CN110139315A - A kind of wireless network fault detection method based on self-teaching - Google Patents

A kind of wireless network fault detection method based on self-teaching Download PDF

Info

Publication number
CN110139315A
CN110139315A CN201910343031.XA CN201910343031A CN110139315A CN 110139315 A CN110139315 A CN 110139315A CN 201910343031 A CN201910343031 A CN 201910343031A CN 110139315 A CN110139315 A CN 110139315A
Authority
CN
China
Prior art keywords
fault detection
label
self
record
kpi
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.)
Granted
Application number
CN201910343031.XA
Other languages
Chinese (zh)
Other versions
CN110139315B (en
Inventor
潘志文
陈彦
尤肖虎
刘楠
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.)
Southeast University
Original Assignee
Southeast University
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 Southeast University filed Critical Southeast University
Priority to CN201910343031.XA priority Critical patent/CN110139315B/en
Publication of CN110139315A publication Critical patent/CN110139315A/en
Application granted granted Critical
Publication of CN110139315B publication Critical patent/CN110139315B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/06Management of faults, events, alarms or notifications
    • H04L41/0677Localisation of faults
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • H04L43/06Generation of reports
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • H04L43/08Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W24/00Supervisory, monitoring or testing arrangements
    • H04W24/08Testing, supervising or monitoring using real traffic
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W24/00Supervisory, monitoring or testing arrangements
    • H04W24/10Scheduling measurement reports ; Arrangements for measurement reports

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Environmental & Geological Engineering (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a kind of wireless network fault detection method based on self-teaching, specific as follows: acquisition network key performance indicator KPI is recorded and is saved as data acquisition system;Uneven processing is carried out using SMOTEENN method to the data acquisition system, the equilibrium data collection after being balanced;By sparse self-encoding encoder from no label auxiliary data focusing study base vector;It is the form of each base vector linear combination by the equilibrium data set representations after balance, and obtains the normally disaggregated model with fault category through the training of support vector machines method on the equilibrium data collection under new expression herein;Classified using well-established disaggregated model to the KPI data record that network generates in real time, and then achievees the purpose that fault detection.The present invention is more accurate and network failure is effectively detected;And the form of self-teaching facilitates migration, can obtain the model of fault detection quickly under new network environment, improve the fault detection efficiency of previous methods.

Description

A kind of wireless network fault detection method based on self-teaching
Technical field
The present invention relates to the network technique field in wireless communication, especially a kind of wireless network event based on self-teaching Hinder detection method.
Background technique
It is one of the critical function of wireless self-organization network from healing, accurately and efficiently carrying out fault detection is to realize certainly Cure the important step of function.Changed in Current wireless communication network using macro base station shown in FIG. 1-Home eNodeB heterogeneous structure The in-door covering of kind macro base station provides capacity gain.Compared to traditional homogeneous network, the Home eNodeB of user's autonomous deployment may Since operating mistake is easier to that the allocation problems such as the improper, channel confliction of transimission power setting occur;In addition, the quantity of Home eNodeB is remote More than macro base station, distributed structure/architecture is but also family is easier to break down.Therefore it is badly in need of a kind of self-organizing for heterogeneous network Network fault detecting method.
The method of the current fault detection for homogeneous network is not often suitable for heterogeneous network, and reason essentially consists in together Structure Network Fault Detection is mostly based on mode classification, and the accuracy rate of classification depends on network history data quantity to a certain extent Size.And the available data of Home eNodeB are often limited in heterogeneous network.There are two the reason of leading to this, is on the one hand portion Administration's Home eNodeB indoors usually services that number of users is less, and it is few that obtainable data compare homogeneous network;On the other hand it is Home eNodeB deployment is determined that the operations such as switching, redeploy regularly causes network topology structure dynamically to change by user, The historical data of base station long-term accumulation cannot will directly be used to analyze the current state in the base station, i.e. network history data exists The low problem of timeliness.Femtocell user data rareness, historical data in heterogeneous network exist in the prior art in summary Easily out-of-date problem.
Summary of the invention
It is provided the technical problem to be solved by the present invention is to overcome the deficiencies in the prior art a kind of based on self-teaching Wireless network fault detection method, the fault detection method based on self-teaching that uses can be used in Home eNodeB in the present invention User data is few, generate the easy mistake of data in the case of, it is more accurate and network failure is effectively detected;And the form of self-teaching Facilitate migration, the model of fault detection can be obtained quickly under new network environment, improve the fault detection of previous methods Efficiency.
The present invention uses following technical scheme to solve above-mentioned technical problem:
A kind of wireless network fault detection method based on self-teaching proposed according to the present invention, comprising the following steps:
Step 1: acquisition network key performance indicator KPI is recorded and is saved as data acquisition system;
It combines synthesis minority class oversampling technique SMOTE Step 2: being used to the data acquisition system and screens arest neighbors ENN's SMOTEENN method carries out uneven processing, the equilibrium data collection after being balanced;
Step 3: by sparse self-encoding encoder from no label auxiliary data focusing study base vector;
Step 4: being the form of each base vector linear combination by the equilibrium data set representations after balance, and new expression herein Under equilibrium data collection on through support vector machines method training obtain the normally disaggregated model with fault category;
Step 5: being classified using well-established disaggregated model to the KPI data record that network generates in real time, in turn Achieve the purpose that fault detection.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Normally refer to non-faulting.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Step 1 is specific as follows:
In heterogeneous network, user periodically reports all kinds of KPI information in the form of measurement report;
From the KPI information for obtaining reporting of user in certain time in the heterogeneous network for needing to carry out fault detection, and according to The reporting of user moment is to the normal KPI information corresponding with malfunction label of its serving BS;KPI information and correspondence markings are protected Save as data acquisition systemForm, m be T in element total number, i.e., altogether acquisition M item record;I-th of element in TInRepresent the n dimension KPI information of certain moment reporting of user, subscript L indicates label data,I-th to report has label KPI information, RnFor n-dimensional vector space, i=1,2 ..., m;y(i)It is correspondingLabel have normal condition and two kinds of malfunction to report moment user service base station state in which.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Y when normal condition(i)=+1, y when malfunction(i)=-1.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Normal condition refers to non-faulting state.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Step 2 is specific as follows:
2.1, to every record that label in T is failureAccording to the KPI information in recordIt is chosen from T Comprising KPI information withApart from the nearest similar KPI record of k item, failure refers to y(i)=-1 separately takes a new data set T'=T;
2.2, r item is extracted from the KPI information of the similar record of k item with putting back to, r is taken as 1 to the integer value between k, often When extracting one this record include KPI withLine on randomly select and a little generate new failure classes KPI xl_new, Subscript l_new, which indicates new, label data;Enable xl_newCorresponding label be -1;
2.3,2.1,2.2 processes are repeated, are recorded in every failure classesIt is upper to generate the new failure classes record of r itemIt is incorporated in data set T', i.e. update T' is For the v articles new KPI of generation Information;Indicate the v articles new record generated;∪ expression takes its front and back union of sets collection;
2.4, it to every record in T', chooses and it is recorded apart from nearest k' item;
2.5, judge whether the label to occupy the majority in k' item record is consistent with the label of record to be judged, protects if consistent It stays this to record, this record is rejected if inconsistent;
2.6, final equilibrium data collection is thus obtainedWhat se was puts down The number of element in weighing apparatus data set;Q-th of element in T'Iny'(q)It is correspondingLabel, y'(q) Normal condition, y' are indicated when=+ 1(q)=-1 indicates malfunction;Q-th of element is concentrated for equilibrium data In include KPI information, q=1,2 ..., se.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, It is as follows without label auxiliary data collection acquisition modes in step 3:
One is obtained from existing a large amount of heterogeneous network user's history measurement report to assist containing g sample without label Data setG is sample total number in A, p-th of unlabeled exemplarsP=1,2 ..., G,Each dimension is the KPI information in user's measurement report, and subscript u indicates no label data.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Step 3 is specific as follows:
By formula (1) to p-th of unlabeled exemplars in no label auxiliary data collection AIt is operated, p=1, 2 ..., g obtains the potential character representation without label auxiliary data collection;
Auxiliary data is concentrated p-th of unlabeled exemplars by first item in formulaIt is expressed as the form of b weighted linear combination, Wherein b={ b1,b2,…,bsThe basal orientation duration set that is, s is the total number of base vector, bj∈Rn, j=1,2 ..., s, bj It is to indicateJ-th of base vector;A={ a(1),a(2),...,a(g)It is activation vector set, p-th of element a in a(p)∈ Rs, RsFor s dimensional vector space, a(p)J-th of componentFor bjIt is correspondingActivation amount;Section 2 is L1 penalty term, β in formula For degree of rarefication control parameter, value determines that the relative importance of first item and Section 2, s.t. indicate to obey pact thereafter Beam condition.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Step 4 is specific as follows:
To in equilibrium data collection T'It, will wherein by formula (2)It is expressed as learning to arrive in step 3 Basal orientation duration set b in element combinations mode;
Wherein, h(q)∈RsFor each element pair in basal orientation duration set bActivation vector, h(q)J-th of componentFor bjIt is correspondingActivation amount;α is punishment term coefficient;What solution obtainedAsNew table at basal orientation duration set b Show;In T'InIt is operated through this and obtains its new expression on b, thus equilibrium data collectionReplacement is expressed asForm,For the balance number under new representation According to collection;
It willAs the training data of svm classifier method, solved using method of Lagrange multipliers Formula (3) obtains disaggregated model (ω, d), and ω, d respectively correspond the normal vector and displacement item of categorised decision plane;
Subscript tran in constraint condition indicates the transposition of vector.
Scheme is advanced optimized as a kind of wireless network fault detection method based on self-teaching of the present invention, Step 5 is specific as follows:
Using obtained disaggregated model (ω, d), KPI information x is tieed up to the n of network user's real-time reportrt∈RnThrough formula (4) the label y of the user reported is determinedrt, subscript rt expression is in real time;
The invention adopts the above technical scheme compared with prior art, has following technical effect that
(1) this method is not limited to carry out network state analysis using the limited user data of Home eNodeB, and by certainly My learning method obtains the promotion in fault detection performance in the form of auxiliary data collection;In addition, it is contemplated that in a network Fault rate is lower, and the data that base station fault is reacted in the Home eNodeB and user data that can be collected into are less, normal Data are more, i.e. the unbalanced problem of Home eNodeB data set, also introduce in the present invention in conjunction with SMOTE (Synthetic Minority Oversampling Technique synthesizes minority class oversampling technique) and ENN (Edited Nearest Neighbor, screen arest neighbors) method of sampling SMOTEENN unbalanced dataset processing method, obtain equilibrium data collection energy It is enough effectively prevented from minority class fault sample to be ignored in disaggregated model, causes what fault detection accuracy rate was greatly lowered to ask Topic;
(2) present invention is embodied in three aspects using the main advantage that the method for self-teaching carries out fault detection: first is that It only needs a small amount of Home eNodeB to have the sample of label (whether illustrate failure) in method flow, meets real network condition, save It has saved to data set tagged plenty of time and human cost;Second is that auxiliary data collection may be from arbitrarily with Home eNodeB number According to the available data with general character, these data are often largely present among network and are not efficiently used;In the present invention Method fully consider and apply these data so that data user rate is improved in network;Third is that timeliness is strong, In new network state environment, it is not necessary to which time-consuming collects mass data and establishes model, improves the efficiency of fault detection;
(3) in the present invention fault detection method based on self-teaching that uses can it is few in family's base station user data, It is more accurate and network failure is effectively detected in the case of generating the easy mistake of data;And the form of self-teaching facilitates migration, The model that can obtain fault detection quickly under new network environment, improves the fault detection efficiency of previous methods.
Detailed description of the invention
Fig. 1 is two layers of macro-home base station network scene.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with the accompanying drawings and the specific embodiments The present invention will be described in detail.
Technical solution of the present invention mainly includes two stages: modelling phase and fault detection stage, process is such as Shown in table 1.Modelling phase mainly includes five steps, is acquiring and saving network KPI (Key Performance Indicator, Key Performance Indicator) it is recorded as after data acquisition system form, the side SMOTEENN is used to the data acquisition system first Method carries out uneven processing;Later by sparse self-encoding encoder from one group of auxiliary data focusing study " base vector ", these " basal orientations It include the feature of image watermarking in amount ";It is again the form of each " base vector " linear combination by the KPI data set representations after balance And it is instructed on the equilibrium data collection under new expression through SVM (Supporting Vector Machine, support vector machines) method herein Get the disaggregated model normally with fault category.
Self-organizing network fault detection method of the table 1 based on self-teaching
The fault detection stage divides the KPI data record that network generates in real time using a stage well-established model Class, and then achieve the purpose that fault detection.
Step 1: it in heterogeneous network pickup area, is obtained by way of the measurement report that user periodically reports All kinds of KPI information.The Reference Signal Received Power RSRP that user measures is collected hereins(Reference Signal Receiving Power), serving cell Reference Signal Received Quality RSRQs(Reference Signal Receiving Quality), maximum adjacent area Reference Signal Received Power (RSRPn), maximum adjacent area Reference Signal Received Quality (RSRQn) etc. four classes Information, wherein subscript s and n respectively indicates serving cell serving and adjacent area neighbor, is used as the KPI of fault detection Information.
The above-mentioned four classes KPI information of the reporting of user in 5 second time is obtained, and according to the status indication of user service base station Corresponding K PI information.
It saves KPI and respective markers is data acquisition systemForm, on Mark m expression acquires m item record.(i=1,2 ..., m) a element i-th in TInSubscript l is the initial of labeled, indicates label data;y(i)It is corresponding Its label reports moment user's state in which, there is normal (y(i)=+1) and failure (y(i)=-1) two kinds of state.In addition, From existing a large amount of heterogeneous network reporting of user without taking one to be used as auxiliary data collection in label measurement reportSubscript g is sample total number in A.Wherein a sample of pth (p=1,2 ..., g)Under Mark u is the initial of unlabeled, indicates no label data.
Step 2: step 1 obtain data set T in label be positive normal element be much more than label be failure element, Data acquisition system T is unbalanced dataset.So SMOTEENN method is used to pre-process to obtain equilibrium data data set T Collect T'.Detailed process is as follows:
(1) to every record that label in T is failureIn KPI informationIt is chosen from T with its distance most The similar KPI information of close k item, k=6 in this example.Enable T'=T.
(2) have extraction 3 put back to from this 6 similar KPI information, it is every when extracting one in itself and xlLine on Machine, which is chosen, a little generates new failure classes KPI, and enabling its corresponding label is -1.
(3) (1), (2) process are repeated, is recorded in every failure classesOn produce 3 new failure classes recordsIt is incorporated in T', that is, updates
(4) to every record in T', choose and it take k'=3 in nearest k'(this example) item record.
(5) judge whether the label to occupy the majority in k' item record and the record label are consistent.Retain this note if consistent This record is rejected in record if inconsistent.
(6) final equilibrium data collection is thus obtainedWhat se was puts down The number of element in weighing apparatus data set.
A element of q in T' (q=1,2 ..., se)InSubscript l is the lead-in of labeled Mother indicates label data;y'(q)Its corresponding label, there is normal condition y'(q)=+1, malfunction y'(q)=-1 two kind.
Step 3: by formula (1) to p-th yuan of prime element in auxiliary data collection AP=1,2 ..., g is grasped Make the basal orientation duration set b of our available no label auxiliary data collection.
Step 4: to each element in equilibrium data collection T'(q=1,2 ..., se), by formula (2), It will whereinBe expressed as the mode of element combinations in the above-mentioned basal orientation duration set b learnt, obtain its it is corresponding activate to Amount
Each element in T'InCan obtain its new expression on b through this operation, thus put down Weigh data setIt can replace and be expressed asForm.
Step 5: willAs the training data of svm classifier method, according to bright using glug Day multiplier method solution formula (3) obtains disaggregated model (ω, d).
Two, the fault detection stage
The disaggregated model (ω, d) obtained using the modelling phase, to the four-dimensional KPI information of network user's real-time report xrt∈R4, subscript rt expression is real-time, and the label y for the user that the moment reports is determined through formula (4)rt
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by those familiar with the art, all answers It is included within the scope of protection of the present invention.

Claims (10)

1. a kind of wireless network fault detection method based on self-teaching, which comprises the following steps:
Step 1: acquisition network key performance indicator KPI is recorded and is saved as data acquisition system;
It combines synthesis minority class oversampling technique SMOTE Step 2: being used to the data acquisition system and screens arest neighbors ENN's SMOTEENN method carries out uneven processing, the equilibrium data collection after being balanced;
Step 3: by sparse self-encoding encoder from no label auxiliary data focusing study base vector;
Step 4: being the form of each base vector linear combination by the equilibrium data set representations after balance, and herein under new expression The normally disaggregated model with fault category is obtained through the training of support vector machines method on equilibrium data collection;
Step 5: being classified using well-established disaggregated model to the KPI data record that network generates in real time, and then reach The purpose of fault detection.
2. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that just Often refer to non-faulting.
3. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that step Rapid one is specific as follows:
In heterogeneous network, user periodically reports all kinds of KPI information in the form of measurement report;
From the KPI information for obtaining reporting of user in certain time in the heterogeneous network for needing to carry out fault detection, and according to user It reports constantly to the normal KPI information corresponding with malfunction label of its serving BS;KPI information and correspondence markings are saved as Data acquisition systemForm, m be T in element total number, i.e., acquire m altogether Item record;I-th of element in TInRepresent the n dimension KPI information of certain moment reporting of user, subscript l table Label data is shown with,I-th to report has label KPI information, RnFor n-dimensional vector space, i=1,2 ..., m;y(i)It is right It answersLabel have normal condition and two kinds of malfunction to report moment user service base station state in which.
4. a kind of wireless network fault detection method based on self-teaching according to claim 3, which is characterized in that just Y when normal state(i)=+1, y when malfunction(i)=-1.
5. a kind of wireless network fault detection method based on self-teaching according to claim 3, which is characterized in that just Normal state refers to non-faulting state.
6. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that step Rapid two is specific as follows:
2.1, to every record that label in T is failureAccording to the KPI information in recordIt is chosen from T and includes KPI information withApart from the nearest similar KPI record of k item, failure refers to y(i)=-1 separately takes a new data set T'=T;
2.2, r item is extracted from the KPI information of the similar record of k item with putting back to, r is taken as 1 to the integer value between k, every extraction At one this record include KPI withLine on randomly select and a little generate new failure classes KPI xl_new, subscript L_new, which indicates new, label data;Enable xl_newCorresponding label be -1;
2.3,2.1,2.2 processes are repeated, are recorded in every failure classesIt is upper to generate the new failure classes record of r itemIt is incorporated in data set T', i.e. update T' is For the v articles new KPI of generation Information;Indicate the v articles new record generated;∪ expression takes its front and back union of sets collection;
2.4, it to every record in T', chooses and it is recorded apart from nearest k' item;
2.5, judge whether the label to occupy the majority in k' item record is consistent with the label of record to be judged, retaining if consistent should Item record rejects this record if inconsistent;
2.6, final equilibrium data collection is thus obtainedThe balance number that se is According to the number for concentrating element;Q-th of element in T'Iny'(q)It is correspondingLabel, y'(q)=+1 When indicate normal condition, y'(q)=-1 indicates malfunction;Q-th of element is concentrated for equilibrium dataMiddle packet The KPI information contained, q=1,2 ..., se.
7. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that step It is as follows without label auxiliary data collection acquisition modes in rapid three:
Obtained from existing a large amount of heterogeneous network user's history measurement report one containing g sample without label auxiliary data CollectionG is sample total number in A, p-th of unlabeled exemplars Each dimension is the KPI information in user's measurement report, and subscript u indicates no label data.
8. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that step Rapid three is specific as follows:
By formula (1) to p-th of unlabeled exemplars in no label auxiliary data collection AIt is operated,It obtains Take the potential character representation of no label auxiliary data collection;
Auxiliary data is concentrated p-th of unlabeled exemplars by first item in formulaIt is expressed as the form of b weighted linear combination, wherein b ={ b1,b2,…,bsThe basal orientation duration set that is, s is the total number of base vector, bj∈ Rn, j=1,2 ..., s, bjIt is table ShowJ-th of base vector;A={ a(1),a(2),…,a(g)It is activation vector set, p-th of element a in a(p)∈Rs, RsFor s Dimensional vector space, a(p)J-th of componentFor bjIt is correspondingActivation amount;Section 2 is L1 penalty term in formula, and β is sparse Control parameter is spent, value determines that the relative importance of first item and Section 2, s.t. indicate to obey constraint condition thereafter.
9. a kind of wireless network fault detection method based on self-teaching according to claim 1, which is characterized in that step Rapid four is specific as follows:
To in equilibrium data collection T'It, will wherein by formula (2)It is expressed as the base learnt in step 3 The mode of element combinations in vector set b;
Wherein, h(q)∈RsFor each element pair in basal orientation duration set bActivation vector, h(q)J-th of componentFor bjIt is right It answersActivation amount;α is punishment term coefficient;What solution obtainedAsNew expression at basal orientation duration set b; In T'InIt is operated through this and obtains its new expression on b, thus equilibrium data collectionReplacement is expressed asForm,For the balance number under new representation According to collection;
It willAs the training data of svm classifier method, method of Lagrange multipliers solution formula is utilized (3) it obtains disaggregated model (ω, d), ω, d respectively correspond the normal vector and displacement item of categorised decision plane;
Subscript tran in constraint condition indicates the transposition of vector.
10. a kind of wireless network fault detection method based on self-teaching according to claim 9, which is characterized in that Step 5 is specific as follows:
Using obtained disaggregated model (ω, d), KPI information x is tieed up to the n of network user's real-time reportrt∈RnIt is true through formula (4) The label y of the fixed user reportedrt, subscript rt expression is in real time;
CN201910343031.XA 2019-04-26 2019-04-26 Wireless network fault detection method based on self-learning Active CN110139315B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910343031.XA CN110139315B (en) 2019-04-26 2019-04-26 Wireless network fault detection method based on self-learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910343031.XA CN110139315B (en) 2019-04-26 2019-04-26 Wireless network fault detection method based on self-learning

Publications (2)

Publication Number Publication Date
CN110139315A true CN110139315A (en) 2019-08-16
CN110139315B CN110139315B (en) 2021-09-28

Family

ID=67575190

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910343031.XA Active CN110139315B (en) 2019-04-26 2019-04-26 Wireless network fault detection method based on self-learning

Country Status (1)

Country Link
CN (1) CN110139315B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110972174A (en) * 2019-12-02 2020-04-07 东南大学 Wireless network interruption detection method based on sparse self-encoder
CN112906727A (en) * 2019-12-04 2021-06-04 中国电信股份有限公司 Method and system for real-time online detection of virtual machine state
WO2022236807A1 (en) * 2021-05-14 2022-11-17 Qualcomm Incorporated Model status monitoring, reporting, and fallback in machine learning applications

Citations (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105704103A (en) * 2014-11-26 2016-06-22 中国科学院沈阳自动化研究所 Modbus TCP communication behavior abnormity detection method based on OCSVM double-contour model
CN105745868A (en) * 2013-11-26 2016-07-06 瑞典爱立信有限公司 Method and apparatus for anomaly detection in a network
CN106101015A (en) * 2016-07-19 2016-11-09 广东药科大学 A kind of mobile Internet traffic classes labeling method and system
CN107066759A (en) * 2017-05-12 2017-08-18 华北电力大学(保定) A kind of Vibration Fault Diagnosis of Turbine Rotor method and device
CN107171819A (en) * 2016-03-07 2017-09-15 北京华为数字技术有限公司 A kind of network fault diagnosis method and device
US20170270429A1 (en) * 2016-03-21 2017-09-21 Xerox Corporation Methods and systems for improved machine learning using supervised classification of imbalanced datasets with overlap
CN107370617A (en) * 2017-03-10 2017-11-21 南京航空航天大学 Cellular network fault diagnosis system based on SVM
CN107426199A (en) * 2017-07-05 2017-12-01 浙江鹏信信息科技股份有限公司 A kind of method and system of Network anomalous behaviors detection and analysis
CN107784325A (en) * 2017-10-20 2018-03-09 河北工业大学 Spiral fault diagnosis model based on the fusion of data-driven increment
CN107943916A (en) * 2017-11-20 2018-04-20 安徽大学 A kind of webpage abnormality detection method based on online classification
CN108322445A (en) * 2018-01-02 2018-07-24 华东电力试验研究院有限公司 A kind of network inbreak detection method based on transfer learning and integrated study
CN108366386A (en) * 2018-05-11 2018-08-03 东南大学 A method of using neural fusion wireless network fault detect
CN108540330A (en) * 2018-04-24 2018-09-14 南京邮电大学 A kind of network fault diagnosis method based on deep learning under heterogeneous network environment
CN108696379A (en) * 2017-04-07 2018-10-23 南京航空航天大学 Cellular network fault diagnosis system based on integrated study and SMOTE
CN108764366A (en) * 2018-06-07 2018-11-06 南京信息职业技术学院 Feature selecting and cluster for lack of balance data integrate two sorting techniques
CN108768772A (en) * 2018-05-29 2018-11-06 南京航空航天大学 The fault detection method of self-organizing network based on cost-sensitive
CN109150564A (en) * 2017-06-19 2019-01-04 中国移动通信集团广东有限公司 A kind of prediction technique and device for cell fault warning
CN109617715A (en) * 2018-11-27 2019-04-12 中盈优创资讯科技有限公司 Network fault diagnosis method, system
CN109670446A (en) * 2018-12-20 2019-04-23 泉州装备制造研究所 Anomaly detection method based on linear dynamic system and depth network

Patent Citations (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105745868A (en) * 2013-11-26 2016-07-06 瑞典爱立信有限公司 Method and apparatus for anomaly detection in a network
CN105704103A (en) * 2014-11-26 2016-06-22 中国科学院沈阳自动化研究所 Modbus TCP communication behavior abnormity detection method based on OCSVM double-contour model
CN107171819A (en) * 2016-03-07 2017-09-15 北京华为数字技术有限公司 A kind of network fault diagnosis method and device
US20170270429A1 (en) * 2016-03-21 2017-09-21 Xerox Corporation Methods and systems for improved machine learning using supervised classification of imbalanced datasets with overlap
CN106101015A (en) * 2016-07-19 2016-11-09 广东药科大学 A kind of mobile Internet traffic classes labeling method and system
CN107370617A (en) * 2017-03-10 2017-11-21 南京航空航天大学 Cellular network fault diagnosis system based on SVM
CN108696379A (en) * 2017-04-07 2018-10-23 南京航空航天大学 Cellular network fault diagnosis system based on integrated study and SMOTE
CN107066759A (en) * 2017-05-12 2017-08-18 华北电力大学(保定) A kind of Vibration Fault Diagnosis of Turbine Rotor method and device
CN109150564A (en) * 2017-06-19 2019-01-04 中国移动通信集团广东有限公司 A kind of prediction technique and device for cell fault warning
CN107426199A (en) * 2017-07-05 2017-12-01 浙江鹏信信息科技股份有限公司 A kind of method and system of Network anomalous behaviors detection and analysis
CN107784325A (en) * 2017-10-20 2018-03-09 河北工业大学 Spiral fault diagnosis model based on the fusion of data-driven increment
CN107943916A (en) * 2017-11-20 2018-04-20 安徽大学 A kind of webpage abnormality detection method based on online classification
CN108322445A (en) * 2018-01-02 2018-07-24 华东电力试验研究院有限公司 A kind of network inbreak detection method based on transfer learning and integrated study
CN108540330A (en) * 2018-04-24 2018-09-14 南京邮电大学 A kind of network fault diagnosis method based on deep learning under heterogeneous network environment
CN108366386A (en) * 2018-05-11 2018-08-03 东南大学 A method of using neural fusion wireless network fault detect
CN108768772A (en) * 2018-05-29 2018-11-06 南京航空航天大学 The fault detection method of self-organizing network based on cost-sensitive
CN108764366A (en) * 2018-06-07 2018-11-06 南京信息职业技术学院 Feature selecting and cluster for lack of balance data integrate two sorting techniques
CN109617715A (en) * 2018-11-27 2019-04-12 中盈优创资讯科技有限公司 Network fault diagnosis method, system
CN109670446A (en) * 2018-12-20 2019-04-23 泉州装备制造研究所 Anomaly detection method based on linear dynamic system and depth network

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
JOSEY MATHEW等: "Classification of Imbalanced Data by Oversampling in Kernel Space of Support Vector Machines", 《 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS》 *
YI CUI: "Improvement of power transformer insulation diagnosis using oil characteristics data preprocessed by SMOTEBoost technique", 《 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION》 *
张涛: "基于支持向量机的不平衡数据分类方法研究与应用", 《中国优秀硕士学位论文库》 *
李卓然: "基于主动学习的非均衡数据分类研究", 《中国优秀硕士学位论文库》 *
沈学利,覃淑娟: "基于SMOTE 和深度信念网络的异常检测", 《计算机应用》 *
王健: "面向样本不平衡的故障特征提取方法", 《中国博士学位论文库》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110972174A (en) * 2019-12-02 2020-04-07 东南大学 Wireless network interruption detection method based on sparse self-encoder
CN110972174B (en) * 2019-12-02 2022-12-30 东南大学 Wireless network interruption detection method based on sparse self-encoder
CN112906727A (en) * 2019-12-04 2021-06-04 中国电信股份有限公司 Method and system for real-time online detection of virtual machine state
WO2022236807A1 (en) * 2021-05-14 2022-11-17 Qualcomm Incorporated Model status monitoring, reporting, and fallback in machine learning applications

Also Published As

Publication number Publication date
CN110139315B (en) 2021-09-28

Similar Documents

Publication Publication Date Title
CN110139315A (en) A kind of wireless network fault detection method based on self-teaching
WO2022057268A1 (en) Base station energy-saving policy determining method and apparatus, and storage medium
CN102592422B (en) Automatic meter reading system collector and ammeter asset number collect management method
CN103763706B (en) Collaborative spectrum sensing vague fusion method based on clustering
CN105706325A (en) Grid frequency response
CN106131959A (en) A kind of dual-positioning method divided based on Wi Fi signal space
CN111405585B (en) Neighbor relation prediction method based on convolutional neural network
CN106597133B (en) A kind of low voltage electric network uniline phase sequence method for identifying and classifying and device
CN110502547A (en) Internet of Things detection data management system and Internet of Things detection data management method
CN105208580B (en) Physical-layer cell identifier PCI optimization method and system in LTE network
CN108632111A (en) Service link monitoring method based on log
CN114092769A (en) Transformer substation multi-scene inspection analysis method based on federal learning
CN113709779A (en) Cellular network fault diagnosis method
CN103369044A (en) Mobile terminal user network perception diagnosis and treatment method based on cloud knowledge base
CN114336615A (en) Power resource scheduling intelligent management system based on big data analysis of power transmission environment
CN115642652A (en) Electric power adjusting device based on artificial intelligence
CN106227689A (en) A kind of based on Internet of Things and the data collection station of cloud
CN108732477A (en) A kind of system and method for the power distribution network main equipment live detection based on electric power wireless communication
CN105242330A (en) Detection method and device for weather condition, and mobile terminal
CN106294873A (en) The analytical equipment of a kind of machine data and the method for analysis
CN113902583A (en) Distribution network side operation and maintenance method and system using low-voltage network equipment data
CN104182470A (en) SVM (support vector machine) based mobile terminal application classification system and method
CN104834981B (en) The main cause of short-term power station photovoltaic power implies type prediction technique
CN115933750A (en) Data processing-based power inspection method and power inspection system
CN107017835A (en) The fault monitoring system and method for photovoltaic module

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
GR01 Patent grant
GR01 Patent grant