CN104506378B - A kind of device and method of prediction data flow - Google Patents
A kind of device and method of prediction data flow Download PDFInfo
- Publication number
- CN104506378B CN104506378B CN201410727881.7A CN201410727881A CN104506378B CN 104506378 B CN104506378 B CN 104506378B CN 201410727881 A CN201410727881 A CN 201410727881A CN 104506378 B CN104506378 B CN 104506378B
- Authority
- CN
- China
- Prior art keywords
- data
- signal
- traffic
- signals
- traffic signals
- 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
Abstract
This application provides a kind of device and methods of prediction data flow.Device includes: signal acquisition module；Determining module, for determining that the corresponding waveform of data-traffic signals has self-similarity according to data-traffic signals；Signal processing module, for carrying out decomposition and reconstruction to waveform using wavelet analysis technology；Determining module is also used to the signal by the stationarity at least one approximation signal and multiple detail signals less than the first preset threshold, is determined as primary sources flow signal；Stationarity is greater than or equal to the signal of the first preset threshold, is determined as secondary sources flow signal；Computing module, for being predicted using compressed sensing model primary sources flow signal；Secondary sources flow signal is predicted using linear model；Synthesize first kind prediction result and the second class prediction result.Using the device or method of the application, the precision of prediction for data traffic can be improved.
Description
Technical field
This application involves the communications fields, more particularly to a kind of device and method of prediction data flow.
Background technique
With the continuous development of the communication technology, carry out communicating generated data traffic between the electronic equipments such as mobile terminal
Also increasing.
It in the prior art,, can be according to reality during transmitting data stream amount in order to reasonable distribution Internet resources
Data traffic information, the data volume for needing to transmit in following a period of time is predicted.According to prediction result, network is provided
Source is allocated, and can more efficiently be carried out data transmission.
But in the prior art, usually main to be predicted using a certain mode for the prediction of data traffic.And
This kind of mode is usually only for data traffic prediction with higher caused by a certain or certain several certain types of business
Precision, and for data traffic caused by other business, often precision of prediction is lower.
Summary of the invention
The purpose of the application is to provide a kind of device and method of prediction data flow, can be according to the data traffic of business
Waveform characteristic, data traffic is predicted using the prediction mode that the waveform characteristic with data traffic matches, to mention
Precision of prediction of the height for data traffic.
To achieve the above object, this application provides following schemes:
The possible implementation of according to a first aspect of the present application the first, the application provide a kind of prediction data flow
Device, comprising:
Signal acquisition module, for obtaining the data-traffic signals in predetermined time period；
Determining module, for determining that the corresponding waveform of the data-traffic signals has according to the data-traffic signals
Self-similarity；
Signal processing module, for carrying out decomposition and reconstruction to the waveform using wavelet analysis technology, after obtaining reconstruct
Data-traffic signals；Data-traffic signals after the reconstruct include at least one approximation signal and multiple detail signals；
The determining module is also used to the stationarity at least one described approximation signal and multiple detail signals is small
In the signal of the first preset threshold, it is determined as primary sources flow signal；
It is pre- that stationarity at least one described approximation signal and multiple detail signals is greater than or equal to described first
If the signal of threshold value is determined as secondary sources flow signal；
Computing module obtains for predicting using compressed sensing model the primary sources flow signal
A kind of prediction result；
The secondary sources flow signal is predicted using linear model, obtains the second class prediction result；
Synthesize the first kind prediction result and the second class prediction result.
The possible implementation of second with reference to first aspect, the signal acquisition module are specifically used for:
It is sampled according to data traffic of the prefixed time interval to generation, each sampling being sequentially arranged
The corresponding data traffic of point；
From the corresponding data traffic of each sampled point being sequentially arranged, intercept in predetermined time period
The corresponding data traffic of sampled point.
The third possible implementation with reference to first aspect, the determining module are specifically used for:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；
Determine that the value of the Hurst Exponent is greater than the second preset threshold.
A kind of specific implementation of the possible implementation of second with reference to first aspect, the determining module, tool
Body is used for:
According to formulaCalculate at least one described approximation signal and multiple details
The sample autocorrelation function of signal；
It willSignal be determined as the primary sources flow signal；
Wherein, XiFor the corresponding data traffic of ith sample point in predetermined time period；EXFor the mean value of X；N is described
The number of sampled point in predetermined time period；K=1,2,3 ... K；θ is first preset threshold.
The 4th kind of possible implementation with reference to first aspect, the determining module are also used to:
Determine that the corresponding waveform of the data-traffic signals does not have self-similarity；
The computing module, is also used to:
When the determining module determines that the corresponding waveform of the data-traffic signals does not have self-similarity, using linear
Data-traffic signals described in model prediction.
The possible implementation of according to a second aspect of the present application the first, the application provide a kind of prediction data flow
Method, comprising:
Obtain the data-traffic signals in predetermined time period；
According to the data-traffic signals, determine that the corresponding waveform of the data-traffic signals has self-similarity；
Decomposition and reconstruction is carried out to the waveform using wavelet analysis technology, the data-traffic signals after being reconstructed；Institute
Data-traffic signals after stating reconstruct include at least one approximation signal and multiple detail signals；
Letter by the stationarity at least one described approximation signal and multiple detail signals less than the first preset threshold
Number, it is determined as primary sources flow signal；
It is pre- that stationarity at least one described approximation signal and multiple detail signals is greater than or equal to described first
If the signal of threshold value is determined as secondary sources flow signal,
The primary sources flow signal is predicted using compressed sensing model, obtains first kind prediction result；
The secondary sources flow signal is predicted using linear model, obtains the second class prediction result；
Synthesize the first kind prediction result and the second class prediction result.
In conjunction with second of possible implementation of second aspect, the data traffic letter obtained in predetermined time period
Number, it specifically includes:
It is sampled according to data traffic of the prefixed time interval to generation, each sampling being sequentially arranged
The corresponding data traffic of point；
From the corresponding data traffic of each sampled point being sequentially arranged, intercept in predetermined time period
The corresponding data traffic of sampled point.
In conjunction with the third possible implementation of second aspect, the corresponding waveform of the determination data-traffic signals
With self-similarity, specifically include:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；
Determine that the value of the Hurst Exponent is greater than the second preset threshold.
In conjunction with a kind of concrete implementation mode of second of possible implementation of second aspect, the general is described at least
Stationarity in one approximation signal and multiple detail signals is determined as primary sources less than the signal of the first preset threshold
Flow signal specifically includes:
According to formulaCalculate at least one described approximation signal and multiple details
The sample autocorrelation function of signal；
It willSignal be determined as the primary sources flow signal；
Wherein, XiFor the corresponding data traffic of ith sample point in predetermined time period；EXFor the mean value of X；N is described
The number of sampled point in predetermined time period；K=1,2,3 ... K；θ is first preset threshold.
In conjunction with the 4th kind of possible implementation of second aspect, further includes:
Determine that the corresponding waveform of the data-traffic signals does not have self-similarity；
Using data-traffic signals described in Linear Model for Prediction.
According to specific embodiment provided by the present application, this application discloses following technical effects:
The device or method of prediction data flow disclosed in the present application pass through the corresponding wave of the determination data-traffic signals
Shape has self-similarity, carries out decomposition and reconstruction to the waveform using wavelet analysis technology；According to stationarity to the reconstruct
Data-traffic signals afterwards are classified；Signal lower for stationarity, is predicted using compressed sensing model；For flat
The higher signal of stability, is predicted using linear model；Synthesize each prediction result；It can be according to the data traffic of business
Waveform characteristic predicts data traffic using the prediction mode that the waveform characteristic with data traffic matches, to improve
For the precision of prediction of data traffic.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the application
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is the schematic diagram of the application scenarios of the method for the prediction data flow of the application；
Fig. 2 is a kind of structure chart of the device of prediction data flow of the application；
Fig. 3 is a kind of flow chart of the method for prediction data flow of the application；
Fig. 4 is the flow chart of the method for another prediction data flow of the application；
Fig. 5 is the structure chart of the calculate node of the application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall in the protection scope of this application.
In order to make the above objects, features, and advantages of the present application more apparent, with reference to the accompanying drawing and it is specific real
Applying mode, the present application will be further described in detail.
Fig. 1 is the schematic diagram of the application scenarios of the method for the prediction data flow of the application.Application scenarios in Fig. 1 are
Example does not include all application scenarios.As shown in Figure 1, including: in the network architecture in the scene
Gateway 10, first community 101, second community 102 and third cell 103.It include at least one in each cell
Base station and user terminal.It can be in communication with each other between the base station of each cell and between each base station and gateway 10.The network rack
Structure can be cellular network.The gateway 10 can be located at core net, and base station can be located at access net.
The method and device of prediction data flow of the invention, can apply at gateway 10, can also apply in base station
Place.
Fig. 2 is a kind of structure chart of the device of prediction data flow of the application.As shown in Fig. 2, the apparatus may include:
Signal acquisition module 201, for obtaining the data-traffic signals in predetermined time period；
The data-traffic signals can indicate the data volume in certain time Jing Guo network transmission.The preset time
It is available in length to arrive multiple data-traffic signals.The multiple data-traffic signals have sequencing in time.
Determining module 202, for determining the corresponding waveform tool of the data-traffic signals according to the data-traffic signals
There is self-similarity；
Popular says, the object of self similarity is intimate or really and it a part of similar.If saying curve self phase
Seemingly, i.e., the curve of every part has a fritter similar with it.It is the important speciality of point shape that self is similar.
Usually the waveform stationarity with self-similarity is poor, and the waveform for not having self-similarity (i.e. non-self similarity) is steady
Property it is higher, can be predicted using linear model.
Signal processing module 203 is reconstructed for carrying out decomposition and reconstruction to the waveform using wavelet analysis technology
Data-traffic signals afterwards；Data-traffic signals after the reconstruct include at least one approximation signal and multiple details letter
Number；
Ha Er (Haar) wavelet analysis technology and mallat algorithm can be used, the waveform is decomposed at least one
Approximation signal and multiple detail signals.In practical application, the number of the detail signal after decomposition, can according to actual needs into
Row selection setting.For example, the waveform can be decomposed at least one approximation signal and three detail signals.
The determining module 202, being also used to will be steady at least one described approximation signal and multiple detail signals
Property is determined as primary sources flow signal less than the signal of the first preset threshold；
It is pre- that stationarity at least one described approximation signal and multiple detail signals is greater than or equal to described first
If the signal of threshold value is determined as secondary sources flow signal.
About stationarity, it will be understood that: assuming that time series X={ Xt, t=1,2 ... inner each element
It is that the random process for obeying some probability distribution by one generates at random, if X meets the following conditions:
1, mean value E (X)=μ is the constant unrelated with time t；
2, variance var (X)=σ2It is the constant unrelated with time t；
3, covariance cov (Xt,Xt+k)=γkIt is only with time interval k in relation to but with the unrelated constant of time t.
So, time series X is stable.
By define Check-Out Time sequence stationarity when, it is desirable that series infinite is long, is difficult to realize in reality.In general,
What is acquired in practical applications is the sample sequence of limited length.Sample is judged by sample autocorrelation function in the embodiment of the present invention
The stationarity of this sequence.Below about the detailed description of auto-correlation function.
It can judge approximation signal and each detail signal judge whether the stationarity of each signal is small respectively
In the first preset threshold.When the stationarity of signal is less than preset threshold, it is possible to determine that the signal does not have stationarity；When signal
Stationarity is greater than or equal to the first preset threshold, it can be determined that the signal has stationarity.
Computing module 204 is obtained for being predicted using compressed sensing model the primary sources flow signal
First kind prediction result；
The secondary sources flow signal is predicted using linear model, obtains the second class prediction result；
Synthesize the first kind prediction result and the second class prediction result.
Compressed sensing (Compressed sensing), also referred to as compression sampling (Compressive sampling),
Sparse sampling (Sparse sampling), compression sensing.Compressed sensing is a new sampling theory, passes through exploitation signal
Sparse characteristic can obtain the discrete sample of signal with stochastical sampling, then under conditions of being much smaller than Nyquist sample rate
Pass through non-linear algorithm for reconstructing reconstruction signal.
In general, a part in data-traffic signals after the reconstruct belongs to primary sources flow signal, another portion
Belong to secondary sources flow signal.It is available for each signal in the data-traffic signals after the reconstruct
One prediction result, the prediction result can indicate the partial data amount for needing to transmit in future time instance network.The synthesis, can
To be to be added each prediction result.Available total prediction result after synthesis, total prediction result can indicate future
The total amount of data for needing to transmit in moment network.According to the total amount of data for needing to transmit in the future time instance network of prediction, network
In relevant device Internet resources can be allocated, to more efficiently carry out data transmission.
In conclusion having by the corresponding waveform of the determination data-traffic signals from phase in embodiment shown in Fig. 2
Like property, decomposition and reconstruction is carried out to the waveform using wavelet analysis technology；According to stationarity to the data flow after the reconstruct
Amount signal is classified；Signal lower for stationarity, is predicted using compressed sensing model；It is higher for stationarity
Signal is predicted using linear model；Synthesize each prediction result；Can according to the waveform characteristic of the data traffic of business,
Data traffic is predicted using the prediction mode that the waveform characteristic with data traffic matches, to improve for data flow
The precision of prediction of amount.
In practical application, the signal acquisition module 201 specifically can be used for:
It is sampled according to data traffic of the prefixed time interval to generation, each sampling being sequentially arranged
The corresponding data traffic of point；
From the corresponding data traffic of each sampled point being sequentially arranged, intercept in predetermined time period
The corresponding data traffic of sampled point.
In practical application, the determining module 202 specifically can be used for:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；
Determine that the value of the Hurst Exponent is greater than the second preset threshold.
In practical application, the determining module 202 specifically be can be also used for:
According to formulaCalculate at least one described approximation signal and multiple details
The sample autocorrelation function of signal；
It willSignal be determined as the primary sources flow signal；
Wherein, XiFor the corresponding data traffic of ith sample point in predetermined time period；EXFor the mean value of X；N is described
The number of sampled point in predetermined time period；K=1,2,3 ... K；θ is first preset threshold.
In practical application, the determining module 202 be can be also used for:
Determine that the corresponding waveform of the data-traffic signals does not have self-similarity；
The computing module 204, can be also used for:
When the determining module determines that the corresponding waveform of the data-traffic signals does not have self-similarity, using linear
Data-traffic signals described in model prediction.
Present invention also provides a kind of methods of prediction data flow.
Fig. 3 is a kind of flow chart of the method for prediction data flow of the application.As shown in figure 3, this method may include:
Step 301: obtaining the data-traffic signals in predetermined time period；
The data-traffic signals can indicate the data volume in certain time Jing Guo network transmission.The preset time
It is available in length to arrive multiple data-traffic signals.The multiple data-traffic signals have sequencing in time.
Step 302: according to the data-traffic signals, determining that the corresponding waveform of the data-traffic signals has self similarity
Property；
Popular says, the object of self similarity is intimate or really and it a part of similar.If saying curve self phase
Seemingly, i.e., the curve of every part has a fritter similar with it.It is the important speciality of point shape that self is similar.
Usually the waveform stationarity with self-similarity is poor, and the waveform for not having self-similarity (i.e. non-self similarity) is steady
Property it is higher, can be predicted using linear model.
Step 303: decomposition and reconstruction being carried out to the waveform using wavelet analysis technology, the data traffic after being reconstructed
Signal；Data-traffic signals after the reconstruct include at least one approximation signal and multiple detail signals；
Ha Er (Haar) wavelet analysis technology and mallat algorithm can be used, the waveform is decomposed at least one
Approximation signal and multiple detail signals.In practical application, the number of the detail signal after decomposition, can according to actual needs into
Row selection setting.For example, the waveform can be decomposed at least one approximation signal and three detail signals.
Step 304: the stationarity at least one described approximation signal and multiple detail signals is preset less than first
The signal of threshold value is determined as primary sources flow signal；
About stationarity, it will be understood that: assuming that time series X={ Xt, t=1,2 ... inner each element
It is that the random process for obeying some probability distribution by one generates at random, if X meets the following conditions:
1, mean value E (X)=μ is the constant unrelated with time t；
2, variance var (X)=σ2It is the constant unrelated with time t；
3, covariance cov (Xt,Xt+k)=γkIt is only with time interval k in relation to but with the unrelated constant of time t.
So, time series X is stable.
By define Check-Out Time sequence stationarity when, it is desirable that series infinite is long, is difficult to realize in reality.In general,
What is acquired in practical applications is the sample sequence of limited length.Sample is judged by sample autocorrelation function in the embodiment of the present invention
The stationarity of this sequence.Below about the detailed description of auto-correlation function.
It can judge approximation signal and each detail signal judge whether the stationarity of each signal is small respectively
In the first preset threshold.When the stationarity of signal is less than preset threshold, it is possible to determine that the signal does not have stationarity；When signal
Stationarity is greater than or equal to the first preset threshold, it can be determined that the signal has stationarity.
Step 305: the stationarity at least one described approximation signal and multiple detail signals is greater than or equal to institute
The signal for stating the first preset threshold is determined as secondary sources flow signal；
Step 306: the primary sources flow signal being predicted using compressed sensing model, it is pre- to obtain the first kind
Survey result；
Compressed sensing (Compressed sensing), also referred to as compression sampling (Compressive sampling),
Sparse sampling (Sparse sampling), compression sensing.Compressed sensing is a new sampling theory, passes through exploitation signal
Sparse characteristic can obtain the discrete sample of signal with stochastical sampling, then under conditions of being much smaller than Nyquist sample rate
Pass through non-linear algorithm for reconstructing reconstruction signal.
Step 307:: the secondary sources flow signal is predicted using linear model, obtains the prediction of the second class
As a result；
Step 308: synthesizing the first kind prediction result and the second class prediction result.
In general, a part in data-traffic signals after the reconstruct belongs to primary sources flow signal, another portion
Belong to secondary sources flow signal.It is available for each signal in the data-traffic signals after the reconstruct
One prediction result, the prediction result can indicate the partial data amount for needing to transmit in future time instance network.The synthesis, can
To be to be added each prediction result.Available total prediction result after synthesis, total prediction result can indicate future
The total amount of data for needing to transmit in moment network.According to the total amount of data for needing to transmit in the future time instance network of prediction, network
In relevant device Internet resources can be allocated, to more efficiently carry out data transmission.
In conclusion having by the corresponding waveform of the determination data-traffic signals from phase in embodiment shown in Fig. 3
Like property, decomposition and reconstruction is carried out to the waveform using wavelet analysis technology；According to stationarity to the data flow after the reconstruct
Amount signal is classified；Signal lower for stationarity, is predicted using compressed sensing model；It is higher for stationarity
Signal is predicted using linear model；Synthesize each prediction result；Can according to the waveform characteristic of the data traffic of business,
Data traffic is predicted using the prediction mode that the waveform characteristic with data traffic matches, to improve for data flow
The precision of prediction of amount.
It should be noted that in above-described embodiment, when determining that the corresponding waveform of the data-traffic signals do not have from phase
It, can be the following steps are included: using data-traffic signals described in Linear Model for Prediction when like property.Because of the waveform of non-self similarity
Usual stationarity is higher, it is therefore possible to use linear model directly predicts the data-traffic signals.
It should also be noted that, in above-mentioned steps, the data-traffic signals obtained in predetermined time period specifically may be used
With in the following ways:
It is sampled according to data traffic of the prefixed time interval to generation, each sampling being sequentially arranged
The corresponding data traffic of point；From the corresponding data traffic of each sampled point being sequentially arranged, when interception is default
Between the corresponding data traffic of sampled point in length.
The corresponding waveform of the determination data-traffic signals has self-similarity, specifically can be in the following ways:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；Described in judgement
Whether the value of Hurst Exponent is greater than the second preset threshold.
Rescaled range analysis (Rescaled Range Analysis), also referred to as R/S analytic approach, are hydrologist Hurst
A kind of method proposed on the basis of a large amount of positive researches.
In general, second preset threshold can be set to 0.5.
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis, can be taken
Such as under type:
The corresponding data traffic of sampled point in predetermined time period may be constructed business hours sequence.For the business hours
Sequence X={ Xi, i > 1 }, it is assumed that its sequence length is N, i.e. X={ X1,X2,...,XN}.The sequence is divided intoA sub- sequence
Column, wherein each sub-sequence length is identical, it is n.
Its mean value and standard deviation are calculated to k-th of subsequence
To k-th of subsequence, wherein each sample point Z is calculatedi,kWith EkDeviation, i.e. Xi,k=Zi,k-Ek, subscript i, k table
Show i-th of element in k-th of subsequence.Calculate the Accumulated deviation of k-th of subsequencei
=1,2 ..., n is found out the Accumulated deviation limit difference R of the subsequence by Accumulated deviationk=max { Y1,k,...,Yn,k}-min
{Y1,k,...,Yn,k}。
Calculate the S of all subsequenceskAnd Rk, the R/S statistics of original series is finally calculated, i.e.,
Obviously,It is the function about n.Statistics showsRelationship approximate representation with n is (R/S)n~cnH, both sides take pair
Number, obtains log (R/S)n=logc+Hlogn, wherein logc is constant.When describing n in logarithmic coordinates system and taking different value, own
(logn,log(R/S)n) point, these points can be observed and be located approximately on straight line, Linear Quasi is carried out by least square method
Conjunction finds out the straight slope, as H parameter.
H parameter can state the self-similarity of business.The value interval of H is 0 < H < 1, if 0.5 < H < 1, shows have certainly
Similitude；H value is bigger, illustrates that self-similarity is stronger.
The stationarity of data-traffic signals after the determination reconstruct can specifically be used less than the first preset threshold
Following manner:
According to formulaThe sampled point calculated in the predetermined time period is corresponding
The sample autocorrelation function of data traffic；
JudgementIt is whether true；
Wherein, XiFor the corresponding data traffic of ith sample point in the predetermined time period；EXFor the mean value of X；N is
The number of sampled point in the predetermined time period；K=1,2,3 ... K；K be k the number upper limit, can according to demand into
Row setting.θ is first preset threshold.
The value of K and θ is related to N, and when N is greater than 100, K takes 15, θ to take 1.5, can be according to specific prediction in actual emulation
Object adjustment.
The primary sources flow signal uses compressed sensing model to be predicted, specifically can be in the following ways:
Step A: the first detail signal matrix of construction；
Specifically, extracting business hours sequence for primary sources flow signal.
Business hours sequence is subjected to rarefaction expression.That is, by each element and rarefaction threshold in business hours sequence
Value comparison.If being lower than rarefaction threshold value, which is set to 0, is otherwise remained unchanged.The second threshold can be according to emulation
As a result it determines.
Detail signal time series after rarefaction is indicated is as training data, in conjunction with the number to be predicted indicated with 0 value
According to the first detail signal matrix X of construction (m, n respectively indicate number of users and timeslot number).The matrix is necessary for sparse matrix, i.e. square
The number of 0 element is much larger than non-zero element number in battle array.
Step B: the temporal correlation of the first detail signal matrix is determined
The present invention is designed Time correlation matrix by the characteristic of D1 detail signal matrix X after analysis rarefaction are as follows:
Wherein ωm,n(i), i=1 ..., NTIt is by known N in the first detail signal matrix XTA data linear regression institute
Obtained weight coefficient is obtained by following Solving Linear.
Step C: considering temporal correlation, establishes the Optimized model decomposed about the first detail signal approximate matrix.
Consider the first detail signal matrix temporal correlation corresponding with its, establish Optimized model are as follows:
Whereinβ and α is weight parameter, and Time correlation matrix T describes the first details
Stationarity feature of the data on time dimension in signal matrix, contains the correlation between matrix element on time dimension
Spend size information.Wherein, the value of α and β is related to matrix decomposition low-rank characteristic and temporal correlation.Even α is bigger than normal, indicates square
Battle array decomposition result has more low-rank characteristic, if β is bigger than normal, representing matrix decomposition result has more temporal correlation.Under normal circumstances, α is built
View value is that 0.1, β recommended value is 0.001, can be adjusted according to the actual situation.
Step D: solving optimization model determines factorization matrix L and R.
Using alternately leastsquares algorithm solving optimization problem, low-rank decomposition matrix L and R are obtained.
Step E: the first detail signal matrix of reconstruct obtains data to be predicted.
According to obtained low-rank decomposition matrix L and R, original D1 detail signal matrix is reconstructed:Obtain first
The predicted value of data to be predicted in detail signal matrix.Specifically, data to be predicted are in step, original square is substituted into 0 value
Battle array, after reconstruct, the data on original 0 value corresponding position are the predicted value (usually non-zero value) for data to be predicted.Matrix
In in chronological sequence tactic predicted value correspond to different time points.
It is described that the secondary sources flow signal is predicted using linear model in above-described embodiment, specifically may be used
With in the following ways:
By the traffic list at current time be shown as before n moment portfolio linear weighted function with, i.e.,
Wherein { at, t=1,2 ..., N is white noise sequence
CoefficientUsing Least Square Method, model order n can be found out according to Final prediction error criterion function (FPE).Wherein,
NID indicates independent normal distribution.
Least Square Method parameterDetailed process may is that
Lower linear equation group can be able to by the expression formula of the linear weighted function sum of the portfolio at preceding n moment
It is expressed as with matrix formWherein
It is theoretical according to multiple regression,Least-squares estimation be
The calculation formula of FPE function can beWherein N is original series length, and n is model order
Number,It is being given above, i.e.,
N value when the variance of FPE criterion using one-step prediction error approaches minimum value is as model order, i.e.,L is the preset model order upper bound, and value is related with training sample number.
Fig. 4 is the flow chart of the method for another prediction data flow of the application.As shown in figure 4, this method can wrap
It includes:
Step 401: obtaining the data-traffic signals in predetermined time period；
Step 402: according to the data-traffic signals, judging whether the corresponding waveform of the data-traffic signals has certainly
Similitude obtains judging result；When the judging result indicates that the corresponding waveform of the data-traffic signals does not have self similarity
Property when, execute step 410.
Step 403: when the judging result indicates that the corresponding waveform of the data-traffic signals has self-similarity, adopting
Decomposition and reconstruction is carried out to the waveform with wavelet analysis technology, the data-traffic signals after being reconstructed；After the reconstruct
Data-traffic signals include at least one approximation signal and multiple detail signals；
Step 404: whether the stationarity of the data-traffic signals after judging the reconstruct is less than the first preset threshold；If
Less than the first preset threshold, step 405 is executed；Otherwise, step 406 is executed；
Step 405: the data-traffic signals after the reconstruct are determined as primary sources flow signal；
Step 406: the data-traffic signals after the reconstruct are determined as secondary sources flow signal；
Step 407: the primary sources flow signal being predicted using compressed sensing model, it is pre- to obtain the first kind
Survey result；
Step 408: the secondary sources flow signal being predicted using linear model, obtains the second class prediction knot
Fruit；
Step 409: synthesizing the first kind prediction result and the second class prediction result.
Step 410: using data-traffic signals described in Linear Model for Prediction.
In embodiment shown in Fig. 4, by judging whether the corresponding waveform of the data-traffic signals has self-similarity,
When with similitude, decomposition and reconstruction is carried out to the waveform using wavelet analysis technology；According to stationarity to the reconstruct
Data-traffic signals afterwards are classified；Signal lower for stationarity, is predicted using compressed sensing model；For flat
The higher signal of stability, is predicted using linear model；Synthesize each prediction result；When the data-traffic signals are corresponding
When waveform does not have similitude, using data-traffic signals described in Linear Model for Prediction；For the data traffic with similitude
Signal and data-traffic signals without similitude can carry out data traffic using the prediction mode to match pre-
It surveys, to improve the precision of prediction for data traffic, extends the scope of application of the method for the prediction data flow of the application.
In addition, calculate node may be the master comprising computing capability the embodiment of the present application also provides a kind of calculate node
Machine server personal computer PC or portable portable computer or terminal etc., the application are specifically real
Example is applied not limit the specific implementation of calculate node.
Fig. 5 is the structure chart of the calculate node of the application.As shown in figure 5, calculate node 500 includes:
Processor (processor) 510, communication interface (Communications Interface) 520, memory
(memory) 530, bus 540.
Processor 510, communication interface 520, memory 530 complete mutual communication by bus 540.
Processor 510, for executing program 532.
Specifically, program 532 may include program code, and said program code includes computer operation instruction.
Processor 510 may be a central processor CPU or specific integrated circuit ASIC (Application
Specific Integrated Circuit), or be arranged to implement the integrated electricity of one or more of the embodiment of the present application
Road.
Memory 530, for storing program 532.Memory 530 may include high speed RAM memory, it is also possible to further include
Nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.
For storing computer executed instructions, the processor 510 passes through the memory 530 with the memory 530
The bus connection, when the operation of the device of the prediction data flow, the processor 510 executes the memory 530 and deposits
The computer executed instructions 532 of storage, so that the processor executes the prediction data flow provided in the embodiment of the present application
Method.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes that
A little elements, but also including other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Through the above description of the embodiments, those skilled in the art can be understood that the application can be by
Software adds the mode of required hardware platform to realize, naturally it is also possible to all implemented by hardware, but in many cases before
Person is more preferably embodiment.Based on this understanding, the technical solution of the application contributes to background technique whole or
Person part can be embodied in the form of software products, which can store in storage medium, such as
ROM/RAM, magnetic disk, CD etc., including some instructions are used so that a computer equipment (can be personal computer, service
Device or the network equipment etc.) execute method described in certain parts of each embodiment of the application or embodiment.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
For, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is said referring to method part
It is bright.
Specific examples are used herein to illustrate the principle and implementation manner of the present application, and above embodiments are said
It is bright to be merely used to help understand the present processes and its core concept；At the same time, for those skilled in the art, foundation
The thought of the application, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as the limitation to the application.
Claims (8)
1. a kind of device of prediction data flow characterized by comprising
Signal acquisition module, for obtaining the data-traffic signals in predetermined time period；
Determining module, for determining that the corresponding waveform of the data-traffic signals has from phase according to the data-traffic signals
Like property；
Signal processing module, for carrying out decomposition and reconstruction to the waveform using wavelet analysis technology, the number after being reconstructed
According to flow signal；Data-traffic signals after the reconstruct include at least one approximation signal and multiple detail signals；
The determining module is also used to the stationarity at least one described approximation signal and multiple detail signals less than
The signal of one preset threshold is determined as primary sources flow signal；
Stationarity at least one described approximation signal and multiple detail signals is greater than or equal to the described first default threshold
The signal of value is determined as secondary sources flow signal；
Computing module obtains the first kind for being predicted using compressed sensing model the primary sources flow signal
Prediction result；
The secondary sources flow signal is predicted using linear model, obtains the second class prediction result；
Synthesize the first kind prediction result and the second class prediction result；
The determining module is also used to:
Determine that the corresponding waveform of the data-traffic signals does not have self-similarity；
The computing module, is also used to:
When the determining module determines that the corresponding waveform of the data-traffic signals does not have self-similarity, using linear model
Predict the data-traffic signals.
2. the apparatus according to claim 1, which is characterized in that the signal acquisition module is specifically used for:
It is sampled according to data traffic of the prefixed time interval to generation, each sampled point pair being sequentially arranged
The data traffic answered；
From the corresponding data traffic of each sampled point being sequentially arranged, the sampling in predetermined time period is intercepted
The corresponding data traffic of point.
3. the apparatus according to claim 1, which is characterized in that the determining module is specifically used for:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；
Determine that the value of the Hurst Exponent is greater than the second preset threshold.
4. the apparatus of claim 2, which is characterized in that the determining module is specifically used for:
According to formulaCalculate at least one described approximation signal and multiple detail signals
Sample autocorrelation function；
It willSignal be determined as the primary sources flow signal；
Wherein, XiFor the corresponding data traffic of ith sample point in predetermined time period；EXFor the mean value of X；N is described default
The number of sampled point in time span；K=1,2,3 ... K；θ is first preset threshold.
5. a kind of method of prediction data flow characterized by comprising
Obtain the data-traffic signals in predetermined time period；
According to the data-traffic signals, determine that the corresponding waveform of the data-traffic signals has self-similarity；
Decomposition and reconstruction is carried out to the waveform using wavelet analysis technology, the data-traffic signals after being reconstructed；It is described heavy
Data-traffic signals after structure include at least one approximation signal and multiple detail signals；
Signal by the stationarity at least one described approximation signal and multiple detail signals less than the first preset threshold, really
It is set to primary sources flow signal；
Stationarity at least one described approximation signal and multiple detail signals is greater than or equal to the described first default threshold
The signal of value is determined as secondary sources flow signal,
The primary sources flow signal is predicted using compressed sensing model, obtains first kind prediction result；
The secondary sources flow signal is predicted using linear model, obtains the second class prediction result；
Synthesize the first kind prediction result and the second class prediction result；
Determine that the corresponding waveform of the data-traffic signals does not have self-similarity；
Using data-traffic signals described in Linear Model for Prediction.
6. according to the method described in claim 5, it is characterized in that, the data traffic letter obtained in predetermined time period
Number, it specifically includes:
It is sampled according to data traffic of the prefixed time interval to generation, each sampled point pair being sequentially arranged
The data traffic answered；
From the corresponding data traffic of each sampled point being sequentially arranged, the sampling in predetermined time period is intercepted
The corresponding data traffic of point.
7. according to the method described in claim 5, it is characterized in that, the corresponding waveform tool of the determination data-traffic signals
There is self-similarity, specifically include:
The Hurst Exponent of the corresponding waveform of the data-traffic signals is calculated using Rescaled range analysis；
Determine that the value of the Hurst Exponent is greater than the second preset threshold.
8. according to the method described in claim 6, it is characterized in that, described by least one described approximation signal and multiple thin
The stationarity in signal is saved less than the signal of the first preset threshold, is determined as primary sources flow signal, specifically includes:
According to formulaCalculate at least one described approximation signal and multiple detail signals
Sample autocorrelation function；
It willSignal be determined as the primary sources flow signal；
Wherein, XiFor the corresponding data traffic of ith sample point in predetermined time period；EXFor the mean value of X；N is described default
The number of sampled point in time span；K=1,2,3 ... K；θ is first preset threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410727881.7A CN104506378B (en) | 2014-12-03 | 2014-12-03 | A kind of device and method of prediction data flow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410727881.7A CN104506378B (en) | 2014-12-03 | 2014-12-03 | A kind of device and method of prediction data flow |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104506378A CN104506378A (en) | 2015-04-08 |
| CN104506378B true CN104506378B (en) | 2019-01-18 |
Family
ID=52948093
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410727881.7A Active CN104506378B (en) | 2014-12-03 | 2014-12-03 | A kind of device and method of prediction data flow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104506378B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106936611B (en) * | 2015-12-30 | 2019-12-17 | 杭州华为数字技术有限公司 | Method and device for predicting network state |
| CN106157615B (en) * | 2016-06-29 | 2018-11-30 | 泰州市龙泽环境科技有限公司 | A kind of traffic flow information management handheld terminal |
| CN105913654B (en) * | 2016-06-29 | 2018-06-01 | 深圳市前海绿色交通有限公司 | A kind of Intelligent traffic management systems |
| CN106128139B (en) * | 2016-06-29 | 2018-12-14 | 徐州海德力工业机械有限公司 | A kind of unmanned vehicle of automatic dodging congestion route |
| CN106097712B (en) * | 2016-06-29 | 2018-12-04 | 江苏顺泰交通集团有限公司 | A kind of flow-optimized guidance system of traffic |
| CN105913665B (en) * | 2016-06-29 | 2018-06-26 | 山东至信信息科技有限公司 | A kind of intelligence zebra stripes guardrail |
| CN106128098B (en) * | 2016-06-29 | 2018-08-03 | 山西通畅工程勘察设计咨询有限公司 | A kind of multi-display apparatus that can carry out traffic flow forecasting |
| CN105913664B (en) * | 2016-06-29 | 2018-06-12 | 广州浩瑞尔信息科技有限公司 | A kind of magnitude of traffic flow monitoring and controlling forecast system |
| CN106128122B (en) * | 2016-06-29 | 2018-11-06 | 江苏润仕达交通设施有限公司 | A kind of intelligent traffic lamp |
| CN106128102B (en) * | 2016-06-29 | 2018-11-20 | 泰兴市智谷科技孵化器中心 | A kind of traffic jam prior-warning device |
| CN109873712B (en) * | 2018-05-18 | 2022-03-22 | 新华三信息安全技术有限公司 | Network traffic prediction method and device |
| CN109389100A (en) * | 2018-11-01 | 2019-02-26 | 河南思维轨道交通技术研究院有限公司 | A kind of filtering noise-reduction method, device, electronic equipment and storage medium |
| CN110839016B (en) * | 2019-10-18 | 2022-07-15 | 平安科技（深圳）有限公司 | Abnormal flow monitoring method, device, equipment and storage medium |
| CN113055923B (en) * | 2019-12-27 | 2022-06-17 | 中国移动通信集团湖南有限公司 | Mobile network traffic prediction method, device and equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6917711B1 (en) * | 1998-08-10 | 2005-07-12 | Digital Accelerator Corporation | Embedded quadtree wavelets in image compression |
| CN1744718A (en) * | 2004-09-01 | 2006-03-08 | 三菱电机株式会社 | In-frame prediction for high-pass time filtering frame in small wave video coding |
| CN101599871A (en) * | 2009-05-27 | 2009-12-09 | 南京欣网视讯科技股份有限公司 | A kind of SFARIMA network flow prediction method |
| 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 |
| CN103095496A (en) * | 2013-01-10 | 2013-05-08 | 周亚建 | Prediction method and device for network flow |
-
2014
- 2014-12-03 CN CN201410727881.7A patent/CN104506378B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6917711B1 (en) * | 1998-08-10 | 2005-07-12 | Digital Accelerator Corporation | Embedded quadtree wavelets in image compression |
| CN1744718A (en) * | 2004-09-01 | 2006-03-08 | 三菱电机株式会社 | In-frame prediction for high-pass time filtering frame in small wave video coding |
| CN101599871A (en) * | 2009-05-27 | 2009-12-09 | 南京欣网视讯科技股份有限公司 | A kind of SFARIMA network flow prediction method |
| 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 |
| CN103095496A (en) * | 2013-01-10 | 2013-05-08 | 周亚建 | Prediction method and device for network flow |
Non-Patent Citations (4)
| Title |
|---|
| 基于小波变换与组合模型的网络流量预测算法;魏永涛;汪晋宽;王翠荣;张琨;《东北大学学报(自然科学版)》;20111015;第32卷(第10期);第1382-1389页 |
| 基于小波技术的网络流量分析和预测;王西锋;高岭;张晓孪;《计算机应用与软件》;20080815;第25卷(第8期);第70-72页 |
| 基于流量特征的网络流量预测研究;张凤荔;赵永亮;王丹;王豪;《计算机科学》;20140415;第41卷(第4期);第86-98页 |
| 小波分析和AR-LSSVM的网络流量预测;冯华丽;刘渊;《计算机工程与应用》;20110711;第47卷(第20期);第88-162页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104506378A (en) | 2015-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104506378B (en) | A kind of device and method of prediction data flow | |
| Singh et al. | Repeated wavelet transform based ARIMA model for very short-term wind speed forecasting | |
| Hsieh et al. | Deep learning-based indoor localization using received signal strength and channel state information | |
| CN103139907B (en) | A kind of indoor wireless positioning method utilizing fingerprint technique | |
| CN106933649B (en) | Virtual machine load prediction method and system based on moving average and neural network | |
| CN109901107A (en) | A kind of time difference positioning method, device, computer equipment and readable storage medium storing program for executing | |
| CN103559888A (en) | Speech enhancement method based on non-negative low-rank and sparse matrix decomposition principle | |
| CN111898578B (en) | Crowd density acquisition method and device and electronic equipment | |
| CN104424235A (en) | Method and device for clustering user information | |
| CN103001805A (en) | Web user flow producing method based on user behavior analysis | |
| CN104333424B (en) | A kind of frequency spectrum detection and unknown noise variance tracking method of estimation and device | |
| JP5463516B2 (en) | Encryption key generation system, encryption key generation method, and encryption key generation program | |
| CN106128466A (en) | Identity vector processing method and device | |
| CN109163683A (en) | Track wave grinds disease screening method and apparatus | |
| CN109143371A (en) | A kind of noise remove method and device of seismic data | |
| US20150161231A1 (en) | Data sampling method and data sampling device | |
| RU2495504C1 (en) | Method of reducing transmission rate of linear prediction low bit rate voders | |
| CN105553583B (en) | The indoor white frequency spectrum detecting method from training based on Bayes's compressed sensing | |
| CN110263810B (en) | LoRa signal source identification method | |
| Li et al. | An innovated integrated model using singular spectrum analysis and support vector regression optimized by intelligent algorithm for rainfall forecasting | |
| CN113822331A (en) | Encrypted traffic classification method and device and electronic equipment | |
| Sharbati et al. | Optimized cauchy-gaussian blend model for stochastic-parametric simulation of seismic ground motions | |
| CN113723006B (en) | LS-SVM (least squares-support vector machine) -based single-station earth change magnetic field modeling prediction method and system | |
| CN117351993B (en) | Audio transmission quality evaluation method and system based on audio distribution | |
| Luo et al. | Frequency-division combination forecasting of stock market based on wavelet multiresolution analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |