CN106507398B - A network self-optimization method based on continuous learning - Google Patents
A network self-optimization method based on continuous learning Download PDFInfo
- Publication number
- CN106507398B CN106507398B CN201611234233.3A CN201611234233A CN106507398B CN 106507398 B CN106507398 B CN 106507398B CN 201611234233 A CN201611234233 A CN 201611234233A CN 106507398 B CN106507398 B CN 106507398B
- Authority
- CN
- China
- Prior art keywords
- network
- optimization
- continuous learning
- method based
- time delay
- 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.)
- Active
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 57
- 230000000694 effects Effects 0.000 claims description 23
- 238000011156 evaluation Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013135 deep learning Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000002360 preparation method Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 4
- 239000000463 material Substances 0.000 abstract description 4
- 230000000875 corresponding effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000000275 quality assurance Methods 0.000 description 1
- 230000007115 recruitment Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明公开了一种基于持续学习的网络自优化方法；包含持续学习过程和网络优化过程；本发明提供的网络自优化方法可以大大减少人力物力的投入，节约成本，缩短优化流程，提高优化效率，同时解决上述发明优化时间冗长，可能不是最佳优化策略的缺陷;快速地发现网络中出现的问题，并能够缩短网络故障的持续时间，及时恢复网络正常的工作状态，达到优化网络性能的目的。
The invention discloses a network self-optimization method based on continuous learning; it includes a continuous learning process and a network optimization process; the network self-optimization method provided by the invention can greatly reduce the input of manpower and material resources, save costs, shorten the optimization process, and improve the optimization efficiency. At the same time, it solves the defect of the above-mentioned invention that the optimization time is long and may not be the best optimization strategy; quickly finds problems in the network, and can shorten the duration of network failures, restore the normal working state of the network in time, and achieve the purpose of optimizing network performance .
Description
Technical field
The invention belongs to network optimization field more particularly to a kind of network self-organization methods based on continuous learning.
Background technique
With the extensive growth of mobile device, wireless traffic is also grown rapidly, and user is for wireless network cut-in quality
Requirement also increasingly increase, so if obtain any user can with quality assurance at any time and any place
Service, it is necessary to which network is constantly optimized.
Wireless communication system becomes network management and optimization very due to the explosive growth of user and business
It is difficult.The matter of utmost importance that network operator faces is that there are thousands of a network elements (such as base station) in network, and wherein each
Network element is likely to go wrong or failure, some in these problems may be not obvious at its initial stage, if will rely on
Manually go find and check if, that workload will be it is huge even not possible with.Until now, the network optimization still needs
A large amount of professional and superior equipment, this considerably increases the costs of operator, reduce profit space.Operator spends at present
The great number operation maintenance cost of expense has been difficult to improve bring income by extra-service performance to compensate, because of each user
Average yield constantly decline.
Such as application No. is a kind of method, apparatus of 201410371501.0 network self-organization；The one of invention offer
The method, apparatus of kind network self-organization obtains special optimisation strategy first, and then sequence, which executes in special project optimisation strategy, includes
The sub- optimisation strategy of each single item, and recruitment evaluation is carried out to judge whether to have reached pre- to the sub- optimisation strategy of each single item being finished
If effectiveness indicator, just complete the process of entire self-optimizing when reached, terminate the self-optimizing process, otherwise continue to execute down
One sub- optimisation strategy, repeats the above steps.Compared with artificial optimization's method in the prior art, send elsewhere bright by the inclusion of a series of
The special optimisation strategy of sub- optimisation strategy realizes network self-organization, and carries out effect after the completion of each sub- optimisation strategy executes
Assessment, for it is a kind of with self-evaluating mechanism towards strategy self-organization method, relative to artificial discovery problem, problem analysis,
The Optimizing Flow solved the problems, such as, hand labor it is less thus can reduce the investment of corresponding manpower and material resources in optimization process, save
Cost, simultaneously because the method for self-optimizing provided by the invention is that can execute automatically after triggering, the degree of automation is higher, can
Shorten Optimizing Flow, improve optimization efficiency.But only simply sequence executes optimisation strategy for the invention, until effect of optimization changes
It is kind.Defect present in it has two o'clock: 1. selecting sequences execute the mode of optimisation strategy, and it is tediously long to will lead to the optimization time；2. working as
Effect of optimization just terminates entire self-optimizing process after being improved, may result in the optimisation strategy is not optimal policy.
Summary of the invention
The technical problem to be solved by the present invention is to propose that one kind is based on holding to solve shortcoming and defect in the prior art
The network self-organization method of continuous study；Its investment for reducing manpower and material resources, save the cost shorten Optimizing Flow, improve optimization effect
Rate, at the same solve foregoing invention optimization the time it is tediously long, may not be the defect of optimum optimization strategy.
The present invention uses following technical scheme to solve above-mentioned technical problem
A kind of network self-organization method based on continuous learning includes continuous learning process and network optimization process；
The continuous learning process specifically includes the following steps；
Step 1, prepare the network optimization problem classified；
Step 2, the characteristic information for the network optimization problem that extraction step 1 prepares, and the characteristic information extracted is trained
At a prior model；
Step 3, classified using prior model to subsequent network problem；
Step 4, if there is new network optimization problem, new network optimization problem is extracted using unsupervised learning
Characteristic information；
Step 5, the characteristic information obtained according to step 4 is by Active Learning come the priori mould of renolation step 2 training
Type；
The network optimization process specifically includes the following steps:
Step 6, after the classification of known network optimization problem, the optimisation strategy for belonging to such network optimization problem is obtained；
Step 7, the optimisation strategy of such network optimization problem is executed；
Step 8, effect of optimization corresponding with the optimisation strategy executed is obtained；
Step 9, it is evaluated according to each effect of optimization of preset Indexes of Evaluation Effect to acquisition；
Step 10, all evaluations are compared, select optimal Network Optimization Strategy.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, in step 2,
The characteristic information of the network optimization problem of step 1 preparation is extracted using deep learning method.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the optimization
Strategy is respectively the optimisation strategy of covering class problem and the optimisation strategy of signal quality class problem；
Wherein, covering class problem optimisation strategy includes adjustment antenna feeder, adjustment pilot power, adjustment system ovelay range, inspection
Whether normal look into neighbor stations RxLev；
The optimisation strategy of signal quality class includes PCI optimization, adjustment antenna feeder, the leading covering of enhancing, adjustment pilot power.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the effect is commented
Valence index includes covering class index, call setup class index, calling holding class index and time delay class index
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the covering class
Index includes RSRP, RSSI, RSRQ, SINR.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the call setup
Vertical class index includes that RRC connection is created as that power, RRC connection is created as power, E-RAB is created as power, wireless percent of call completed.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the calling is protected
Holding class index includes RRC connection abnormal cutting off rate, E-RAB cutting off rate.
As a kind of further preferred scheme of the network self-organization method based on continuous learning of the present invention, the time delay class
Index include UE converted from Idle state to Active state time delay, Attach time delay, user plane time delay, in system in X2 service switchover
S1 service switchover interrupts time delay in disconnected time delay, system, different system service switchover interrupts time delay.
The invention adopts the above technical scheme compared with prior art, has following technical effect that
Network self-organization method provided by the invention can greatly reduce the input cost of manpower and material resources, reduce network construction
With maintenance cost；Automatically, it rapidly finds the problem in network, and the duration of network failure can be shortened, in time
Restore the normal working condition of network, achievees the purpose that optimize network performance.
Detailed description of the invention
Fig. 1 is continuous learning process of the invention；
Fig. 2 is prior model schematic diagram of the present invention；
Fig. 3 is the updated prior model schematic diagram of the present invention；
Fig. 4 is network optimization process of the invention；
Fig. 5 is the database that present example 1 provides.
Specific embodiment
Technical solution of the present invention is described in further detail with reference to the accompanying drawing:
A kind of network self-organization process based on continuous learning is as follows: including continuous learning process and network optimization process；
As shown in Figure 1, the process of continuous learning is as follows:
Step 1, prepare the network optimization problem classified；The network of step 1 preparation is extracted using deep learning method
The characteristic information of optimization problem.
Step 2, the characteristic information for the network optimization problem that extraction step 1 prepares, and the characteristic information extracted is trained
At a prior model；
Step 3, classified using prior model to subsequent network problem；
Step 4, if there is new network optimization problem, new network optimization problem is extracted using unsupervised learning
Characteristic information；
Step 5, the characteristic information obtained according to step 4, come progressive updating sophisticated model, is reached and is continued using Active Learning
The destination of study.
The purpose of continuous learning is: improving prior model, classifies to different network optimization problems.It can after classification
To use different optimisation strategies for different optimization problems.Wherein prior model is in the nature characteristic information and problem category
Corresponding relationship, as shown in the figure:
When there is new network problem, the appearance of new characteristic information is also meaned that.Assuming that prior model such as Fig. 2 institute
Show, when new network problem occurs, characteristic information is characterized information 1, characteristic information 3, characteristic information 4 and new feature
Information 9, then the network problem will be classified as classification 1, at the same time, prior model can be also updated, characteristic information 9
Problem category 1 can be pointed to.The prior model of Fig. 2 will be updated to as shown in Figure 3.
After classifying to network optimization problem, start the self-optimizing of network, as shown in figure 4, detailed process is as follows:
Step 6, after the classification of known network problem, the optimisation strategy for belonging to such optimization problem is obtained；
Step 7, such all optimisation strategy is executed；
Step 8, effect of optimization corresponding with performed optimisation strategy is obtained；
Step 9, each effect of optimization is evaluated according to preset Indexes of Evaluation Effect；
Step 10, all evaluations are compared, optimisation strategy described in optimal Network Optimization Strategy is selected to distinguish
To cover the optimisation strategy of class problem and the optimisation strategy of signal quality class problem；
Wherein, covering class problem optimisation strategy includes adjustment antenna feeder, adjustment pilot power, adjustment system ovelay range, inspection
Whether normal look into neighbor stations RxLev；
The optimisation strategy of signal quality class includes PCI optimization, adjustment antenna feeder, the leading covering of enhancing, adjustment pilot power.
The Indexes of Evaluation Effect includes covering class index, call setup class index, calling holding class index and time delay class
Index
The covering class index includes RSRP, RSSI, RSRQ, SINR.
The call setup class index includes that RRC connection is created as that power, RRC connection is created as power, E-RAB is created as
Power, wireless percent of call completed.
It includes RRC connection abnormal cutting off rate, E-RAB cutting off rate that the calling, which keeps class index,.
The time delay class index include UE converted from Idle state to Active state time delay, Attach time delay, user plane time delay,
X2 service switchover interrupts time delay in system, S1 service switchover interrupts time delay, different system service switchover interrupts time delay in system.
It is excellent to train a continuous renewal for the method that network self-organization method provided by the invention uses continuous learning first
The prior model of change later classifies to non-classified network optimization problem, and then selection belongs to such optimisation strategy pair
Network optimizes, and is finally evaluated using preset evaluation index effect, selects optimum optimization strategy.
Specific embodiment is as follows:
Example 1, when occurring unknown problem in network, and when the problem is new network problem, processing step is as follows:
Step 1: extraction characteristic information first；
Step 2: being classified later using prior model to the problem；
Step 3: and then be updated using the characteristic information extracted to prior model；
Step 4: assuming that the problem is classified as downlink signal quality problem (network problem classification 1), for such problem,
There are three types of optimisation strategies；
Step 5: executing such three kinds of all optimisation strategy, obtain corresponding effect of optimization；
Step 6: three kinds of effect of optimization being evaluated according to preset Indexes of Evaluation Effect；
Step 7: three kinds of evaluations being compared, optimal Network Optimization Strategy is selected；
Involved database is as shown in Figure 5 in this example:
Example 2, when going wrong in network, and when the problem has been classified, processing step is as follows:
Step 1: extraction characteristic information first；
Step 2: being classified later using prior model to the problem；
Step 3: assuming that the problem is classified as downlink signal quality problem (network problem classification 1), for such problem,
There are three types of optimisation strategies；
Step 4: executing such three kinds of all optimisation strategy, obtain corresponding effect of optimization；
Step 5: three kinds of effect of optimization being evaluated according to preset Indexes of Evaluation Effect；
Step 6: three kinds of evaluations being compared, optimal Network Optimization Strategy is selected.
Key point in continuous learning is: preparing a series of network optimization problems classified and uses deep learning side
Method or model extract characteristic information therein and are trained a prior model, and it is complete that the two operations do not need user
At and after network operator trains a prior model, which can be constantly updated and perfect by continuous learning.
Key when executing optimisation strategy is: executing, but holds until meeting preset evaluation index not in accordance with sequence
All optimisation strategies under the row category select the strategy with optimal evaluation effect to be applied in network.
Claims (8)
1. a kind of network self-organization method based on continuous learning, it is characterised in that: include continuous learning process and the network optimization
Process；The continuous learning process specifically includes the following steps；
Step 1, prepare the network optimization problem classified；
Step 2, the characteristic information for the network optimization problem that extraction step 1 prepares, and the characteristic information extracted is trained to one
A prior model；
Step 3, classified using prior model to subsequent network problem；
Step 4, if there is new network optimization problem, the spy of new network optimization problem is extracted using unsupervised learning
Reference breath；
Step 5, the characteristic information obtained according to step 4 is by Active Learning come the prior model of renolation step 2 training；
The network optimization process specifically includes the following steps:
Step 6, after the classification of known network optimization problem, the optimisation strategy for belonging to such network optimization problem is obtained；
Step 7, the optimisation strategy of such network optimization problem is executed；
Step 8, effect of optimization corresponding with the optimisation strategy executed is obtained；
Step 9, it is evaluated according to each effect of optimization of preset Indexes of Evaluation Effect to acquisition；
Step 10, all evaluations are compared, select optimal Network Optimization Strategy.
2. a kind of network self-organization method based on continuous learning according to claim 1, it is characterised in that: in step 2
In, the characteristic information of the network optimization problem of step 1 preparation is extracted using deep learning method.
3. a kind of network self-organization method based on continuous learning according to claim 1, it is characterised in that: in step 6
In, the optimisation strategy is respectively the optimisation strategy of the optimisation strategy and signal quality class problem that cover class problem；
Wherein, covering class problem optimisation strategy includes adjustment antenna feeder, adjustment pilot power, adjustment system ovelay range, checks phase
Whether neighboring station RxLev is normal；
The optimisation strategy of signal quality class includes PCI optimization, adjustment antenna feeder, the leading covering of enhancing, adjustment pilot power.
4. a kind of network self-organization method based on continuous learning according to claim 1, it is characterised in that: in step 9
In, the Indexes of Evaluation Effect includes covering class index, call setup class index, calling holding class index and time delay class index.
5. a kind of network self-organization method based on continuous learning according to claim 4, it is characterised in that: the covering
Class index includes RSRP, RSSI, RSRQ, SINR.
6. a kind of network self-organization method based on continuous learning according to claim 4, it is characterised in that: the calling
Establishing class index includes that RRC connection is created as power, E-RAB is created as power, wireless percent of call completed.
7. a kind of network self-organization method based on continuous learning according to claim 4, it is characterised in that: the calling
Keeping class index includes RRC connection abnormal cutting off rate, E-RAB cutting off rate.
8. a kind of network self-organization method based on continuous learning according to claim 4, it is characterised in that: the time delay
Class index includes that UE converts time delay, Attach time delay, user plane time delay, X2 service switchover in system from Idle state to Active state
S1 service switchover interrupts time delay in interruption time delay, system, different system service switchover interrupts time delay.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611234233.3A CN106507398B (en) | 2016-12-28 | 2016-12-28 | A network self-optimization method based on continuous learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611234233.3A CN106507398B (en) | 2016-12-28 | 2016-12-28 | A network self-optimization method based on continuous learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106507398A CN106507398A (en) | 2017-03-15 |
| CN106507398B true CN106507398B (en) | 2019-06-21 |
Family
ID=58334555
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611234233.3A Active CN106507398B (en) | 2016-12-28 | 2016-12-28 | A network self-optimization method based on continuous learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106507398B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107517474B (en) * | 2017-10-20 | 2021-11-12 | 北京神州泰岳软件股份有限公司 | Network analysis optimization method and device |
| CN109474497A (en) * | 2018-12-19 | 2019-03-15 | 四川艾贝斯科技发展有限公司 | A kind of reliable network maintenance terminal deep learning algorithm |
| CN111385822B (en) * | 2018-12-29 | 2021-11-09 | 华为技术有限公司 | Configuration method and controller |
| CN112188505B (en) * | 2019-07-02 | 2024-05-10 | 中兴通讯股份有限公司 | Network optimization method and device |
| CN112512059A (en) * | 2020-05-24 | 2021-03-16 | 中兴通讯股份有限公司 | Network optimization method, server, network side equipment, system and storage medium |
| CN113705067B (en) * | 2021-10-29 | 2022-02-18 | 中国电力科学研究院有限公司 | A method, system, device and storage medium for generating a microgrid optimization operation strategy |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877857A (en) * | 2009-04-28 | 2010-11-03 | 五邑大学 | Intelligent Decision Support System for Mobile Communication Network Optimization |
| CN101998411A (en) * | 2010-12-10 | 2011-03-30 | 上海百林通信软件有限公司 | Sweep frequency and propagation model coverage prediction-based frequency optimization method |
| CN102014412A (en) * | 2010-12-10 | 2011-04-13 | 上海百林通信软件有限公司 | Wireless network telephone traffic balanced optimization method based on antenna parameter adjustment |
| CN102404756A (en) * | 2011-11-15 | 2012-04-04 | 上海百林通信网络科技有限公司 | Antenna parameter optimization method based on mobile phone measurement report |
| WO2012134913A1 (en) * | 2011-04-01 | 2012-10-04 | At&T Mobility Ii Llc | Provisioning of a femto access point with parameter profiles of a nearby femto access point |
| US9369351B2 (en) * | 2014-01-06 | 2016-06-14 | Cisco Technology, Inc. | Using learning machine-based prediction in multi-hopping networks |
-
2016
- 2016-12-28 CN CN201611234233.3A patent/CN106507398B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877857A (en) * | 2009-04-28 | 2010-11-03 | 五邑大学 | Intelligent Decision Support System for Mobile Communication Network Optimization |
| CN101998411A (en) * | 2010-12-10 | 2011-03-30 | 上海百林通信软件有限公司 | Sweep frequency and propagation model coverage prediction-based frequency optimization method |
| CN102014412A (en) * | 2010-12-10 | 2011-04-13 | 上海百林通信软件有限公司 | Wireless network telephone traffic balanced optimization method based on antenna parameter adjustment |
| WO2012134913A1 (en) * | 2011-04-01 | 2012-10-04 | At&T Mobility Ii Llc | Provisioning of a femto access point with parameter profiles of a nearby femto access point |
| CN102404756A (en) * | 2011-11-15 | 2012-04-04 | 上海百林通信网络科技有限公司 | Antenna parameter optimization method based on mobile phone measurement report |
| US9369351B2 (en) * | 2014-01-06 | 2016-06-14 | Cisco Technology, Inc. | Using learning machine-based prediction in multi-hopping networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106507398A (en) | 2017-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106507398B (en) | A network self-optimization method based on continuous learning | |
| CN109151881B (en) | Network load balancing optimization method based on user data | |
| CN103929781B (en) | Cross-layer interference coordination optimization method in super dense heterogeneous network | |
| US9473956B2 (en) | Antenna tilt optimization in a wireless communications network | |
| CN103858461A (en) | Neighbor relation optimization methods and device | |
| CN101790173A (en) | Network coverage detection method and system | |
| CN101964985A (en) | Coverage and capacity self-optimization device of self-organization network in LTE/LTE-A and method thereof | |
| CN103179580A (en) | Method and device for self-adaptively optimizing coverage | |
| CN103354647B (en) | Adjacent cell optimization method in mobile communication system and apparatus thereof | |
| CN102377462A (en) | Method and system for interactively selecting auxiliary cell | |
| CN102612070B (en) | Method for configuring wireless resource management and measurement set under multi-point coordination and base station | |
| CN102685783B (en) | Method and device for searching network coverage dead zone in mobile communication network | |
| WO2019119866A1 (en) | Method and device for automatic addition of neighboring cell | |
| CN104113864A (en) | Self-optimizing method and device of network | |
| CN108243435A (en) | A parameter optimization method and device in LTE cell scene division | |
| CN106034311B (en) | An information processing method and device | |
| Guan et al. | Research on evaluation method of wireless network cross-boundary coverage based on smart terminals location data | |
| CN105933923B (en) | A kind of super-intensive networking cell cluster-dividing method | |
| CN103517308A (en) | Cell interrupt detecting method based on neighbor cell reference signal receiving power list | |
| CN107241749A (en) | A kind of network optimized approach and device | |
| CN104754609A (en) | Network optimization method and system based on TD-LTE (Time Division-Long Term Evolution) and WLAN (Wireless Local Area Network) | |
| US9277509B1 (en) | Dynamic power boosting based on contiguous and non-contiguous coverage | |
| CN102781015A (en) | Method and system for quantitatively evaluating complexity of network structures | |
| EP4111728A1 (en) | Cellular network indoor traffic auto-detection | |
| WO2016065759A1 (en) | Method and apparatus for optimizing neighbour cell list |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |