CN109088754A - A kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method - Google Patents
A kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method Download PDFInfo
- Publication number
- CN109088754A CN109088754A CN201810856699.XA CN201810856699A CN109088754A CN 109088754 A CN109088754 A CN 109088754A CN 201810856699 A CN201810856699 A CN 201810856699A CN 109088754 A CN109088754 A CN 109088754A
- Authority
- CN
- China
- Prior art keywords
- communication network
- learning machine
- failure
- characteristic parameter
- network performances
- 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
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
-
- 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)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Algebra (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method, communication network failure diagnostic techniques fields.The method obtains the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed, using the learning machine COMMUNICATION NETWORK PERFORMANCES fault analysis model that transfinites pre-established, the characteristic parameter of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed is handled, to analyze the corresponding COMMUNICATION NETWORK PERFORMANCES failure mode of this characteristic parameter.The present invention realizes COMMUNICATION NETWORK PERFORMANCES failure mode and effectively accurately analyzes, and by choosing sigmoid activation primitive and Gaussian radial basis function both different activation primitives, enhancing tradition transfinites the generalization ability of learning machine model, improves model accuracy.
Description
Technical field
The invention belongs to communication network failure diagnostic techniques field, it is related to a kind of COMMUNICATION NETWORK PERFORMANCES failure reason and feature
Parameter association relationship analysis method more particularly to a kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter based on improvement ELM
Association relationship analysis method.
Background technique
The influence factor of communication network failure is numerous, and multifactor complicated coupling, dynamic time-varying, carry out failure reason with
When characteristic parameter incidence relation is analyzed, needs to consider configuration, the coupling of whole network system, show complicated association between failure
Relationship: a failure often causes other multiple failures, these failure dynamic dependencies, constantly association triggering.In engineering, in face of same
Multiple failures of Shi Fasheng, in the case where its incidence relation is unknown, engineering staff can only check one by one, troubleshooting inefficiency.
Currently, mainly passing through expert practical experience in analysis of communications networks performance fault reason.In addition, tradition research
Middle analyzing and associating relationship, in addition to expertise, there are also the methods of the reasonings of rule-based summary, the reasoning based on example.However,
Network fault influence factor is numerous, and influence of the different factors for performance fault is different, and this relationship is extremely complex, with failure
Scale increases, and expertise is difficult quickly and accurately analyzing and associating relationship.Recently as the development of machine learning techniques, it is
We provide new COMMUNICATION NETWORK PERFORMANCES fault methods.Wherein transfinite learning machine (Extreme Learning Machine)
(bibliography [1]: Huang G B, Wang D H, Lan Y.Extreme learning machines:a survey [J]
.International Journal of Machine Learning&Cybernetics.2011,2 (2): 107-122. is translated:
Bibliography [1]: Huang Guangbin, Wang Dianhui, Lan Y. transfinite learning machine: investigation [J] machine learning and cybernetics international phase
Print .2011,2 (2): 107-122.) it is more model in recent years, its essence is completing to be input to output mapping,
But traditional learning machine that transfinites is a kind of feedforward neural networks with single hidden layer model, and due to its design feature, generalization ability is poor
(generalization ability refers to the adaptability to new samples), and model accuracy is to be improved.
Summary of the invention
The present invention is the problem analysis in order to solve COMMUNICATION NETWORK PERFORMANCES failure reason.In COMMUNICATION NETWORK PERFORMANCES failure reason
When analyzing with characteristic parameter incidence relation, traditional learning machine model that transfinites, and model accuracy poor there are generalization ability has
The problem of to be improved, the present invention, aiming to the above problems, provides a kind of COMMUNICATION NETWORK PERFORMANCES failure reasons, and pass is associated with characteristic parameter
It is analysis method, the learning machine model that transfinites of specially a kind of heterogeneous hidden node.By choosing sigmoid activation primitive and height
Both different activation primitives of this radial basis function enhance the extensive energy of the learning machine model that transfinites of heterogeneous hidden node
Power improves model accuracy.
A kind of COMMUNICATION NETWORK PERFORMANCES failure reason provided by the invention and characteristic parameter association relationship analysis method, feature
Be the described method includes:
S1: the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed are obtained.
S2: using the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites pre-established, to be analyzed is led to
The characteristic parameter of communication network performance fault is handled, and is caused to analyze the corresponding COMMUNICATION NETWORK PERFORMANCES failure of this characteristic parameter
Cause.
Wherein, the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites is based on heterogeneous hidden node
It transfinites made of learning machine model training, training process are as follows:
S21: the training sample of the learning machine model that transfinites of heterogeneous hidden node is generated using Network Simulation Software (such as OPNET)
This (or the data for using actual measurement to obtain as training sample);
S22: using the learning machine model that transfinites of heterogeneous hidden node, being trained the training sample, obtains described
The learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites.
Wherein, the activation primitive of the learning machine model use of transfiniting of heterogeneous hidden node are as follows:WhereinFor sigmoid activation primitive,For gaussian radial basis function activation primitive, μ is the center of gaussian radial basis function activation primitive, and σ is Gauss diameter
To the width of base activation primitive, as shown in Figure 1, it is constant, i and j difference that a and b, which are the parameter obtained after transfiniting learning machine training,
Indicate the serial number of hidden node, x is the sample for inputting parameter, and e is that natural logrithm is constant.
Wherein, the learning machine model that transfinites that heterogeneous hidden node is generated using Network Simulation Software (such as OPNET)
Training sample process specifically:
S211: it determines the base unit information of network, determines network topology structure;
S212: control variable determines influence factor, such as network topology structure number of nodes, router memory;
S213: the characteristic parameter of the performance fault to be collected of determination, including packet loss, time delay, the bit error rate;
S214: setting simulation time runs simulated program;
S215: the training sample of generation is exported.
Wherein, the process training sample being trained described in step S22 specifically:
S221: step S21 training sample obtained is randomly divided into training set and verifying collects, training set accounts for p (0.6≤a
< 1, and a is rational), verifying collection accounts for 1-p;
S222: the learning machine model that transfinites that training set obtained by step S221 is input to heterogeneous hidden node is trained
To the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites;
S223: verifying collection obtained by step S221 is input to the learning machine communication network that transfinites that step S222 training obtains
Can failure causation analysis model, the analysis result that learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model obtains of transfiniting and true
Real result compares, and verifying gained transfinites the validity of learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model and accurate
Property.
A kind of COMMUNICATION NETWORK PERFORMANCES failure reason provided by the invention and characteristic parameter association relationship analysis method, such as Fig. 1
Shown in traditional learning machine that transfinites for showing relative to Fig. 2 of the learning machine model that transfinites based on the theoretical heterogeneous hidden node of suit
Model is compared, and advantage essentially consists in two o'clock:
(1) the learning machine model that transfinites of heterogeneous hidden node, which uses, mixes heterogeneous hidden node, i.e., hidden node is sharp
Function type and quantity living are all different.According to Adaboost theory, when two kinds of activation primitives and corresponding input, output
Weight matrix can be combined into a Weak Classifier, if the Weak Classifier that both single hidden nodes are constituted all has lower mistake
Rate, then mixing these heterogeneous nodes can be improved the precision of model.The selected two kinds of activation primitive sigmoid activation of the present invention
Function and gaussian radial basis function activation primitive can meet above-mentioned condition, and therefore, the improved learning machine model that transfinites is able to ascend
The precision of model.
(2) the learning machine model that transfinites of heterogeneous hidden node improves Weak Classifier generalization ability.So-called generalization ability is
Refer to the adaptability to new samples, traditional learning machine model error that transfinites is mainly since generalization ability is poor, current research
In transfinite learning machine model due to the design feature of its Single hidden layer feedforward neural networks, although its structure is simple, the speed of service is fast,
Then its generalization ability has certain limitation.Adaboost theory machine learning field be proved to be able to improve it is extensive
A kind of method of ability, core are that multiple " Weak Classifiers " are combined into " strong classifier ".We are theoretical by Bagging,
Evaluation function sequence is generated, generalization ability is improved using voting mechanism.
Detailed description of the invention
Fig. 1 is the ELM model operation principle schematic diagram of the theoretical heterogeneous hidden node of suit.
Fig. 2 is traditional ELM model operation principle schematic diagram.
Fig. 3 is the learning machine model training process that transfinites of heterogeneous hidden node.
Specific embodiment
Below in conjunction with attached drawing and preferred embodiment, the invention will be further described.It is emphasized that following the description
It is only exemplary, the range and application being not intended to be limiting of the invention.
Embodiment 1
Further to verify model efficiency and validity, the fault data of emulation is chosen as experimental data, is saved as inside
4GB, processor is Interl (R) Core (TM) i5-2430M CPU 2.40GHz 2.40GHz, operation cabinet is 64
ELM model and integrated ELM of the invention are realized by Python2.7+Anaconda6.0 programming on the computer of Window10
Model.
S1: the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed are obtained.
In order to obtain the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed, we are carried out using OPNET herein
Emulation, the emulation data of the characteristic parameter of the correlated performance failure of collection, characteristic parameter include packet loss, time delay, the bit error rate, are incited somebody to action
Emulation data are encapsulated as four-dimensional variable (LABEL, TD, PLR, BER), and wherein LABEL is label, and three classes reason TD, PLR, BER will
As input variable, our every kind of failure reason sample data randomly selects 5000 here, three classes reason totally 15000 samples
This.The sample that will acquire is separated into training set and test set at random, and training set accounts for 75%, and verifying collection accounts for 25%.Here we use
Train_test_split module in Python platform sklearn.cross_valiation for dividing data, then from
StandardScaler is imported in Python platform inside sklearn.preprocessing and is used for standardized data, is guaranteed every
The characteristic variance of a latitude is 1, mean value 0, so that prediction result will not be dominated by the excessive characteristic value of certain latitudes.
Here it is the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed that we, which take verifying collection,.
S2: using the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites pre-established to communication to be analyzed
The characteristic parameter of network performance failure is handled, and is caused to analyze the corresponding COMMUNICATION NETWORK PERFORMANCES failure of this characteristic parameter
Cause.
As Fig. 3 obtains step S1 in order to which training obtains the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites
The training set of statistical nature parameter be input in the learning machine model that transfinites of heterogeneous hidden node, obtained by training iteration
The output weight matrix of the learning machine model that transfinites of heterogeneous hidden node, to obtain learning machine COMMUNICATION NETWORK PERFORMANCES event of transfiniting
Hinder causation analysis model.We are herein with Sigmoid function
With Gaussian radial basis functionAs the activation primitive for the hidden node that case uses, μ
It is the center of gaussian radial basis function activation primitive, σ is the width of gaussian radial basis function activation primitive, and a and b are after transfiniting learning machine training
Obtained parameter is constant, and i and j respectively indicate the serial number of hidden node, and x is the sample for inputting parameter, and e is that natural logrithm is.
Using Bagging-ELM integrated classifier, output category result of voting.
It can be by test set for the verifying for the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model accuracy that transfinites
Test result obtains.There is the extraction Bagging sample put back at random from test set, the input training of Bagging sample is obtained
ELM model in, computation model obtains the ratio of prediction result and actual result, the ratio be precision.By comparing traditional ELM
Model and the precision for improving ELM model, it may be verified that the precision of improved model.
Verifying for generalization ability is to be mixed under same sample data by the number and change of integrated study device T
The ratio of node is closed, to compare two kinds using traditional ELM model and improve and pass through control in obtained model under ELM model case
The number of T realizes the analysis of integrated result, controls the ratio of two kinds of activation primitive nodes, if α=1, generation by using α value
Table hidden layer node is homogeneity node, i.e., only a kind of activation primitive sigmoid constitutes hidden node;If 0 < α < 1, represents
The ratio of two kinds of mixed nodes, in order to verify generalization ability, T value is respectively 1 and 20,0 < α < 1=0.5 and α=1 by we
As the comparison for improving ELM model and conventional model.α=0.5 is taken, hidden layer node, T=20, setting base classifier are set
Number.The sample data of training set is input to training in model and obtains output matrix β;By the data of test set to trained model
Accuracy validation illustrates that model is effective if accuracy can reach 94% or more, typically, what test data obtained
Nicety of grading, which reaches 94%, can think that model is effective.α=1, T=1 are compared, improved model and traditional ELM are carried out
Comparison inputs the training set sample of same ratio, comparison and precision when α=0.5, T=20, if α=0.5, T=20 are showed
Precision it is higher, then illustrate generalization ability promotion be proven.
With split () method of Scipython, choosing 70% is training data, 30% verify data.Verifying is completed,
The results are shown in Table 1 for analysis.
Model | Precision |
ELM | 0.9343 |
Improve ELM | 0.9448 |
Table 1
By emulating data, it has been found that its ratio of precision tradition ELM model of improved ELM model promotes 1% or more, card
Its bright precision is higher than tradition ELM model, and precision is more than 94%, it was demonstrated that its generalization ability is strong, it was demonstrated that its method is effective.
Claims (5)
1. a kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method, it is characterised in that: the method
Include:
S1: the characteristic parameter data of COMMUNICATION NETWORK PERFORMANCES failure to be analyzed are obtained;
S2: using the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites pre-established, to communication network to be analyzed
The characteristic parameter of network performance fault is handled, to analyze the corresponding COMMUNICATION NETWORK PERFORMANCES failure reason of this characteristic parameter;
Wherein, the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites is transfiniting based on heterogeneous hidden node
Made of learning machine model training.
2. a kind of COMMUNICATION NETWORK PERFORMANCES failure reason according to claim 1 and characteristic parameter association relationship analysis method,
It is characterized by: the learning machine model training that transfinites based on heterogeneous hidden node, process are specific as follows:
S21: the training sample of the learning machine model that transfinites of heterogeneous hidden node is generated using Network Simulation Software, or using practical
Obtained data are measured as training sample;
S22: using the learning machine model that transfinites of heterogeneous hidden node, being trained the training sample, obtains described surpass
Limit learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model.
3. a kind of COMMUNICATION NETWORK PERFORMANCES failure reason according to claim 2 and characteristic parameter association relationship analysis method,
It is characterized by: wherein, the activation primitive of the learning machine model use of transfiniting of heterogeneous hidden node in step S21 and S22 are as follows:WhereinFor sigmoid activation primitive,For gaussian radial basis function activation primitive, μ is the center of gaussian radial basis function activation primitive, and σ is Gauss diameter
To the width of base activation primitive, a and b are that the parameter obtained after transfiniting learning machine training is constant, and i and j respectively indicate hidden layer section
The serial number of point, x are the sample for inputting parameter, and e is that natural logrithm is constant.
4. a kind of COMMUNICATION NETWORK PERFORMANCES failure reason according to claim 2 and characteristic parameter association relationship analysis method,
It is characterized by: the instruction of the learning machine model that transfinites of heterogeneous hidden node is generated described in step S21 using Network Simulation Software
Practice the process of sample specifically:
S211: it determines the base unit information of network, determines network topology structure;
S212: control variable determines influence factor;
S213: the characteristic parameter of the performance fault to be collected of determination, including packet loss, time delay, the bit error rate;
S214: setting simulation time runs simulated program;
S215: the training sample of generation is exported.
5. a kind of COMMUNICATION NETWORK PERFORMANCES failure reason according to claim 2 and characteristic parameter association relationship analysis method,
It is characterized by: the process being trained described in step S22 to the training sample specifically:
S221: being randomly divided into training set for step S21 training sample obtained and verifying collect, and training set accounts for p, and verifying collection accounts for 1-
p;Wherein, p is 0.6≤a < 1, and a is rational;
S222: the learning machine model that transfinites that training set obtained by step S221 is input to heterogeneous hidden node is trained and is surpassed
Limit learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model;
S223: verifying collection obtained by step S221 is input to the learning machine COMMUNICATION NETWORK PERFORMANCES event of transfiniting that step S222 training obtains
Hinder causation analysis model, by the analysis result and true knot that learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model obtains that transfinite
Fruit compares, the validity and accuracy of the learning machine COMMUNICATION NETWORK PERFORMANCES failure causation analysis model that transfinites obtained by verifying.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810856699.XA CN109088754A (en) | 2018-07-31 | 2018-07-31 | A kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810856699.XA CN109088754A (en) | 2018-07-31 | 2018-07-31 | A kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109088754A true CN109088754A (en) | 2018-12-25 |
Family
ID=64831043
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810856699.XA Pending CN109088754A (en) | 2018-07-31 | 2018-07-31 | A kind of COMMUNICATION NETWORK PERFORMANCES failure reason and characteristic parameter association relationship analysis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109088754A (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102394765A (en) * | 2011-09-01 | 2012-03-28 | 上海大学 | Method for positioning faults of cable television network management system |
CN104536435A (en) * | 2014-12-18 | 2015-04-22 | 中国科学院电工研究所 | Online diagnosis method for linear control system network |
CN106130786A (en) * | 2016-07-26 | 2016-11-16 | 腾讯科技(深圳)有限公司 | The detection method of a kind of network failure and device |
CN106707221A (en) * | 2017-01-05 | 2017-05-24 | 云南电网有限责任公司电力科学研究院 | Fault diagnosis method and system for sensor in electric energy metering device |
CN107018030A (en) * | 2017-06-01 | 2017-08-04 | 厦门华厦学院 | Mobile network's trouble analysis system based on intelligent terminal |
CN107370617A (en) * | 2017-03-10 | 2017-11-21 | 南京航空航天大学 | Cellular network fault diagnosis system based on SVM |
US20170351227A1 (en) * | 2016-06-01 | 2017-12-07 | Accenture Global Solutions Limited | Real-time engineering analysis based system control |
-
2018
- 2018-07-31 CN CN201810856699.XA patent/CN109088754A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102394765A (en) * | 2011-09-01 | 2012-03-28 | 上海大学 | Method for positioning faults of cable television network management system |
CN104536435A (en) * | 2014-12-18 | 2015-04-22 | 中国科学院电工研究所 | Online diagnosis method for linear control system network |
US20170351227A1 (en) * | 2016-06-01 | 2017-12-07 | Accenture Global Solutions Limited | Real-time engineering analysis based system control |
CN106130786A (en) * | 2016-07-26 | 2016-11-16 | 腾讯科技(深圳)有限公司 | The detection method of a kind of network failure and device |
CN106707221A (en) * | 2017-01-05 | 2017-05-24 | 云南电网有限责任公司电力科学研究院 | Fault diagnosis method and system for sensor in electric energy metering device |
CN107370617A (en) * | 2017-03-10 | 2017-11-21 | 南京航空航天大学 | Cellular network fault diagnosis system based on SVM |
CN107018030A (en) * | 2017-06-01 | 2017-08-04 | 厦门华厦学院 | Mobile network's trouble analysis system based on intelligent terminal |
Non-Patent Citations (1)
Title |
---|
郑小禄等: "《基于超限学习机的通信网络弹性预测方法》", 《通信技术》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106529090B (en) | A kind of aerospace electron class Reliability Assessment method | |
CN105893256B (en) | software fault positioning method based on machine learning algorithm | |
Cai et al. | Advice classes of parameterized tractability | |
CN106483449B (en) | Analog-circuit fault diagnosis method based on deep learning and Complex eigenvalues | |
CN107909118A (en) | A kind of power distribution network operating mode recording sorting technique based on deep neural network | |
CN113792825A (en) | Fault classification model training method and device for electricity information acquisition equipment | |
CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
CN108520301A (en) | A kind of circuit intermittent fault diagnostic method based on depth confidence network | |
US20180100894A1 (en) | Automatic Generation of Test Sequences | |
CN108985279A (en) | The method for diagnosing faults and device of double-unit traction controller waveform | |
EP1290459A2 (en) | Method and apparatus for maximizing test coverage | |
CN109298993A (en) | A kind of method, apparatus and computer readable storage medium detecting failure | |
CN105572572B (en) | Analog-circuit fault diagnosis method based on WKNN-LSSVM | |
CN102707256A (en) | Fault diagnosis method based on BP-Ada Boost nerve network for electric energy meter | |
CN105606914A (en) | IWO-ELM-based Aviation power converter fault diagnosis method | |
CN108305619A (en) | Voice data collection training method and apparatus | |
CN112464995A (en) | Power grid distribution transformer fault diagnosis method and system based on decision tree algorithm | |
CN109033513A (en) | Method for diagnosing fault of power transformer and diagnosing fault of power transformer device | |
CN109375948A (en) | A kind of software pricing method of intelligent function point identification | |
CN101871994A (en) | Method for diagnosing faults of analog circuit of multi-fractional order information fusion | |
CN107391452A (en) | A kind of software defect estimated number method based on data lack sampling and integrated study | |
CN104635146B (en) | Analog circuit fault diagnosis method based on random sinusoidal signal test and HMM (Hidden Markov Model) | |
CN117056678B (en) | Machine pump equipment operation fault diagnosis method and device based on small sample | |
CN112926686B (en) | BRB and LSTM model-based power consumption anomaly detection method and device for big power data | |
CN114330486A (en) | Power system bad data identification method based on improved Wasserstein GAN |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181225 |