into the long-short term memory network and the full convolution network, and combining the output of the long-short term memory network aiming at the operation indexes in the historical time range with the output of the full convolution network aiming at the operation indexes in the historical time range to obtain the output of the neural network model;
evaluating a training result of the neural network model by using a preset loss function and the output of the neural network model;
and adjusting the neural network model according to the evaluation result.
10. The method of claim 9, further comprising:
acquiring characteristic data of the operation indexes in the historical time range, inputting the characteristic data into an attention mechanism model, and increasing the weight value of the characteristic data close to the current time by using the attention mechanism model;
the step of inputting the acquired operation indexes of the plurality of network nodes in the historical time range into the long-short term memory network comprises the following steps:
and inputting the characteristic data processed by the attention mechanism model into the long-short term memory network to train the long-short term memory network.
11. The method of claim 1, further comprising:
obtaining the first operation index of the network node;
inputting the first operation index into a time sequence prediction model, predicting the operation condition of the network node in a future time range by using the time sequence prediction model, wherein the time sequence prediction model is obtained by performing periodic characteristic training based on the operation index in a historical time range and performing periodic adjustment on a prediction result predicted by the time sequence prediction model.
12. An apparatus for managing a network node, comprising: the node operation index acquisition module, the node operation numerical value acquisition module and the node operation condition determination module are used for acquiring node operation indexes; wherein the content of the first and second substances,
the node operation index acquisition module is used for respectively acquiring a first operation index of a network node and second operation indexes of a plurality of other network nodes related to the network node, wherein the first operation index and the second operation index both comprise multidimensional indexes;
the node operation value acquisition module is used for determining a first operation value of the network node according to the multidimensional indexes included in the first operation indexes; determining a second operation numerical value of the network node according to the multidimensional index included by the first operation index and the multidimensional index included by the second operation index;
the node operation condition determining module is configured to determine an operation condition of the network node based on the first operation value and the second operation value, so as to manage the network node based on the operation condition of the network node.
13. A system for managing network nodes, comprising: a plurality of network nodes, at least one network node having the apparatus for managing network nodes of claim 11; at least any two network nodes of the plurality of network nodes are communicatively coupled.
14. An electronic device, comprising:
one or more processors;
a storage device for storing one or more programs,
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method recited in any of claims 1-11.
15. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-11.
CN202211142185.0A 2022-09-20 2022-09-20 Method, device and system for managing network nodes Pending CN115567406A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211142185.0A CN115567406A (en) 2022-09-20 2022-09-20 Method, device and system for managing network nodes

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211142185.0A CN115567406A (en) 2022-09-20 2022-09-20 Method, device and system for managing network nodes

Publications (1)

Publication Number Publication Date
CN115567406A true CN115567406A (en) 2023-01-03

Family

ID=84741726

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211142185.0A Pending CN115567406A (en) 2022-09-20 2022-09-20 Method, device and system for managing network nodes

Country Status (1)

Country Link
CN (1) CN115567406A (en)

Similar Documents

Publication Publication Date Title
CN109471783B (en) Method and device for predicting task operation parameters
CN110929799B (en) Method, electronic device, and computer-readable medium for detecting abnormal user
CN114138625A (en) Method and system for evaluating health state of server, electronic device and storage medium
CN113837596B (en) Fault determination method and device, electronic equipment and storage medium
CN107392259B (en) Method and device for constructing unbalanced sample classification model
CN110766184A (en) Order quantity prediction method and device
WO2023134188A1 (en) Index determination method and apparatus, and electronic device and computer-readable medium
US11645540B2 (en) Deep graph de-noise by differentiable ranking
CN115034596A (en) Risk conduction prediction method, device, equipment and medium
CN114500339A (en) Node bandwidth monitoring method and device, electronic equipment and storage medium
CN114780338A (en) Host information processing method and device, electronic equipment and computer readable medium
CN107291835B (en) Search term recommendation method and device
CN116684330A (en) Traffic prediction method, device, equipment and storage medium based on artificial intelligence
CN112508723B (en) Financial risk prediction method and device based on automatic preferential modeling and electronic equipment
CN112486784A (en) Method, apparatus and medium for diagnosing and optimizing data analysis system
CN112231299A (en) Method and device for dynamically adjusting feature library
CN116128135A (en) Data processing method and device, electronic equipment and storage medium
CN116186603A (en) Abnormal user identification method and device, computer storage medium and electronic equipment
CN115567406A (en) Method, device and system for managing network nodes
CN114637809A (en) Method, device, electronic equipment and medium for dynamic configuration of synchronous delay time
CN111429257B (en) Transaction monitoring method and device
CN110413482B (en) Detection method and device
CN112906723A (en) Feature selection method and device
CN115550259B (en) Flow distribution method based on white list and related equipment
CN113052509B (en) Model evaluation method, model evaluation device, electronic apparatus, and storage medium

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