CN105471631A - Network traffic prediction method based on traffic trend - Google Patents
Network traffic prediction method based on traffic trend Download PDFInfo
- Publication number
- CN105471631A CN105471631A CN201510793377.1A CN201510793377A CN105471631A CN 105471631 A CN105471631 A CN 105471631A CN 201510793377 A CN201510793377 A CN 201510793377A CN 105471631 A CN105471631 A CN 105471631A
- Authority
- CN
- China
- Prior art keywords
- moment
- network
- traffic trends
- network traffic
- error
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
Abstract
The invention provides a network traffic prediction method based on traffic trend. The method is implemented according to the following steps: S1, extracting the network traffic trend under n time periods before a current time period i and the network traffic trend from the first moment to a moment c in the current time period i, wherein n is a positive integer; S2, predicting the network traffic trend at a future moment according to the extracted network traffic trend; S3, calculating the error between an extracted network traffic value and the network traffic trend, and predicting the traffic error; S4, predicting the network traffic predicted value at a future moment according to the network traffic trend predicted in S2 and the traffic error predicted in S3; and S5, letting c=c+k, ending the process if c is greater than or equal to the ending moment of the time period i, or returning to S2. The number of training samples needed for prediction is reduced greatly while the prediction precision is improved. The method can be applied to actual network management and measurement more easily.
Description
Technical field
The present invention relates to computer network field, be specifically related to a kind of network flow prediction method based on traffic trends.
Background technology
Predicting network flow tool in network control and management has very important significance, network traffics planning is conducive to the long-term prediction of network traffics, and the network problem that better reply is possible, the short-term prediction of network traffics is then conducive to real-time Dynamic Programming disparate networks resource, as bandwidth sum route etc.
The fast development of network itself in recent years, the explosive growth of the network user, enriching constantly of network application kind makes the complexity of network itself greatly increase, and the characteristic of network traffics also great variety occurs.Traditional Network Traffic Forecast Model, as ARIMA, SARIMA, the model that the short-term correlation of the flow Network Based such as Markov model proposes can not accurately describe current network flow characteristic, volume forecasting poor effect.
Effectively can describe the correlation of network traffics under each time scale by all kinds of for the introducing of fractal thought based on after the model of ARMA, and then successfully carry out volume forecasting.But too high based on fractal model complexity, amount of calculation is large, is difficult to be applied to reality.
Artificial neural net obtains noticeable achievement at the application and development in the field such as signal transacting, pattern recognition.Because it can reach target by study, ANN has great advantage at process non-linear process.Learnt by the feature of training set to network traffics in process predicting network flow, compared with traditional Forecasting Methodology, prediction based on ANN method does not need the Mathematical Modeling of predefined sample data, namely can predict quite accurately by means of only learning sample data, therefore tool has many advantages, thus be widely used in predicting network flow, particularly in volume forecasting in short-term.Conventional method has BP model, FNT model and all kinds of in conjunction with genetic algorithm, the ANN model etc. that the optimized algorithms such as genetic algorithm improve.But require that when carrying out volume forecasting a large amount of samples carries out training to obtain optimized parameter and structure based on the forecast model of ANN, considerably increase cost and the complexity of prediction.Therefore, in real-time prediction, the forecast model based on ANN has significant limitation.
Summary of the invention
In order to overcome the defect existed in above-mentioned prior art, the invention provides a kind of network flow prediction method based on traffic trends.
To achieve these goals, the invention provides a kind of network flow prediction method based on traffic trends, carry out according to following steps:
S1: set current time as c, the time cycle at current time c place is i, extract the network traffic trends under n the time cycle before current time period i, and from the network traffic trends of the 1st moment to moment c in current time period i, described n is positive integer;
S2: according to the network traffic trends in k1 moment before the current time c extracted, the network traffic trends in following k the moment of prediction current time c, described k1, k are positive integer;
S3: calculate the error between the network flow value in k1 the moment of extracting and its network traffic trends, the network traffics error in following k the moment of prediction current time c;
S4: according to the network traffics error predicted in the network traffic trends predicted in step S2 and S3, the predicting network flow value in following k the moment of prediction current time c;
S5: make c=c+k, if c is more than or equal to the finish time of time cycle i, then EP (end of program); Otherwise return step S2.
The present invention maintains the periodic feature of network traffics and the partial structurtes feature of each time cycle down-off by the network traffic trends extracted, and is convenient to the more accurate network traffic trends to future time instance and predicts.
In the preferred embodiment of the present invention, the extracting method of described network traffic trends is:
S11: known network flow sequence s, the network traffics value sequence inscribed when comprising S, a time cycle is made to comprise J moment, be that row are reassembled as traffic matrix TR with time cycle by network traffics sequence s, total [S/J] line time cycle arranges, [] for giving up the integer of remainder, the wherein flow value in J moment in the time cycle of each row record;
S12: calculated flow rate trend matrix T T, formula is:
Wherein I is unit matrix, second differnce matrix D ∈ R
(J-2) × J, D
i,i=1, D
i, i+1=-2, D
i, i+2=1, R is real number, L=diag (sum (simC)), and matrix simC represents the similitude of flow between each time cycle in traffic trends matrix T T, and sum (), to the row summation of matrix, obtains vector; Diag () carries out diagonalization to vector, and the element obtained in new matrix on each diagonal is the value in vector,
for Kronecker product; Vec () is for being converted into vector by matrix; λ 1, λ 2 is parameter, and represent flatness and the ratio of local similarity shared by trend abstraction of flow successively, span is [0,1].
The network traffic trends that extracts of this step can the prediction of repeated application some k moment flow in subsequent step, greatly reduces flow error prediction and the number of training required for flow estimation, saves the training time; And the network traffic trends extracted not only highlights the periodic characteristic of flow under each time cycle, also maintains the partial structurtes feature of flow.
In the preferred embodiment of the present invention, the Forecasting Methodology of described network traffic trends is:
S21: the average discharge trend ar_tt calculating current time period according to the network traffic trends of n time cycle before the current time period i extracted
i, computing formula is:
ar_tt
i＝(θ
n*tt
i-n+θ
n-1*tt
i-n+1+θ
2*tt
i-2+θ
1*tt
i-1)/(θ
n+θ
n-1+...+θ
2+θ
1)，
Wherein tt
i-n, tt
i-n+1..., tt
i-2, tt
i-1the network traffic trends of the i-th-n under front n the time cycle being expressed as current time period i successively, the i-th-n+1 network traffic trends ..., the i-th-2 network traffic trends, the i-th-1 network traffic trends; θ
1, θ
2..., θ
n-1, θ
nrepresent the weights that the corresponding time cycle is occupied in average discharge trend calculates respectively;
S22: make moment c=mk+k1, calculate the network traffic trends error between the moment (mk+1) to the network traffic trends and average discharge trend in moment (mk+k1), formula is:
[re_tt
(i,mk+1),re_tt
(i,mk+2),...,re_tt
(i,mk+k1)]'＝
([tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'-[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]')./
[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]'
[re_tt
(i, mk+1), re_tt
(i, mk+2)..., re_tt
(i, mk+k1)] ' represent network traffic trends error;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the network traffic trends of moment (mk+1) to moment (mk+k1) successively;
[ar_tt
(i, mk+1), ar_tt
(i, mk+2)..., ar_tt
(i, mk+k1)] ' be expressed as the average discharge trend in moment (mk+1) to moment (mk+k1) successively;
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle, the value of M is:
./represent two waits the element between long vector on correspondence position to be divided by;
S23: according to the network traffic trends error of S22 gained, utilize neural network model, error between the network traffic trends in following k the moment in prediction c moment and average discharge trend, obtains the network traffic trends error in a following k moment, i.e. [pr_re_tt
(i, mk+k1+1), pr_re_tt
(i, mk+k1+2)..., pr_re_tt
(i, mk+k1+k)] ',
S24: the network traffic trends calculating a following k moment again according to the network traffic trends error in k moment of prediction and average discharge trend, formula is:
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'＝
(E+[pr_re_tt
(i,mk+k1+1),pr_re_tt
(i,mk+k1+2),...,pr_re_tt
(i,mk+k1+k)]').*
[ar_tt
(i,mk+k1+1),ar_tt
(i,mk+k1+2),...,ar_tt
(i,mk+k1+k)]'
Wherein E is the column vector that all elements is 1, and .* represents the element multiplication between two long vectors such as grade on correspondence position.
Be more preferably in execution mode in one of the present invention, n gets 4, θ in the step s 21
1=1, θ
2=1, θ 3=0.5, θ
4=0.5, its formula is:
ar_tt
i＝(0.5*tt
i-4+0.5*tt
i-3+tt
i-2+tt
i-1)/3。
In the preferred embodiment of the present invention, the Forecasting Methodology of network traffics error is:
S31: calculate the error between the moment (mk+1) to the network flow value and network traffic trends in moment (mk+k1), formula is:
[re
(i,mk+1),re
(i,mk+2),...,re
(i,mk+k1)]'＝
([tr
(i,mk+1),tr
(i,mk+2),...,tr
(i,mk+k1)]'-[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]')./
[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle ./represent that two are waited the element between long vectors on correspondence position to be divided by;
[re
(i, mk+1), re
(i, mk+2)..., re
(i, mk+k1)] ' represent following error;
[tr
(i, mk+1), tr
(i, mk+2)..., tr
(i, mk+k1)] ' be expressed as the network traffics of moment (mk+1) to moment (mk+k1) successively;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the traffic trends of moment (mk+1) to moment (mk+k1) successively;
S32: based on the flow error of neural network model study gained, and predict the network traffics error in a following k moment, i.e. [pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] ';
In the preferred embodiment of the present invention, the computational methods of predicting network flow value are:
[pr_tr
(i,mk+k1+1),pr_tr
(i,mk+k1+2),...,pr_tr
(i,mk+k1+k)]'＝
(E+[pr_re
(i,mk+k1+1),pr_re
(i,mk+k1+2),...,pr_re
(i,mk+k1+k)]').*
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle,
[pr_tr
(i, mk+k1+1), pr_tr
(i, mk+k1+2)..., pr_tr
(i, mk+k1+k)] ' represent predicting network flow value;
[pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] ' represent the network traffics error predicted;
[pr_tt
(i, mk+k1+1), pr_tt
(i, mk+k1+2)..., pr_tt
(i, mk+k1+k)] ' represent the network traffic trends predicted.
The present invention is by the trend feature to network traffics sequence in the existing time cycle, and in current time period, the feature of known mass flow learns, and realizes the prediction to future network traffic trends and flow value in current time period; While improve precision of prediction, greatly reduce the number of training required for prediction, be easier to be applied in real network management and measurement.
Accompanying drawing explanation
Above-mentioned and/or additional aspect of the present invention and advantage will become obvious and easy understand from accompanying drawing below combining to the description of embodiment, wherein:
Fig. 1 is the volume forecasting schematic flow sheet of model of the present invention on certain link in i-th time cycle.
Fig. 2 is schematic flow sheet of the present invention.
Fig. 3 is the present invention within i-th time cycle, and the flow value in a known k1 moment predicts the schematic flow sheet of the network traffic trends in a following k moment.
Fig. 4 is the present invention within i-th time cycle, and the flow value in a known k1 moment predicts the schematic flow sheet of the network traffics in a following k moment.
Embodiment
Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.
The flow sequence that the present invention have selected Abilene data set and lbl data set has done volume forecasting, when processing network traffics sequence, namely the traffic matrix (the former is 1 day the time cycle, and the time cycle of the latter is 1 second) being often classified as a time cycle is converted into after data scrubbing is carried out to sequence.Carry out predicting the network traffics representing k1 the moment before in certain time cycle known in the present invention to network traffics sequence, the flow in all moment following in this time cycle is predicted.Be illustrated in figure 1 the volume forecasting block diagram of model of the present invention on certain network link in time cycle i.The network traffics sequence of time cycle i is divided into flow sequence that (M+1) individual length is k1 (in order to be used as the network traffics sequence in k1 moment before the current time c of training set in artificial actual application), and to each k1 sequence carry out predicting can obtain predicted flow rate sequence that corresponding length is k (in order in artificial actual application by the network traffics sequence in following k moment of a front k1 sequence prediction), namely the sequence merging predicted the most at last obtains the flow of all predictions in this time cycle from the k1+1 moment.
The invention provides a kind of network flow prediction method of traffic trends, as shown in Figure 2, carry out according to following steps:
The first step, if current time is c, the time cycle at current time c place is i, extracts the network traffic trends under n the time cycle before current time period i, and from the network traffic trends of the 1st moment to moment c in current time period i, described n is positive integer.In the present embodiment, the extracting method of network traffic trends is:
S11: known network flow sequence s, the network traffics value sequence inscribed when comprising S, a time cycle is made to comprise J moment, be that row are reassembled as traffic matrix TR with time cycle by network traffics sequence s, total [S/J] line time cycle arranges, [] for giving up the integer of remainder, the wherein flow value in J moment in the time cycle of each row record.Wherein, the value of S, J is according to the difference of data set, and the value of S with J is different; Such as, the length S=36000 of data set used in an experiment; J is also the value different according to different DSDs; If a data set fetches data in units of 5min, a time cycle is 1 day, then J=288, if a data set is in units of 0.1s, a time cycle is 1min, then J=600.In addition, if S=36001, J=600, then [S/J]=60, namely get front 36000 flow values and form matrix, last is given up.
S12: calculated flow rate trend matrix T T, formula is:
Wherein I is unit matrix, second differnce matrix D ∈ R
(J-2) × J, D
i,i=1, D
i, i+1=-2, D
i, i+2=1, R is real number, L=diag (sum (simC)), and matrix simC represents the similitude of flow between each time cycle in traffic trends matrix T T, and sum (), to the row summation of matrix, obtains vector; Diag () carries out diagonalization to vector, and the element obtained in new matrix on each diagonal is the value in vector,
for Kronecker product; Vec () is for being converted into vector by matrix; λ 1, λ 2 is parameter, and represent flatness and the ratio of local similarity shared by trend abstraction of flow successively, span is [0,1].
Second step, according to the network traffic trends in k1 moment before the current time c extracted, the network traffic trends in following k the moment of prediction current time c, described k1, k are positive integer.In the present embodiment, as shown in Figure 3, the Forecasting Methodology of network traffic trends is:
S21: the average discharge trend ar_tt calculating current time period according to the network traffic trends of n time cycle before the current time period i extracted
i, ar_tt
ibe a vector, it comprises J numerical value, and it element comprised can be denoted as: ar_tt
(i, mk+1),ar_tt
(i, mk+2) ...,ar_tt
(i, mk+k1) ...,ar_tt
(i, mk+k1+k).average discharge trend ar_tt
icomputing formula be:
ar_tt
i＝(θ
n*tt
i-n+θ
n-1*tt
i-n+1+θ
2*tt
i-2+θ
1*tt
i-1)/(θ
n+θ
n-1+...+θ
2+θ
1)，
Wherein tt
i-n, tt
i-n+1..., tt
i-2, tt
i-1the network traffic trends of the i-th-n under front n the time cycle being expressed as current time period i successively, the i-th-n+1 network traffic trends ..., the i-th-2 network traffic trends, the i-th-1 network traffic trends; θ
1, θ
2..., θ
n-1, θ
nrepresent the weights that the corresponding time cycle is occupied in average discharge trend calculates respectively.In the present embodiment, when n gets 4, θ
1=1, θ
2=1, θ 3=0.5, θ
4=0.5, its formula is: ar_tt
i=(0.5*tt
i-4+ 0.5*tt
i-3+ tt
i-2+ tt
i-1)/3.
S22: make moment c=mk+k1, calculate the network traffic trends error between the moment (mk+1) to the network traffic trends and average discharge trend in moment (mk+k1), formula is:
[re_tt
(i,mk+1),re_tt
(i,mk+2),...,re_tt
(i,mk+k1)]'＝
([tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'-[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]')./
[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]'
[re_tt
(i, mk+1), re_tt
(i, mk+2)..., re_tt
(i, mk+k1)] ' represent network traffic trends error;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the network traffic trends of moment (mk+1) to moment (mk+k1) successively;
[ar_tt
(i, mk+1), ar_tt
(i, mk+2)..., ar_tt
(i, mk+k1)] ' be expressed as the average discharge trend in moment (mk+1) to moment (mk+k1) successively;
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle ./represent that two are waited the element between long vectors on correspondence position to be divided by;
S23: according to the network traffic trends error of S22 gained, utilize neural network model, error between the network traffic trends in following k the moment in prediction c moment and average discharge trend, obtains the network traffic trends error in a following k moment, i.e. [pr_re_tt
(i, mk+k1+1), pr_re_tt
(i, mk+k1+2)..., pr_re_tt
(i, mk+k1+k)] '.Wherein, neural network model adopts prior art, concrete parameter is set in advance when testing, what all adopt with the prediction of later step S32 is three layers of bp neural network model herein, every node layer number is 2 respectively, 2,1, every layer of transfer function used is tanh sigmoid function, tansig function, logarithm S shape transfer function, logsig and linear function purelin respectively; The training function used is Gradient Descent autoadapted learning rate training function, the learning rate Ir=0.01 of traingdx definition, maximum iteration time epochs=2000, the target error goal=0.01 of training.
S24: the network traffic trends calculating a following k moment again according to the network traffic trends error in k moment of prediction and average discharge trend, formula is:
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'＝
(E+[pr_re_tt
(i,mk+k1+1),pr_re_tt
(i,mk+k1+2),...,pr_re_tt
(i,mk+k1+k)]').*
[ar_tt
(i,mk+k1+1),ar_tt
(i,mk+k1+2),...,ar_tt
(i,mk+k1+k)]'
Wherein E is the column vector that all elements is 1, and .* represents the element multiplication between two long vectors such as grade on correspondence position.
3rd step, calculates the error between the network flow value in k1 the moment of extracting and its network traffic trends, the network traffics error in following k the moment of prediction current time c.In the present embodiment, as shown in Figure 4, the Forecasting Methodology of network traffics error is:
S31: calculate the error between the moment (mk+1) to the network flow value and network traffic trends in moment (mk+k1), formula is:
[re
(i,mk+1),re
(i,mk+2),...,re
(i,mk+k1)]'＝
([tr
(i,mk+1),tr
(i,mk+2),...,tr
(i,mk+k1)]'-[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]')./，
[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'
Wherein ./represent that two are waited the element between long vector on correspondence position to be divided by;
[re
(i, mk+1), re
(i, mk+2)..., re
(i, mk+k1)] ' represent following error;
[tr
(i, mk+1), tr
(i, mk+2)..., tr
(i, mk+k1)] ' be expressed as the network traffics of moment (mk+1) to moment (mk+k1) successively;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the traffic trends of moment (mk+1) to moment (mk+k1) successively.
S32: based on the flow error of neural network model study gained, and predict the network traffics error in a following k moment, i.e. [pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] '.
4th step, according to the network traffics error that the network traffic trends predicted in second step and the 3rd step are predicted, the predicting network flow value in following k the moment of prediction current time c.In the present embodiment, the computational methods of predicting network flow value are:
[pr_tr
(i,mk+k1+1),pr_tr
(i,mk+k1+2),...,pr_tr
(i,mk+k1+k)]'＝
(E+[pr_re
(i,mk+k1+1),pr_re
(i,mk+k1+2),...,pr_re
(i,mk+k1+k)]').*
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'，
Wherein:
[pr_tr
(i, mk+k1+1), pr_tr
(i, mk+k1+2)..., pr_tr
(i, mk+k1+k)] ' represent predicting network flow value;
[pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] ' represent the network traffics error predicted;
[pr_tt
(i, mk+k1+1), pr_tt
(i, mk+k1+2)..., pr_tt
(i, mk+k1+k)] ' represent the network traffic trends predicted.
5th step, makes c=c+k, if c is more than or equal to the finish time of time cycle i, then and EP (end of program); Otherwise return second step.
In the description of this specification, specific features, structure, material or feature that the description of reference term " embodiment ", " some embodiments ", " example ", " concrete example " or " some examples " etc. means to describe in conjunction with this embodiment or example are contained at least one embodiment of the present invention or example.In this manual, identical embodiment or example are not necessarily referred to the schematic representation of above-mentioned term.And the specific features of description, structure, material or feature can combine in an appropriate manner in any one or more embodiment or example.
Although illustrate and describe embodiments of the invention, those having ordinary skill in the art will appreciate that: can carry out multiple change, amendment, replacement and modification to these embodiments when not departing from principle of the present invention and aim, scope of the present invention is by claim and equivalents thereof.
Claims (6)
1. based on a network flow prediction method for traffic trends, it is characterized in that, carry out according to following steps:
S1: set current time as c, the time cycle at current time c place is i, extract the network traffic trends under n the time cycle before current time period i, and from the network traffic trends of the 1st moment to moment c in current time period i, described n is positive integer;
S2: according to the network traffic trends in k1 moment before the current time c extracted, the network traffic trends in following k the moment of prediction current time c, described k1, k are positive integer;
S3: calculate the error between the network flow value in k1 the moment of extracting and its network traffic trends, the network traffics error in following k the moment of prediction current time c;
S4: according to the network traffics error predicted in the network traffic trends predicted in step S2 and S3, the predicting network flow value in following k the moment of prediction current time c;
S5: make c=c+k, if c is more than or equal to the finish time of time cycle i, then EP (end of program); Otherwise return step S2.
2. the network flow prediction method based on traffic trends according to claim 1, is characterized in that, the extracting method of described network traffic trends is:
S11: known network flow sequence s, the network traffics value sequence inscribed when comprising S, a time cycle is made to comprise J moment, be that row are reassembled as traffic matrix TR with time cycle by network traffics sequence s, total [S/J] line time cycle arranges, [] for giving up the integer of remainder, the wherein flow value in J moment in the time cycle of each row record;
S12: calculated flow rate trend matrix T T, formula is:
Wherein I is unit matrix, second differnce matrix D ∈ R
(J-2) × J, D
i,i=1, D
i, i+1=-2, D
i, i+2=1, R is real number, L=diag (sum (simC)), and matrix simC represents the similitude of flow between each time cycle in traffic trends matrix T T, and sum (), to the row summation of matrix, obtains vector; Diag () carries out diagonalization to vector, and the element obtained in new matrix on each diagonal is the value in vector,
for Kronecker product; Vec () is for being converted into vector by matrix; λ 1, λ 2 is parameter, and represent flatness and the ratio of local similarity shared by trend abstraction of flow successively, span is [0,1].
3. the network flow prediction method based on traffic trends according to claim 1, is characterized in that, the Forecasting Methodology of described network traffic trends is:
S21: the average discharge trend ar_tt calculating current time period according to the network traffic trends of n time cycle before the current time period i extracted
i, computing formula is:
ar_tt
i＝(θ
n*tt
i-n+θ
n-1*tt
i-n+1+θ
2*tt
i-2+θ
1*tt
i-1)/(θ
n+θ
n-1+...+θ
2+θ
1)，
Wherein tt
i-n, tt
i-n+1..., tt
i-2, tt
i-1the network traffic trends of the i-th-n under front n the time cycle being expressed as current time period i successively, the i-th-n+1 network traffic trends ..., the i-th-2 network traffic trends, the i-th-1 network traffic trends; θ
1, θ
2..., θ
n-1, θ
nrepresent the weights that the corresponding time cycle is occupied in average discharge trend calculates respectively;
S22: make moment c=mk+k1, calculate the network traffic trends error between the moment (mk+1) to the network traffic trends and average discharge trend in moment (mk+k1), formula is:
[re_tt
(i,mk+1),re_tt
(i,mk+2),...,re_tt
(i,mk+k1)]'＝
([tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'-[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]')./
[ar_tt
(i,mk+1),ar_tt
(i,mk+2),...,ar_tt
(i,mk+k1)]'
[re_tt
(i, mk+1), re_tt
(i, mk+2)..., re_tt
(i, mk+k1)] ' represent network traffic trends error;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the network traffic trends of moment (mk+1) to moment (mk+k1) successively;
[ar_tt
(i, mk+1), ar_tt
(i, mk+2)..., ar_tt
(i, mk+k1)] ' be expressed as the average discharge trend in moment (mk+1) to moment (mk+k1) successively;
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle, the value of M is:
./represent two waits the element between long vector on correspondence position to be divided by;
S23: according to the network traffic trends error of S22 gained, utilize neural network model, error between the network traffic trends in following k the moment in prediction c moment and average discharge trend, obtains the network traffic trends error in a following k moment, i.e. [pr_re_tt
(i, mk+k1+1), pr_re_tt
(i, mk+k1+2)..., pr_re_tt
(i, mk+k1+k)] ',
S24: the network traffic trends calculating a following k moment again according to the network traffic trends error in k moment of prediction and average discharge trend, formula is:
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'＝
(E+[pr_re_tt
(i,mk+k1+1),pr_re_tt
(i,mk+k1+2),...,pr_re_tt
(i,mk+k1+k)]').*
[ar_tt
(i,mk+k1+1),ar_tt
(i,mk+k1+2),...,ar_tt
(i,mk+k1+k)]'
Wherein E is the column vector that all elements is 1, and .* represents the element multiplication between two long vectors such as grade on correspondence position.
4. the network flow prediction method based on traffic trends according to claim 3, is characterized in that, in the step s 21, n gets 4, θ
1=1, θ
2=1, θ
3=0.5, θ
4=0.5, its formula is:
ar_tt
i＝(0.5*tt
i-4+0.5*tt
i-3+tt
i-2+tt
i-1)/3。
5. the network flow prediction method of traffic trends according to claim 1, is characterized in that, the Forecasting Methodology of network traffics error is:
S31: calculate the error between the moment (mk+1) to the network flow value and network traffic trends in moment (mk+k1), formula is:
[re
(i,mk+1),re
(i,mk+2),...,re
(i,mk+k1)]'＝
([tr
(i,mk+1),tr
(i,mk+2),...,tr
(i,mk+k1)]'-[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]')./
[tt
(i,mk+1),tt
(i,mk+2),...,tt
(i,mk+k1)]'
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle ./represent that two are waited the element between long vectors on correspondence position to be divided by;
[re
(i, mk+1), re
(i, mk+2)..., re
(i, mk+k1)] ' represent following error;
[tr
(i, mk+1), tr
(i, mk+2)..., tr
(i, mk+k1)] ' be expressed as the network traffics of moment (mk+1) to moment (mk+k1) successively;
[tt
(i, mk+1), tt
(i, mk+2)..., tt
(i, mk+k1)] ' be expressed as the traffic trends of moment (mk+1) to moment (mk+k1) successively;
S32: based on the flow error of neural network model study gained, and predict the network traffics error in a following k moment, i.e. [pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] '.
6. the network flow prediction method based on traffic trends according to claim 1, is characterized in that, the computational methods of predicting network flow value are:
[pr_tr
(i,mk+k1+1),pr_tr
(i,mk+k1+2),...,pr_tr
(i,mk+k1+k)]'＝
(E+[pr_re
(i,mk+k1+1),pr_re
(i,mk+k1+2),...,pr_re
(i,mk+k1+k)]').*
[pr_tt
(i,mk+k1+1),pr_tt
(i,mk+k1+2),...,pr_tt
(i,mk+k1+k)]'
Wherein m represents the number of times of circulation, described m=0,1,2 ..., M+1, described M+1 be maximum cycle,
[pr_tr
(i, mk+k1+1), pr_tr
(i, mk+k1+2)..., pr_tr
(i, mk+k1+k)] ' represent predicting network flow value;
[pr_re
(i, mk+k1+1), pr_re
(i, mk+k1+2)..., pr_re
(i, mk+k1+k)] ' represent the network traffics error predicted;
[pr_tt
(i, mk+k1+1), pr_tt
(i, mk+k1+2)..., pr_tt
(i, mk+k1+k)] ' represent the network traffic trends predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510793377.1A CN105471631B (en) | 2015-11-17 | 2015-11-17 | Network flow prediction method based on traffic trends |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510793377.1A CN105471631B (en) | 2015-11-17 | 2015-11-17 | Network flow prediction method based on traffic trends |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105471631A true CN105471631A (en) | 2016-04-06 |
| CN105471631B CN105471631B (en) | 2018-12-18 |
Family
ID=55608945
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510793377.1A Active CN105471631B (en) | 2015-11-17 | 2015-11-17 | Network flow prediction method based on traffic trends |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105471631B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106355091A (en) * | 2016-08-26 | 2017-01-25 | 西南大学 | Communication source positioning method based on biological intelligence |
| CN108880888A (en) * | 2018-06-20 | 2018-11-23 | 浙江工商大学 | A kind of SDN network method for predicting based on deep learning |
| CN109862585A (en) * | 2019-01-31 | 2019-06-07 | 湖北工业大学 | A kind of dynamic heterogeneous network flow prediction method based on depth Space-time Neural Network |
| CN109873712A (en) * | 2018-05-18 | 2019-06-11 | 新华三信息安全技术有限公司 | A kind of network flow prediction method and device |
| CN109981358A (en) * | 2019-03-13 | 2019-07-05 | 南京理工大学 | A kind of adaptive network performance method for early warning based on built-up pattern |
| CN110009016A (en) * | 2019-03-25 | 2019-07-12 | 新华三信息安全技术有限公司 | Feature extracting method and device |
| CN110621026A (en) * | 2019-02-18 | 2019-12-27 | 北京航空航天大学 | Base station flow multi-time prediction method |
| CN112202736A (en) * | 2020-09-15 | 2021-01-08 | 浙江大学 | Industrial control system communication network abnormity classification method based on statistical learning and deep learning |
| CN112398670A (en) * | 2019-08-15 | 2021-02-23 | 中国移动通信集团浙江有限公司 | Flow prediction method, flow prediction device, computing equipment and computer storage medium |
| CN113472551A (en) * | 2020-03-30 | 2021-10-01 | 中国电信股份有限公司 | Network flow prediction method, device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101695050A (en) * | 2009-10-19 | 2010-04-14 | 浪潮电子信息产业股份有限公司 | Dynamic load balancing method based on self-adapting prediction of network flow |
| CN101834753A (en) * | 2010-03-30 | 2010-09-15 | 华中科技大学 | Network flow prediction method based on covariant orthogonality and neural network |
| US7808903B2 (en) * | 2008-03-25 | 2010-10-05 | Verizon Patent And Licensing Inc. | System and method of forecasting usage of network links |
| CN102369689A (en) * | 2011-08-11 | 2012-03-07 | 华为技术有限公司 | Long-term forecasting method and device of network flow |
| CN103840988A (en) * | 2014-03-17 | 2014-06-04 | 湖州师范学院 | Network traffic measurement method based on RBF neural network |
-
2015
- 2015-11-17 CN CN201510793377.1A patent/CN105471631B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7808903B2 (en) * | 2008-03-25 | 2010-10-05 | Verizon Patent And Licensing Inc. | System and method of forecasting usage of network links |
| CN101695050A (en) * | 2009-10-19 | 2010-04-14 | 浪潮电子信息产业股份有限公司 | Dynamic load balancing method based on self-adapting prediction of network flow |
| CN101834753A (en) * | 2010-03-30 | 2010-09-15 | 华中科技大学 | Network flow prediction method based on covariant orthogonality and neural network |
| CN102369689A (en) * | 2011-08-11 | 2012-03-07 | 华为技术有限公司 | Long-term forecasting method and device of network flow |
| CN103840988A (en) * | 2014-03-17 | 2014-06-04 | 湖州师范学院 | Network traffic measurement method based on RBF neural network |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106355091B (en) * | 2016-08-26 | 2019-01-29 | 西南大学 | Propagating source localization method based on biological intelligence |
| CN106355091A (en) * | 2016-08-26 | 2017-01-25 | 西南大学 | Communication source positioning method based on biological intelligence |
| CN109873712A (en) * | 2018-05-18 | 2019-06-11 | 新华三信息安全技术有限公司 | A kind of network flow prediction method and 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 |
| CN108880888A (en) * | 2018-06-20 | 2018-11-23 | 浙江工商大学 | A kind of SDN network method for predicting based on deep learning |
| CN109862585B (en) * | 2019-01-31 | 2022-04-19 | 湖北工业大学 | Dynamic heterogeneous network flow prediction method based on deep space-time neural network |
| CN109862585A (en) * | 2019-01-31 | 2019-06-07 | 湖北工业大学 | A kind of dynamic heterogeneous network flow prediction method based on depth Space-time Neural Network |
| CN110621026A (en) * | 2019-02-18 | 2019-12-27 | 北京航空航天大学 | Base station flow multi-time prediction method |
| CN110621026B (en) * | 2019-02-18 | 2023-09-05 | 北京航空航天大学 | Multi-moment prediction method for base station flow |
| CN109981358A (en) * | 2019-03-13 | 2019-07-05 | 南京理工大学 | A kind of adaptive network performance method for early warning based on built-up pattern |
| CN110009016A (en) * | 2019-03-25 | 2019-07-12 | 新华三信息安全技术有限公司 | Feature extracting method and device |
| CN110009016B (en) * | 2019-03-25 | 2021-08-27 | 新华三信息安全技术有限公司 | Feature extraction method and device |
| CN112398670A (en) * | 2019-08-15 | 2021-02-23 | 中国移动通信集团浙江有限公司 | Flow prediction method, flow prediction device, computing equipment and computer storage medium |
| CN113472551A (en) * | 2020-03-30 | 2021-10-01 | 中国电信股份有限公司 | Network flow prediction method, device and storage medium |
| CN113472551B (en) * | 2020-03-30 | 2022-11-18 | 中国电信股份有限公司 | Network flow prediction method, device and storage medium |
| CN112202736B (en) * | 2020-09-15 | 2021-07-06 | 浙江大学 | Communication network anomaly classification method based on statistical learning and deep learning |
| CN112202736A (en) * | 2020-09-15 | 2021-01-08 | 浙江大学 | Industrial control system communication network abnormity classification method based on statistical learning and deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105471631B (en) | 2018-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105471631A (en) | Network traffic prediction method based on traffic trend | |
| Li et al. | Prediction for tourism flow based on LSTM neural network | |
| CN110851782B (en) | Network flow prediction method based on lightweight space-time deep learning model | |
| CN109816095B (en) | Network flow prediction method based on improved gated cyclic neural network | |
| CN109829577B (en) | Rail train running state prediction method based on deep neural network structure model | |
| CN109462520B (en) | Network traffic resource situation prediction method based on LSTM model | |
| CN107527091A (en) | Data processing method and device | |
| Bi et al. | Deep neural networks for predicting task time series in cloud computing systems | |
| CN108764568B (en) | Data prediction model tuning method and device based on LSTM network | |
| CN108876044B (en) | Online content popularity prediction method based on knowledge-enhanced neural network | |
| CN113852432B (en) | Spectrum Prediction Sensing Method Based on RCS-GRU Model | |
| CN112910711B (en) | Wireless service flow prediction method, device and medium based on self-attention convolutional network | |
| Petruseva et al. | Neural network prediction model for construction project duration | |
| CN107563801A (en) | Consumer behavior Forecasting Methodology under a kind of consumer's line based on hidden Markov model | |
| CN103942461A (en) | Water quality parameter prediction method based on online sequential extreme learning machine | |
| CN112631717A (en) | Network service function chain dynamic deployment system and method based on asynchronous reinforcement learning | |
| CN106533750A (en) | System and method for predicting non-steady application user concurrency in cloud environment | |
| Alfred | A genetic-based backpropagation neural network for forecasting in time-series data | |
| CN107230349A (en) | A kind of online real-time short time traffic flow forecasting method | |
| CN107704426A (en) | Water level prediction method based on extension wavelet-neural network model | |
| CN111158237A (en) | Industrial furnace temperature multi-step prediction control method based on neural network | |
| CN106709588A (en) | Prediction model construction method and equipment and real-time prediction method and equipment | |
| CN116882708B (en) | Steel process flow control method and device based on digital twin and related equipment | |
| CN101834753A (en) | Network flow prediction method based on covariant orthogonality and neural network | |
| CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium |
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 |