CN103095496A - Prediction method and device for network flow - Google Patents
Prediction method and device for network flow Download PDFInfo
- Publication number
- CN103095496A CN103095496A CN201310008445XA CN201310008445A CN103095496A CN 103095496 A CN103095496 A CN 103095496A CN 201310008445X A CN201310008445X A CN 201310008445XA CN 201310008445 A CN201310008445 A CN 201310008445A CN 103095496 A CN103095496 A CN 103095496A
- Authority
- CN
- China
- Prior art keywords
- component
- network traffics
- network
- outcome
- predicted
- 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
Images
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses prediction method and device for network flow. With the adoption of the technical scheme, the prediction method and device for network flow comprises the following steps in the multiscale frequency application scene: firstly conducting wavelet decomposition and reconfiguration on the to-be-predicted network flow, and generating a plurality of network flow components; then, respectively conducting training and predicting on each network flow component according to corresponding flow characteristics in different network flow modules, and respectively outputting prediction results; finally, conducting stack processing on all the prediction results, and determining the predication results of the to-be-predicted network flow. Consequently, on the condition of conducting decomposition and reconfiguration on the network flow according to the multiscale frequency and respectively conducting network flow prediction according to the respective flow characteristic of each network flow component, the effect of the multiscale frequency on accuracy of the prediction results can be effectively reduced, and the problem that network transmission accuracy requirements can not be met due to the fact that accurate prediction cannot be conducted on the network flow in the multiscale frequency range caused by complicated non-linear relationship of the network flow.
Description
Technical field
The present invention relates to communication technical field, particularly a kind of network flow prediction method and equipment.
Background technology
Along with the expansion of computer network scale and the continuous growth of class of business, P2P(Peer-to-Peer, point-to-point) network obtained develop rapidly as a kind of brand-new Internet technology.P2P uses develops the life ﹠ amusement that greatly facilitates on the one hand people rapidly, not only caused but then the huge consumption of the network bandwidth, even cause network congestion, greatly reduce network performance, deteriorated network service quality has hindered carrying out and crucial use universal of proper network business.
For the Strengthens network management, effectively improve network operation speed and utilance, predicting network flow becomes the treatment technology of an increasingly extensive application.
So-called prediction, be exactly to not yet occuring and present also indefinite information, carry out in advance estimation and supposition according to past and present information, namely under certain Mathematical Modeling to development trend, direction and the possible state of following a period of time internal information make reasonably, deduction in the permissible error scope.
And in the predicting network flow process, the object of predicting is exactly the Change and Development trend of network traffics.Traditional time series predicting model can have reasonable performance for steady network traffics sequence.
In realizing process of the present invention, the inventor finds to exist at least in prior art following problem:
The network traffics time series is the dynamical system of non-linear, a Multiple Time Scales conversion, has the characteristics such as obvious self-similarity, sudden and periodicity.Therefore, traditional time series predicting model is difficult to portray the non-linear relation of network traffics complexity.
Echo state network (Echo State Network, ESN) good approximation capability and precision of prediction have been shown in the time series to Noise not, precision of prediction is compared conventional model and is significantly increased, but few and comprise in the situation of noise for sample, the prediction effect of ESN network is unsatisfactory, and only can limited extent ground process multiple dimensioned frequency range, relatively poor to multiple dimensioned seasonal effect in time series forecasting problem adaptability, be difficult to satisfy the required precision of P2P volume forecasting.
Summary of the invention
The embodiment of the present invention provides a kind of network flow prediction method and equipment, solve in existing technical scheme the non-linear relation due to the network traffics complexity, and can not carry out Accurate Prediction to network traffics in multiple dimensioned frequency range, can't satisfy the problem of Internet Transmission required precision.
For achieving the above object, the embodiment of the present invention provides a kind of network flow prediction method on the one hand, comprises the following steps at least:
Network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generate a plurality of network traffics components;
The characteristic of network traffics component according to each, train prediction with each described network traffics component in corresponding Network Traffic Forecast Model respectively, exports the corresponding network traffics component of each network traffics component and predict the outcome;
The all-network flow component of exporting is predicted the outcome carry out overlap-add procedure, determine predicting the outcome of described network traffics to be predicted.
Preferably, described network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing are reconstructed respectively, generate a plurality of network traffics components, specifically comprise:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
Preferably, the described characteristic of network traffics component according to each respectively, each described network traffics component is trained prediction in corresponding Network Traffic Forecast Model, exports the corresponding network traffics component of each network traffics component and predict the outcome, specifically comprise:
The characteristic of network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component;
According to each input sample sequence, determine the corresponding next input sample of each network traffics component respectively;
Choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model;
Exporting the corresponding network traffics component of each network traffics component predicts the outcome.
Preferably, the characteristic of described network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component, specifically comprises:
The input dimension of the characteristic of network traffics component according to each and described network traffics to be predicted is respectively determined the input sample corresponding with described input dimension of each network traffics component;
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
Preferably, after the corresponding network traffics component of each network traffics component of described output predicts the outcome, also comprise:
Input sample sequence according to each remaining part when carrying out training sample to selection, predict the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
On the other hand, the embodiment of the present invention also provides a kind of predicting network flow equipment, comprises at least:
Decomposed and reconstituted module is used for network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generates a plurality of network traffics components;
The training prediction module, be used for the characteristic of network traffics component according to each respectively, each described network traffics component of described decomposed and reconstituted module institute reconstruct is trained prediction in corresponding Network Traffic Forecast Model, export the corresponding network traffics component of each network traffics component and predict the outcome;
The overlap-add procedure module is used for the all-network flow component that described training prediction module is exported is predicted the outcome and carries out overlap-add procedure, determines predicting the outcome of described network traffics to be predicted.
Preferably, described decomposed and reconstituted module specifically is used for:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
Preferably, described training prediction module specifically is used for:
The characteristic of network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component of described decomposed and reconstituted module institute reconstruct;
Respectively according to each input sample sequence, determine the corresponding next input sample of each network traffics component of described decomposed and reconstituted module institute reconstruct;
Choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model;
The corresponding network traffics component of each network traffics component of exporting the reconstruct of described decomposed and reconstituted module institute predicts the outcome.
Preferably, described training prediction module specifically is used for:
According to the input dimension of the characteristic of each described network traffics component of described decomposed and reconstituted module institute reconstruct and described network traffics to be predicted, determine the input sample corresponding with described input dimension of each network traffics component respectively;
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
Preferably, the reconstruct of described decomposed and reconstituted module institute, also be used for:
After the corresponding network traffics component of each network traffics component of output predicted the outcome, input sample sequence according to each remaining part when carrying out training sample to selection, predicted the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
Compared with prior art, the technical scheme that proposes of the embodiment of the present invention has the following advantages:
by using the technical scheme of the embodiment of the present invention, under the application scenarios of multiple dimensioned frequency, at first to network traffics to be predicted are carried out wavelet decomposition reconstruct, generate a plurality of network traffics components, then, respectively to each network traffics component, according to corresponding discharge characteristic, train prediction in different Model of network traffic, difference prediction of output result, at last, overlap-add procedure is carried out in all predicting the outcome, determine predicting the outcome of network traffics to be predicted, thereby, carrying out decomposed and reconstituted according to multiple dimensioned frequency network traffics, and carry out in the situation of volume forecasting according to frequency characteristic separately respectively, can more effectively reduce multiple dimensioned frequency for the impact of the accuracy that predicts the outcome, solution is due to the non-linear relation of network traffics complexity, and can not carry out Accurate Prediction to network traffics in multiple dimensioned frequency range, can't satisfy the problem of Internet Transmission required precision.
Description of drawings
Fig. 1 is the schematic flow sheet of a kind of network flow prediction method of proposing of the embodiment of the present invention;
Fig. 2 is the model schematic diagram based on the volume forecasting model of wavelet transformation and ESN that the embodiment of the present invention proposes;
Fig. 3 is the structural representation of a kind of predicting network flow equipment of embodiment of the present invention proposition.
Embodiment
As stated in the Background Art, be the dynamical system of non-linear, a Multiple Time Scales conversion due to the network traffics time series, have the characteristics such as obvious self-similarity, sudden and periodicity.So, in the application scenarios of reality, often because the complexity that network traffics change, especially in the situation that network traffics itself have multiple dimensioned frequency, affect the accuracy of predicting network flow, can not satisfy Internet Transmission, for the more and more higher requirement of predicting network flow accuracy.
In order to overcome such defective, the embodiment of the present invention has proposed a kind of network flow prediction method, in the situation that network traffics itself have multiple dimensioned frequency, carry out decomposed and reconstituted according to multiple dimensioned frequency network traffics, carry out after volume forecasting, result being superposeed according to frequency characteristic separately respectively, thereby, can more effectively reduce multiple dimensioned frequency for the impact of the accuracy that predicts the outcome.
Below in conjunction with the accompanying drawing in the present invention, the technical scheme in the present invention is clearly and completely described, obviously, described embodiment is only part embodiment of the present invention, rather than whole embodiment.Based on the embodiment in the present invention, those of ordinary skills belong to the scope of protection of the invention not making all other embodiment that obtain under the creative work prerequisite.
As shown in Figure 1, the schematic flow sheet of a kind of network flow prediction method that proposes for the embodiment of the present invention, the method specifically comprises the following steps:
Step S101, network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generate a plurality of network traffics components.
In concrete application scenarios, the processing of this step specifically comprises:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
By such processing, network traffics to be predicted can be decomposed according to multiple dimensioned frequency, the network traffics component after decomposition will no longer have multiple dimensioned frequency, and the complexity of carrying out volume forecasting also greatly reduces.
Step S102, the characteristic of network traffics component according to each are respectively trained prediction with each described network traffics component in corresponding Network Traffic Forecast Model, export the corresponding network traffics component of each network traffics component and predict the outcome.
Concrete, the processing of this step comprises the following steps:
(1) characteristic of network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component, is described as follows:
The input dimension of the characteristic of network traffics component according to each and described network traffics to be predicted is respectively determined the input sample corresponding with described input dimension of each network traffics component.
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
(2) respectively according to each input sample sequence, determine the corresponding next input sample of each network traffics component.
(3) choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model.
(4) exporting the corresponding network traffics component of each network traffics component predicts the outcome.
By above-mentioned processing, carry out volume forecasting for each network traffics component respectively, exported accordingly and predicted the outcome.
Further, in order to ensure the accuracy of volume forecasting, after completing above-mentioned volume forecasting, can also comprise corresponding result verification mechanism, be described as follows:
Input sample sequence according to each remaining part when carrying out training sample to selection, predict the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
if do not meet preset standard and predict the outcome, can carry out subsequent treatment according to corresponding processing policy, for example, can be to being the corresponding volume forecasting of the heavy new initiation of network traffics that meets preset standard, also can tap into the correction that row predicts the outcome according to corresponding standard straight, perhaps, the directly prompting of prediction of output result verification failure, there are concrete operator or network management side to determine the mode of subsequent treatment, certainly, in practical operation, subsequent treatment mode after such authentication failed can also have a lot, specifically can select according to actual needs, such variation does not affect protection scope of the present invention.
On the other hand, above-mentioned preset standard can be set as the deviation ratio that predicts the outcome, or the Verification standards such as matching degree with sample sequence of predicting the outcome, its purpose is to verify the whole matching situation that predicts the outcome with sample sequence, it is better to mate, and proves that the accuracy that predicts the outcome accordingly is higher.
In above-mentioned processing procedure, why, when carrying out training sample to selection, only select a part of sample sequence, process in order to carry out the follow-up checking that predicts the outcome by the remaining sample sequence that has neither part nor lot in training exactly.
Certainly; the processing scheme that above-mentioned proof procedure just proposes in order further to guarantee to predict the outcome accuracy; can select according to actual needs whether to carry out such checking processes; if do not need checking; when carrying out training sample to selection; can directly train prediction with whole sample sequences, such variation does not affect protection scope of the present invention yet.
Step S103, the all-network flow component of exporting is predicted the outcome carry out overlap-add procedure, determine predicting the outcome of described network traffics to be predicted.
In the application scenarios of reality, can carry out overlap-add procedure by modes such as linear superposition, certainly, this is relevant to decomposed and reconstituted mode in step S101, and according to the difference of is olation, concrete stack strategy also can be adjusted accordingly.
Compared with prior art, the technical scheme that proposes of the embodiment of the present invention has the following advantages:
by using the technical scheme of the embodiment of the present invention, under the application scenarios of multiple dimensioned frequency, at first to network traffics to be predicted are carried out wavelet decomposition reconstruct, generate a plurality of network traffics components, then, respectively to each network traffics component, according to corresponding discharge characteristic, train prediction in different Model of network traffic, difference prediction of output result, at last, overlap-add procedure is carried out in all predicting the outcome, determine predicting the outcome of network traffics to be predicted, thereby, in the situation that network traffics are carried out decomposed and reconstituted and are carried out volume forecasting according to frequency characteristic separately respectively according to multiple dimensioned frequency, can more effectively reduce multiple dimensioned frequency for the impact of the accuracy that predicts the outcome, solution is due to the non-linear relation of network traffics complexity, and can not carry out Accurate Prediction to network traffics in multiple dimensioned frequency range, can't satisfy the problem of Internet Transmission required precision.
Below, in conjunction with concrete application scenarios, the technical scheme that the embodiment of the present invention is proposed describes.
The ESN network has shown good approximation capability and precision of prediction in the time series to Noise not, precision of prediction is compared conventional model and is significantly increased, but few and comprise in the situation of noise for sample, the prediction effect of ESN is unsatisfactory, and only can limited extent ground process multiple dimensioned frequency range, relatively poor to multiple dimensioned seasonal effect in time series forecasting problem adaptability, be difficult to satisfy the required precision of P2P volume forecasting.
For this problem, the embodiment of the present invention has proposed the volume forecasting model based on wavelet transformation and ESN, and its concrete model schematic diagram as shown in Figure 2.
Based on above-mentioned model, the embodiment of the present invention has proposed the network flow prediction method under a kind of concrete application scenarios, and corresponding forecasting process can divide following 3 steps.
(1) wavelet decomposition and reconstruct.
At first, utilize the Mallat algorithm, adopt the fast discrete dyadic wavelet transform that P2P flow time series is decomposed, obtain approximate part and detail section.
By original P2P flow time series y (n) being carried out decomposition and the reconstruct that yardstick is M, can obtain each P2P flow low-and high-frequency component d
jAnd a
M
Wherein, original P2P flow time series y (n) can approximate representation be:
In concrete application scenarios, may there be minimum a part of error in above-mentioned reconstruction result with original P2P flow time series, but the requirement of concrete processing accuracy is satisfied in such processing, can ignore.
Above-mentioned specific algorithm and is olation are a kind of concrete example that the embodiment of the present invention provides for the corresponding technical scheme of clearer explanation, in the practical application scene, in the situation that guarantee same treatment effect, can adjust accordingly.
(2) network traffics component prediction
K is the input dimension of network, for each flow component a
MAnd d
j, the input sample sequence of tectonic network is as follows respectively:
A
M(n)=[a
M(n),a
M(n-1),…,a
M(n-K+1)]
T
D
j(n)=[d
j(n),d
j(n-1),…,d
j(n-K+1)]
T
Based on above-mentioned sample sequence, corresponding output is respectively a
M(n+1) and d
j(n+1).
Choose A
M(n), D
j(n) separately a part of sequence is as the training list entries, and corresponding training sample is to being respectively { A
M(n), a
M(n+1) }, { D
j(n), d
j(n+1) }.
Further, after the ESN model training is completed, A
M(n), D
j(n) remaining part is tested prediction effect as forecasting sequence, if test successfully, obtains predicting the outcome of each network traffics component:
(3) determine the predicting network flow result.
Each component in previous step is predicted the outcome
With
Linear superposition can obtain predicting the outcome to former data flow y (n)
Specific as follows:
Compared with prior art, the technical scheme that proposes of the embodiment of the present invention has the following advantages:
by using the technical scheme of the embodiment of the present invention, under the application scenarios of multiple dimensioned frequency, at first to network traffics to be predicted are carried out wavelet decomposition reconstruct, generate a plurality of network traffics components, then, respectively to each network traffics component, according to corresponding discharge characteristic, train prediction in different Model of network traffic, difference prediction of output result, at last, overlap-add procedure is carried out in all predicting the outcome, determine predicting the outcome of network traffics to be predicted, thereby, in the situation that network traffics are carried out decomposed and reconstituted and are carried out volume forecasting according to frequency characteristic separately respectively according to multiple dimensioned frequency, can more effectively reduce multiple dimensioned frequency for the impact of the accuracy that predicts the outcome, solution is due to the non-linear relation of network traffics complexity, and can not carry out Accurate Prediction to network traffics in multiple dimensioned frequency range, can't satisfy the problem of Internet Transmission required precision.
In order to realize the technical scheme of the embodiment of the present invention, the embodiment of the present invention also provides a kind of predicting network flow equipment, and its structural representation comprises as shown in Figure 3 at least:
Decomposed and reconstituted module 31 is used for network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generates a plurality of network traffics components;
Overlap-add procedure module 33 is used for the all-network flow component that described training prediction module 32 is exported is predicted the outcome and carries out overlap-add procedure, determines predicting the outcome of described network traffics to be predicted.
Preferably, described decomposed and reconstituted module 31 specifically is used for:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
Preferably, described training prediction module 32 specifically is used for:
The characteristic of network traffics component according to each, the corresponding input sample sequence of each network traffics component of definite 31 reconstruct of described decomposed and reconstituted module respectively;
According to each input sample sequence, determine the corresponding next input sample of each network traffics component of 31 reconstruct of described decomposed and reconstituted module respectively;
Choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model;
The corresponding network traffics component of each network traffics component of exporting 31 reconstruct of described decomposed and reconstituted module predicts the outcome.
Preferably, described training prediction module 32 specifically is used for:
According to the input dimension of the characteristic of each described network traffics component of 31 reconstruct of described decomposed and reconstituted module and described network traffics to be predicted, determine the input sample corresponding with described input dimension of each network traffics component respectively;
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
Preferably, 31 reconstruct of described decomposed and reconstituted module, also be used for:
After the corresponding network traffics component of each network traffics component of output predicted the outcome, input sample sequence according to each remaining part when carrying out training sample to selection, predicted the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
Compared with prior art, the technical scheme that proposes of the embodiment of the present invention has the following advantages:
by using the technical scheme of the embodiment of the present invention, consider the amphicheirality who disturbs between wireless communication system, when needs carry out the sensory perceptual system introducing, only in the situation that can guarantee simultaneously the normal operation of authoring system and sensory perceptual system in corresponding frequency and adjacent other frequencies thereof, just corresponding frequency can be defined as available frequency, thereby, guarantee determined available frequency band in the situation that guarantee that the normal operation of authoring system is not subjected to the interference of sensory perceptual system, guarantee that simultaneously the new sensory perceptual system of introducing also can not be subjected to the interference of authoring system and work, improve the communication quality of wireless communication system.
Through the above description of the embodiments, those skilled in the art can be well understood to the embodiment of the present invention and can realize by hardware, also can realize by the mode that software adds necessary general hardware platform.Based on such understanding, the technical scheme of the embodiment of the present invention can embody with the form of software product, it (can be CD-ROM that this software product can be stored in a non-volatile memory medium, USB flash disk, portable hard drive etc.) in, comprise some instructions with so that computer equipment (can be personal computer, server, or network equipment etc.) each implements the described method of scene to carry out the embodiment of the present invention.
It will be appreciated by those skilled in the art that accompanying drawing is a preferred schematic diagram of implementing scene, the module in accompanying drawing or flow process might not be that the enforcement embodiment of the present invention is necessary.
It will be appreciated by those skilled in the art that the module in the device of implementing in scene can be distributed in the device of implementing scene according to implementing scene description, also can carry out respective change and be arranged in the one or more devices that are different from this enforcement scene.The module of above-mentioned enforcement scene can be merged into a module, also can further split into a plurality of submodules.
The invention described above embodiment sequence number does not represent just to description the quality of implementing scene.
Above disclosed be only the several concrete enforcement scene of the embodiment of the present invention, still, the embodiment of the present invention is not limited thereto, the changes that any person skilled in the art can think of all should fall into the traffic limits scope of the embodiment of the present invention.
Claims (10)
1. a network flow prediction method, is characterized in that, comprises the following steps at least:
Network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generate a plurality of network traffics components;
The characteristic of network traffics component according to each, train prediction with each described network traffics component in corresponding Network Traffic Forecast Model respectively, exports the corresponding network traffics component of each network traffics component and predict the outcome;
The all-network flow component of exporting is predicted the outcome carry out overlap-add procedure, determine predicting the outcome of described network traffics to be predicted.
2. the method for claim 1, is characterized in that, described network traffics to be predicted carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generates a plurality of network traffics components, specifically comprises:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
3. the method for claim 1, it is characterized in that, the described characteristic of network traffics component according to each respectively, each described network traffics component is trained prediction in corresponding Network Traffic Forecast Model, export the corresponding network traffics component of each network traffics component and predict the outcome, specifically comprise:
The characteristic of network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component;
According to each input sample sequence, determine the corresponding next input sample of each network traffics component respectively;
Choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model;
Exporting the corresponding network traffics component of each network traffics component predicts the outcome.
4. method as claimed in claim 3, is characterized in that, the characteristic of described network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component, specifically comprises:
The input dimension of the characteristic of network traffics component according to each and described network traffics to be predicted is respectively determined the input sample corresponding with described input dimension of each network traffics component;
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
5. method as claimed in claim 3, is characterized in that, after the corresponding network traffics component of each network traffics component of described output predicts the outcome, also comprises:
Input sample sequence according to each remaining part when carrying out training sample to selection, predict the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
6. a predicting network flow equipment, is characterized in that, comprises at least:
Decomposed and reconstituted module is used for network traffics to be predicted are carried out wavelet decomposition, and the result after decomposing is reconstructed respectively, generates a plurality of network traffics components;
The training prediction module, be used for the characteristic of network traffics component according to each respectively, each described network traffics component of described decomposed and reconstituted module institute reconstruct is trained prediction in corresponding Network Traffic Forecast Model, export the corresponding network traffics component of each network traffics component and predict the outcome;
The overlap-add procedure module is used for the all-network flow component that described training prediction module is exported is predicted the outcome and carries out overlap-add procedure, determines predicting the outcome of described network traffics to be predicted.
7. equipment as claimed in claim 6, is characterized in that, described decomposed and reconstituted module specifically is used for:
Algorithm and decomposition scale according to default carry out wavelet decomposition to network traffics to be predicted, obtain approximate part and detail section;
Each decomposition result is reconstructed respectively, generates high fdrequency component and the low frequency component corresponding with described decomposition scale.
8. equipment as claimed in claim 6, is characterized in that, described training prediction module specifically is used for:
The characteristic of network traffics component according to each is determined respectively the corresponding input sample sequence of each network traffics component of described decomposed and reconstituted module institute reconstruct;
Respectively according to each input sample sequence, determine the corresponding next input sample of each network traffics component of described decomposed and reconstituted module institute reconstruct;
Choose respectively the part of each described input sample sequence, form training sample pair with corresponding next one input sample, training prediction in corresponding Network Traffic Forecast Model;
The corresponding network traffics component of each network traffics component of exporting the reconstruct of described decomposed and reconstituted module institute predicts the outcome.
9. equipment as claimed in claim 8, is characterized in that, described training prediction module specifically is used for:
According to the input dimension of the characteristic of each described network traffics component of described decomposed and reconstituted module institute reconstruct and described network traffics to be predicted, determine the input sample corresponding with described input dimension of each network traffics component respectively;
According to the input sample corresponding with described input dimension of each network traffics component, determine the corresponding input sample sequence of each network traffics component.
10. equipment as claimed in claim 8, is characterized in that, the reconstruct of described decomposed and reconstituted module institute, also be used for:
After the corresponding network traffics component of each network traffics component of output predicted the outcome, input sample sequence according to each remaining part when carrying out training sample to selection, predicted the outcome to corresponding network traffics component and test respectively;
If test result meets preset standard, export the corresponding network traffics component of corresponding network flow component and predict the outcome.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310008445XA CN103095496A (en) | 2013-01-10 | 2013-01-10 | Prediction method and device for network flow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310008445XA CN103095496A (en) | 2013-01-10 | 2013-01-10 | Prediction method and device for network flow |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103095496A true CN103095496A (en) | 2013-05-08 |
Family
ID=48207656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310008445XA Pending CN103095496A (en) | 2013-01-10 | 2013-01-10 | Prediction method and device for network flow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103095496A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104168131A (en) * | 2014-06-05 | 2014-11-26 | 国家电网公司 | Flow generation method of power dispatching exchange network based on multicast communication |
| CN104506378A (en) * | 2014-12-03 | 2015-04-08 | 上海华为技术有限公司 | Data flow prediction device and method |
| CN106789297A (en) * | 2016-12-29 | 2017-05-31 | 淮海工学院 | Predicting network flow system and its method for predicting based on neutral net |
| CN107026763A (en) * | 2017-06-02 | 2017-08-08 | 广东电网有限责任公司中山供电局 | A kind of data communication network method for predicting decomposed based on flow |
| CN107070683A (en) * | 2016-12-12 | 2017-08-18 | 国网北京市电力公司 | The method and apparatus of data prediction |
| CN107113910A (en) * | 2015-01-07 | 2017-08-29 | 华为技术有限公司 | The method and apparatus for building network |
| CN108901033A (en) * | 2018-06-20 | 2018-11-27 | 南京邮电大学 | Base station method for predicting based on echo state network |
| CN109640351A (en) * | 2019-01-25 | 2019-04-16 | 南京邮电大学 | A kind of unified prediction of base station flow |
| CN110839253A (en) * | 2019-11-08 | 2020-02-25 | 西北工业大学青岛研究院 | Method for determining wireless grid network flow |
| CN111884854A (en) * | 2020-07-29 | 2020-11-03 | 中国人民解放军空军工程大学 | Virtual network traffic migration method based on multi-mode hybrid prediction |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050135266A1 (en) * | 2003-12-22 | 2005-06-23 | Gwoboa Horng | Method of detecting distributed denial of service based on grey theory |
| CN101729315A (en) * | 2009-12-24 | 2010-06-09 | 北京邮电大学 | Network flow-predicting method and device based on wavelet package decomposition and fuzzy neural network |
-
2013
- 2013-01-10 CN CN201310008445XA patent/CN103095496A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050135266A1 (en) * | 2003-12-22 | 2005-06-23 | Gwoboa Horng | Method of detecting distributed denial of service based on grey theory |
| CN101729315A (en) * | 2009-12-24 | 2010-06-09 | 北京邮电大学 | Network flow-predicting method and device based on wavelet package decomposition and fuzzy neural network |
Non-Patent Citations (3)
| Title |
|---|
| 冯华丽，刘渊: "小波分析和AR-LSSVM的网络流量预测", 《计算机工程与应用》 * |
| 时飞飞: "基于改进ESN的高炉煤气系统预测方法的研究", 《中国优秀硕士学位论文全文数据库，信息科技辑》 * |
| 王卓群，孙志国: "一种小波分解回声状态网络时间序列预测方法", 《电子测量与仪器学报》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104168131A (en) * | 2014-06-05 | 2014-11-26 | 国家电网公司 | Flow generation method of power dispatching exchange network based on multicast communication |
| CN104506378B (en) * | 2014-12-03 | 2019-01-18 | 上海华为技术有限公司 | A kind of device and method of prediction data flow |
| CN104506378A (en) * | 2014-12-03 | 2015-04-08 | 上海华为技术有限公司 | Data flow prediction device and method |
| CN107113910B (en) * | 2015-01-07 | 2020-01-10 | 华为技术有限公司 | Method and device for constructing network |
| CN107113910A (en) * | 2015-01-07 | 2017-08-29 | 华为技术有限公司 | The method and apparatus for building network |
| CN107070683A (en) * | 2016-12-12 | 2017-08-18 | 国网北京市电力公司 | The method and apparatus of data prediction |
| CN106789297A (en) * | 2016-12-29 | 2017-05-31 | 淮海工学院 | Predicting network flow system and its method for predicting based on neutral net |
| CN107026763B (en) * | 2017-06-02 | 2019-11-26 | 广东电网有限责任公司中山供电局 | A kind of data communication network method for predicting decomposed based on flow |
| CN107026763A (en) * | 2017-06-02 | 2017-08-08 | 广东电网有限责任公司中山供电局 | A kind of data communication network method for predicting decomposed based on flow |
| CN108901033A (en) * | 2018-06-20 | 2018-11-27 | 南京邮电大学 | Base station method for predicting based on echo state network |
| CN109640351A (en) * | 2019-01-25 | 2019-04-16 | 南京邮电大学 | A kind of unified prediction of base station flow |
| CN110839253A (en) * | 2019-11-08 | 2020-02-25 | 西北工业大学青岛研究院 | Method for determining wireless grid network flow |
| CN111884854A (en) * | 2020-07-29 | 2020-11-03 | 中国人民解放军空军工程大学 | Virtual network traffic migration method based on multi-mode hybrid prediction |
| CN111884854B (en) * | 2020-07-29 | 2022-09-02 | 中国人民解放军空军工程大学 | Virtual network traffic migration method based on multi-mode hybrid prediction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103095496A (en) | Prediction method and device for network flow | |
| CN102741921B (en) | Improved subband block based harmonic transposition | |
| US11748595B2 (en) | Convolution acceleration operation method and apparatus, storage medium and terminal device | |
| CN105718848A (en) | Quality evaluation method and apparatus of fingerprint images | |
| CN104541327A (en) | Methods and systems for efficient recovery of high frequency audio content | |
| Ghasemzadeh et al. | Universal audio steganalysis based on calibration and reversed frequency resolution of human auditory system | |
| CN111597971B (en) | Urban rail transit short-term arrival passenger flow prediction method | |
| Zuo et al. | Artificial Intelligence Prediction and Decision Evaluation Model Based on Deep Learning | |
| CN109697083B (en) | Fixed-point acceleration method and device for data, electronic equipment and storage medium | |
| CN111242210A (en) | Short-term load prediction method based on improved Shapley value model | |
| CN114676629A (en) | Multi-means composite light weight processing method for modulation type recognition model | |
| US20240113519A1 (en) | Method and device for forecasting electric load, and electronic device | |
| CN105096957B (en) | Process the method and apparatus of signal | |
| US20180084600A1 (en) | Link Packing in mmWave Networks | |
| CN106056254B (en) | Wind power prediction error simulation method considering influence of output level | |
| CN113823312B (en) | Speech enhancement model generation method and device, and speech enhancement method and device | |
| RU2495504C1 (en) | Method of reducing transmission rate of linear prediction low bit rate voders | |
| CN105283915A (en) | Digital-watermark embedding device, digital-watermark detection device, digital-watermark embedding method, digital-watermark detection method, digital-watermark embedding program, and digital-watermark detection program | |
| US10002615B2 (en) | Inter-channel level difference processing method and apparatus | |
| CN110989363B (en) | Electric energy quality control method and device based on deep learning | |
| CN113179354A (en) | Sound signal processing method and device and electronic equipment | |
| CN115348179B (en) | Network traffic boundary prediction method, device, equipment and medium | |
| US20230395086A1 (en) | Method and apparatus for processing of audio using a neural network | |
| CN110991643B (en) | Model deployment method and device, electronic equipment and storage medium | |
| CN111063368B (en) | Method, apparatus, medium, and device for estimating noise in audio signal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130508 |
|
| RJ01 | Rejection of invention patent application after publication |