CN109120463A - Method for predicting and device - Google Patents
Method for predicting and device Download PDFInfo
- Publication number
- CN109120463A CN109120463A CN201811195940.5A CN201811195940A CN109120463A CN 109120463 A CN109120463 A CN 109120463A CN 201811195940 A CN201811195940 A CN 201811195940A CN 109120463 A CN109120463 A CN 109120463A
- Authority
- CN
- China
- Prior art keywords
- moment
- flow rate
- data
- predicted flow
- flows
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/04—Processing captured monitoring data, e.g. for logfile generation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0876—Network utilisation, e.g. volume of load or congestion level
Abstract
This disclosure relates to a kind of method for predicting and device, comprising: obtain data on flows of the target network in first time period;The data on flows is pre-processed, pretreated data on flows is obtained;According to the pretreated data on flows and the first model, the first predicted flow rate data in second time period are obtained;The corresponding second predicted flow rate data of the first predicted flow rate data are obtained according to the first predicted flow rate data and the second model;According to the predicted flow rate data in second time period described in the first predicted flow rate data and the second predicted flow rate data acquisition.The method for predicting and device of the disclosure, may be implemented the Accurate Prediction of the data on flows of future time section.
Description
Technical field
This disclosure relates to field of computer technology more particularly to a kind of method for predicting and device.
Background technique
With the rapid development of network technology, the business and application carried on network is more and more abundant.It is provided in internet
Service increasingly diversification and complicate while, the pressure of network links carry is also increasing, therefore, alleviate network link
The pressure of carrying, which seems, to be even more important.
In order to alleviate the pressure of network links carry, enterprise needs to understand the business carried in link in time, grasps link
Traffic characteristic, in advance to risk carry out control, to solve network bursting problem in time.
Summary of the invention
In view of this, the present disclosure proposes a kind of method for predicting and device, for network following a period of time
Data on flows is predicted.
According to one aspect of the disclosure, a kind of method for predicting is proposed, which comprises
Obtain data on flows of the target network in first time period;
The data on flows is pre-processed, pretreated data on flows is obtained;
According to the pretreated data on flows and the first model, the first predicted flow rate number in second time period is obtained
According to;
The first predicted flow rate data corresponding second are obtained according to the first predicted flow rate data and the second model
Predicted flow rate data;
According in second time period described in the first predicted flow rate data and the second predicted flow rate data acquisition
Predicted flow rate data.
According to another aspect of the disclosure, a kind of volume forecasting device is proposed, described device includes:
Module is obtained, for obtaining data on flows of the target network in first time period;
Preprocessing module is connected to the acquisition module, for pre-processing to the data on flows, obtains pretreatment
Data on flows afterwards;
First computing module is connected to the preprocessing module, for according to the pretreated data on flows and the
One model obtains the first predicted flow rate data in second time period;
Second computing module is connected to first computing module, for according to the first predicted flow rate data and the
Two models obtain the corresponding second predicted flow rate data of the first predicted flow rate data;
Third computing module is connected to second computing module, for according to the first predicted flow rate data and institute
State the predicted flow rate data in second time period described in the second predicted flow rate data acquisition.
According to another aspect of the present disclosure, a kind of volume forecasting system is provided, comprising: processor;It is handled for storage
The memory of device executable instruction;Wherein, the processor is configured to executing above-mentioned method for predicting.
According to another aspect of the present disclosure, a kind of non-volatile computer readable storage medium storing program for executing is provided, is stored thereon with
Computer program instructions, wherein the computer program instructions realize above-mentioned method for predicting when being executed by processor.
The method for predicting of the disclosure, by obtaining data on flows of the target network in first time period, to described
Data on flows is pre-processed, and pretreated data on flows is obtained, according to the pretreated data on flows and the first mould
Type obtains the first predicted flow rate data in second time period, is obtained according to the first predicted flow rate data and the second model
The corresponding second predicted flow rate data of the first predicted flow rate data, according to the first predicted flow rate data and described second
Predicted flow rate data in second time period described in predicted flow rate data acquisition can carry out the data on flows in future time
Accurate prediction.
According to below with reference to the accompanying drawings to detailed description of illustrative embodiments, the other feature and aspect of the disclosure will become
It is clear.
Detailed description of the invention
Comprising in the description and constituting the attached drawing of part of specification and specification together illustrates the disclosure
Exemplary embodiment, feature and aspect, and for explaining the principles of this disclosure.
Fig. 1 shows the flow chart of the method for predicting according to one embodiment of the disclosure.
Fig. 2 shows the flow charts according to the data prediction of the disclosure one embodiment.
Fig. 3 shows the prediction flow diagram of the first model according to one embodiment of the disclosure.
Fig. 4 shows the acquisition flow diagram of the second predicted flow rate according to one embodiment of the disclosure.
Fig. 5 shows the schematic diagram of the data on flows in the first time period according to one embodiment of the disclosure.
Fig. 6 shows the curve synoptic diagram of predicted value and true value according to one embodiment of the disclosure.
Fig. 7 shows the block diagram of the volume forecasting device according to one embodiment of the disclosure.
Fig. 8 shows the block diagram of the volume forecasting device according to one embodiment of the disclosure.
Fig. 9 shows the block diagram of the volume forecasting system according to one embodiment of the disclosure.
Specific embodiment
Various exemplary embodiments, feature and the aspect of the disclosure are described in detail below with reference to attached drawing.It is identical in attached drawing
Appended drawing reference indicate element functionally identical or similar.Although the various aspects of embodiment are shown in the attached drawings, remove
It non-specifically points out, it is not necessary to attached drawing drawn to scale.
Dedicated word " exemplary " means " being used as example, embodiment or illustrative " herein.Here as " exemplary "
Illustrated any embodiment should not necessarily be construed as preferred or advantageous over other embodiments.
In addition, giving numerous details in specific embodiment below to better illustrate the disclosure.
It will be appreciated by those skilled in the art that without certain details, the disclosure equally be can be implemented.In some instances, for
Method, means, element and circuit well known to those skilled in the art are not described in detail, in order to highlight the purport of the disclosure.
In order to realize the intelligent management of network, it is crucial for carrying out prediction to link flow.
By the measurement and prediction to link flow, not only it will be seen that traffic conditions and trend between link, thus
It is more effectively carried out link optimizing, preferably carries out the design of routing Design and load balancing, and can be with determined link congestion
Control can thus reduce because link congestion bring information is lost and is postponed, make full use of link circuit resource, improve Service Quality
Amount.
The real-time monitoring of link flow is an importance of network management.When abnormal conditions occurs in flow, in net
After network management system issues alarm notification, the problem is addressed by network management personnel, only a kind of row of response type
For the mode of that is, first problematic post-processing.Such mode, it is likely that due to having insufficient time to analysis processing appearance
Problem, and then influence the normal operation of network.If can predicted flow rate overload, by preparatory network managing mode, flowing
Amount overload is analyzed and is solved the problems, such as before occurring, and the availability of link can be significantly improved.
Flux prediction model is the basis for carrying out link performance analysis and link planning and designing, a good chain volume forecasting
Model and prediction technique are equal to design new generation network agreement, network management and diagnosis, the high performance router of design and load
The network hardware equipments such as weighing apparatus and the service quality for improving network are all significant.With the increase of network bandwidth and each
The appearance of kind network service, traditional flux prediction model are also difficult to meet to existing and network flow in future accurate description
And prediction, therefore, for traditional flux prediction model and the limitation of prediction technique, provided for link flow more accurate
Flux prediction model and prediction technique seem very necessary.
Referring to Fig. 1, Fig. 1 shows the flow chart of the method for predicting according to one embodiment of the disclosure.
The method can be applied to server and perhaps be executed by server or terminal to target network in terminal
Flow is predicted.As shown in Figure 1, the method may include:
Step S110 obtains data on flows of the target network in first time period;
Step S120 pre-processes the data on flows, obtains pretreated data on flows;
Step S130 obtains first in second time period according to the pretreated data on flows and the first model
Predicted flow rate data;
Step S140 obtains the first predicted flow rate data pair according to the first predicted flow rate data and the second model
The the second predicted flow rate data answered;
Step S150, when second according to the first predicted flow rate data and the second predicted flow rate data acquisition
Between predicted flow rate data in section.
By above method, the disclosure can according to the first model and the second model to the data on flows of future time section into
Row Accurate Prediction.
For step S110:
In a kind of possible embodiment, target network can be the data source for obtaining data on flows, for example, it may be
The data center at some enterprise network management center is also possible to other data sources.
In a kind of possible embodiment, first time period can be the time of any duration such as 12 hours, 1 day, 1 week
Section.
In a kind of possible embodiment, the data on flows of target network can be obtained by network flow collector,
For example, mesh can be acquired by Simple Network Management Protocol (SNMP, Simple Network Management Protocol)
The data on flows of network is marked, SNMP is the network management standard based on TCP/IP protocol suite, is that one kind manages net in an ip network
The standard agreement of network node (such as server, work station, router, interchanger).
The network of snmp management mainly consists of three parts: equipment, SNMP agent and the Network Management System being managed
(NMS), there are a management information banks (MIB) for collecting simultaneously storage management letter for each equipment being managed in network
Breath.By snmp protocol, NMS can obtain these information.Typically, the data on flows of acquisition in SNMP each minute, and deposit
Enter in database, which can be the data on flows library of special storage data on flows.Acquire the data on flows of target network
Multilink can be divided into be acquired, to each of the links, SNMP can be obtained respectively into direction and the data on flows in direction out, enters
Direction can indicate to acquire the byte number for being injected into the link in certain a period of time, i.e. downlink traffic;Direction can indicate to adopt out
Collect the byte number that the link is left in certain a period of time, i.e. uplink traffic.
In a kind of possible embodiment, the data on flows of acquisition may include link ID, acquisition time, flow value
Deng.Link ID can be used for identifying which link is data on flows belong to, and acquisition time can be used for identifying the flow number of acquisition
According to be in which time (such as SNMP obtains data on flows with minute, this time identifier of 2018/3/23 19:04 is at this
Minute in obtain data on flows), flow value can be used for identifying the data on flows obtained in acquisition time number.
For step S120:
Multiple link IDs are generally included by the data on flows that network flow collector obtains, that is to say, that the stream of acquisition
It is rambling for measuring data, and there may be exceptional values and missing values for certain fields in the data on flows obtained, for example, certain
The value of the value of a little fields bigger than normal perhaps obvious less than normal or certain fields obvious compared to the value of critical field is negative value or is
It is empty.These are disorderly and unsystematic, and even wrong data on flows can generate interference to volume forecasting, will affect the essence of volume forecasting
Degree, in some instances it may even be possible to which the mistake that will lead to prediction result improves volume forecasting therefore, it is necessary to pre-process to data on flows
Accuracy.
Referring to Figure 2 together, Fig. 2 shows the flow charts according to the data prediction of the disclosure one embodiment.
As shown in Fig. 2, in step S210, the data on flows is pre-processed in a kind of possible embodiment,
Pretreated data on flows is obtained, may include:
Step S121 carries out data cleansing to the data on flows.
Data cleansing refers to discovery and corrects one of program of identifiable mistake in data file, including checks data one
Cause property handles invalid value and missing values etc..
Wherein, the invalid value can be greater than first threshold for flow value in the data on flows or less than second threshold
Data on flows, the missing values can be the data on flows of flow value missing in the data on flows.
For invalid value, in a kind of possible embodiment, the data on flows comprising invalid value few for quantity,
It can directly delete, in other embodiments, the abnormal flow data can also rationally be replaced using other methods.
It for example, can be by the way of rationally replacing, to avoid exception when the quantity of abnormal flow data is greater than amount threshold
Data on flows deletes excessive caused prediction result interference.
For missing values, in one possible implementation, can use after nearest neighbor interpolation method fills up the cleaning
Data on flows in missing values.
Using nearest neighbor interpolation method, the distance between each data on flows can be defined, between the data on flows away from
From the similarity degree that can reflect each data on flows, removed in determining data on flows and the data on flows comprising missing values
After the distance of other datas on flows other than data on flows comprising missing values, can using in other datas on flows with comprising
The data on flows of missing values carries out interpolation to missing values as interpolation value apart from the smallest data on flows.
Step S122 is grouped the data on flows according to the link ID of the data on flows after data cleansing, obtains
The data on flows of same link ID.
When the data on flows to following a period of time is predicted, usually the data on flows of a certain link is carried out pre-
It surveys, when obtaining data on flows, the data on flows of acquisition frequently includes multiple link IDs, therefore can be according to different link IDs
Data on flows is grouped, with according to the different corresponding datas on flows of link ID respectively to the data on flows of each link into
Row prediction.
For example, table 1 shows the data on flows that link ID is 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c.
Table 1
Serial number | Link ID | Acquisition time | Downlink traffic | Uplink traffic |
1 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:04 | 5442844 | 3697609 |
2 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:05 | 6164659 | 4120042 |
3 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:06 | 6454498 | 4404634 |
4 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:07 | 5974726 | 4481856 |
5 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:08 | 5370612 | 3462507 |
6 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:09 | 5887826 | 4155548 |
7 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:10 | 6184585 | 4297634 |
8 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:11 | 6589378 | 4558746 |
9 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:12 | 6412367 | 4351408 |
10 | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 19:13 | 5932734 | 4010300 |
… | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | … | … | … |
… | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 20:00 | 5765245 | 4053216 |
… | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | 2018/3/23 20:01 | 6305452 | 4356452 |
… | 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c | … | … | … |
It can be seen that the data traffic of the link ID includes uplink traffic, downlink traffic and corresponding acquisition time.It should be understood that
The acquisition time include a time point and the time point before a period, for example, in table 1 band acquisition time be each
From one minute in time point.It can be " 51c5a002-c8cc-4f16-a45d- to link ID using the data on flows in table 1
The link of 6daf3bd68a4c " carries out volume forecasting, likewise, corresponding stream can be respectively adopted for the link of other ID
It measures data and carries out volume forecasting.
The first time period is divided into multiple periods, wherein in the first time period after division by step S123
Multiple periods are arranged in order.
In a kind of possible embodiment, the duration of each period can be 1 hour, 2 in the multiple period
Any durations such as hour.
The time interval that network flow collector obtains data on flows is usually 1 minute, in this case, time granularity
It is smaller, when need to predict in one day 24 it is small when data on flows when, be interval according to 1 minute, then possibly can not embody flow
Therefore the integral status of data can set the duration of the multiple period to 1 hour.
Step S124 obtains the corresponding data on flows of each period under same link ID in first time period, wherein
Each period, corresponding data on flows respectively included at least one flow value.
At same link ID, each period in multiple periods includes the data on flows of multiple 1 minutes sections,
Multiple 1 minute data on flows were corresponded under the correspondence period being integrated into each multiple periods, so as to subsequent processing.
For example, by each period when it is 1 hour a length of for, as shown in table 1, can obtain link ID is
Each period under " 51c5a002-c8cc-4f16-a45d-6daf3bd68a4c " in multiple periods is multiple 1 points corresponding
The data on flows of clock time section, for example, the data on flows conduct of available 2018/3/2319:01-2018/3/23 20:00
Correspond to period corresponding data on flows.
Step S125, according at least one corresponding flow value of each period in first time period, when calculating first
Between in each period in section same link ID flow average value.
In the present embodiment, multiple 1 points of same link ID in each period in multiple periods can be sought
The average value of flow value in clock, using the average value of same link ID in each period as the same link ID corresponding time
Flow value in section.
In other embodiments, the multiple of same link ID can be summarized in the hope of each period in multiple periods
The summation of flow value in 1 minute, using the summation of the flow value as the flow value of same link ID corresponding period.
Please refer to table 2, the stream that it is 1 hour with each period in multiple periods under same link ID that table 2, which is shown,
Measure schematic diagram of the data in 1 day.
Table 2
From table 2 it can be seen that in 24 hours, the data on flows of the link ID includes uplink traffic, downlink traffic and right
The acquisition time answered.
Step S126, by each period in link ID, the link ID corresponding multiple periods and multiple periods
Corresponding flow average value is as the pretreated data on flows.
By to pretreated data on flows removal exceptional value, interpolation missing values, the flow number for being grouped different link IDs
According to the precision of volume forecasting can be improved as the data on flows that prediction uses for pretreated data on flows.
In a kind of possible embodiment, pretreated data on flows include in first time period the 1st moment arrive
The n data on flows (multiple periods corresponding data on flows) at n-th of moment.
For step S130:
In a kind of possible embodiment, first model may include shot and long term memory network model LSTM
(Long-Short Term Memory, LSTM).
Shot and long term memory network model LSTM is a kind of improved Recognition with Recurrent Neural Network structure, including multiple LSTM units,
It can solve long-term Dependence Problem.In LSTM model, each LSTM unit includes 3 control door controls, 3 control door difference
It is to forget door, input gate and out gate, controls state, input and output in LSTM unit respectively.The eucaryotic cell structure of LSTM
After (LSTM unit) receives input information, each carries out operation according to respective feature, to the input information of separate sources
And screening, determine which input information can pass through.The effect for forgeing door is to make Recognition with Recurrent Neural Network otiose before forgetting
Information, meanwhile, after the transformation of nonlinear function, the two is overlapped to form new status information for the input of input gate.It is defeated
The effect gone out is the output that cell factory generates current time after new state is calculated.The weight of LSTM model can be with
Learnt by training process, LSTM allows information selectively to influence each moment in Recognition with Recurrent Neural Network according to the structure of " door "
State.
LSTM is as follows to the principle of the data on flows prediction of future time:
The forgetting door is obtained according to the following formula in tiThe forgetting data at moment:
Wherein,For tiThe input parameter at moment,For ti-1When
The output at quarter as a result,For ti-1The state parameter at moment, WxfForPreset weight matrix, WhfForPreset power
Value matrix, WcfForPreset weight matrix, bfFor preset first bias vector,For tiThe forgetting data at moment;
The input gate is obtained according to the following formula in tiThe increase parameter at moment:
Wherein, WxdFor tiThe input parameter at momentIt is default
Weight matrix, WhdForPreset weight matrix, WcdForPreset weight matrix, bdFor preset second biasing
Vector,For tiThe increase parameter at moment;
The out gate t is obtained according to the following formulaiThe state parameter at moment:
Wherein, WxcFor tiThe input parameter at momentPreset power
Value matrix, WhcForPreset weight matrix, bcFor preset third bias vector,For tiThe state parameter at moment;
The output parameter of the out gate is obtained according to the following formula:
Wherein, WxoFor tiThe input parameter at momentIt is preset
Weight matrix, WhoForPreset weight matrix, WcoFor state parameterPreset weight matrix, boIt is preset
Four bias vectors,For tiThe output parameter of out gate described in moment.
T is obtained according to the following formulaiThe output result at moment:
Wherein,For tiThe output result at moment.
T is obtained according to the following formulai+1The predicted flow rate data at moment:
It is to be understood that can use the error that loss function determines LSTM model, LSTM is carried out according to the error
Reverse train is to optimize the LSTM model parameter such as weight matrix.In the present embodiment, the loss of neural network can be defined
Function are as follows:
Wherein, prejIndicate j-th of predicted flow rate, yjIndicate j-th of actual flow.
Referring to Fig. 3, Fig. 3 shows the prediction flow diagram of the first model according to one embodiment of the disclosure.
As shown in figure 3, the pre- flow gauge of the first model includes the following steps:
Step S131, according to the data on flows at i-th of moment in the flow at the n continuous moment, (i-1)-th moment
The output of LSTM, (i-1)-th moment LSTM state parameter, obtain the output of i-th of moment LSTM, i-th moment LSTM
The predicted flow rate of state parameter and i+1 moment;Wherein, 1≤i≤n;
As the i=1, the output of (i-1)-th moment LSTM, (i-1)-th moment LSTM state parameter be respectively with
Machine value or 0;
As the i=n, the predicted flow rate at i+1 moment is first prediction at (n+1)th moment in second time period
Flow.
Step S132, according to the predicted flow rate at m-th of moment in second time period, the output of the m-1 moment LSTM,
The state parameter of m-1 moment LSTM obtains the output of m-th of moment LSTM, the state parameter and m+ of m-th moment LSTM
The predicted flow rate at 1 moment;Wherein, n+1≤m≤2n.
Below by the prediction process of the first predicted flow rate data being illustrated to step S130, it should be appreciated that
It is that citing is exemplary below, is not limited to the disclosure.
Assuming that the data of acquisition include on September 29th, 2018 n=3 moment (t1、t2、t3) flow:
t1=2018/9/29 00:00, data on flows are
t2=2018/9/29 08:00, data on flows are
t3=2018/9/29 16:00, data on flows are
For step S131:
Step S131 can be the training process of the first model, and after the training of step S131, the first model can be used
In the prediction for carrying out the first predicted flow rate data.
Firstly, input t1The actual flow at momentThe output of a upper moment modelThe shape of a upper moment model
State parameterIt should be noted that the initial stageLSTM model can be arrived for random value or 0, LSTM model carries out
Following operation is to start the training of model;
Forget door output:
Input gate output:
Out gate output:
Hidden layer output:
t2The predicted flow rate at moment are as follows:
Then, t is inputted2The actual flow at momentThe output of a upper moment modelThe shape of a upper moment model
State parameterTo LSTM model, LSTM model carries out following operation to continue to be trained model:
Forget door output:
Input gate output:
Out gate output:
Hidden layer output:
t3The predicted flow rate at moment are as follows:
Through above procedure it is found that in LSTM model, the actual flow for inputting some moment is available next
The predicted flow rate at moment.Assuming that acquire the flow at n moment, then it is available in addition to n moment by model training stage
In first moment other moment predicted flow rate, can have N in the sequence using these predicted flow rates as a sequence
(then N=n-1) a predicted flow rate, accordingly can be using the corresponding actual flow of these predicted flow rates as a sequence, the sequence
There is N number of actual flow in column, then error can be calculated according to this N number of actual flow and predicted flow rate, error gets over large-sized model essence
Degree is lower, and according to the error being calculated, each weight and biasing in adjustable model are duplicate to execute above-mentioned trained
Journey, until error in reasonable range, then carrys out predicted flow rate using the corresponding model of the error.
For step S132:
After completing to the training of LSTM model, can use trained LSTM model carry out future time section (such as
Second time period) data on flows prediction.
Still using above-mentioned n=3 (t1、t2、t3) be illustrated for the moment, and with above-mentioned moment and corresponding flow
Data predict on September 30th, 2018 corresponding 3 (i.e. n+1~2n) moment (t4: 2018/9/30 00:00, t5: 2018/9/30
08:00、t6: 2018/9/30 16:00) flow, here, t3、t4Moment can be continuously, be also possible to discontinuous.
After LSTM model completes training, it is identical with the calculating process of above-mentioned training process to can use the progress of LSTM model
Operation, to obtain t4The pre- flow measurement at moment (i.e. m=n+1 moment in the (n+1)th~the 2n moment, when n takes 3, m=4)
Amount, step may include:
Input t1The available t of the actual flow at moment2The predicted flow rate at momentThe output of hidden layerState
Parameter
Input t2The available t of the actual flow at moment3The predicted flow rate at momentThe output of hidden layerState
Parameter
Input t3The available t of the actual flow at moment4The predicted flow rate at momentt3The output of the hidden layer at momentt3The state parameter at moment
Utilize t4The predicted flow rate at momentt3The output of the hidden layer at momentt3The state parameter at momentIt can
To be achieved by the steps of the prediction of the first predicted flow rate data:
Firstly, input t4The pre- flow measurement at moment (the m=n+1 moment, n+1≤m≤2n in the (n+1)th~the 2n moment)
Amountt3The output of moment LSTM modelt3The state parameter of moment modelTo LSTM model, same step is executed
The similar calculating process of S131.Pass through the available t of above-mentioned calculating5The first predicted flow rate data at momentt4Shi Kemo
The output of typeThe state parameter of t4 moment modelIt should be noted that in the training processIndicate tiThe reality at moment
Border flow, during predictionIndicate tiThe predicted flow rate at moment, i are the integer more than or equal to 1.
Then, t is inputted5The first predicted flow rate data at momentt4The output of moment modelt4Moment model
State parameterTo LSTM model, above-mentioned similar calculating process, available t are executed6The first predicted flow rate data at momentt5The output of moment modelThe state parameter of t5 moment model
When also to carry out the volume forecasting at other moment, by above step, operation is successively carried out, can be obtained other
The first predicted flow rate data.
For step S140:
In a kind of possible embodiment, the second model can be bi-exponential smoothing model (Second
Exponential Smoothing, SES).
SES model is established based on bi-exponential smoothing method.Bi-exponential smoothing method can be to an index
The method that smooth value makees exponential smoothing again, it can cooperate with Single Exponential Smoothing, establish the mathematical model of prediction, so
Predicted value is determined with mathematical model afterwards.
The first predicted flow rate data that SES model can export LSTM model, which are smoothed, obtains the second pre- flow measurement
Amount.
First one can be obtained by first predicted flow rate at the (n+1)th~the 2n above-mentioned moment according to time-sequencing
Sequence then includes first predicted flow rate at n moment in the sequence, the first pre- flow measurement at i-th of moment in sequence is indicated with zi
Amount.
The principle of operation of SES model may include:
First Smoothness Index is obtained by following formula:
Wherein, α is the decimal between 0~1,First for i-th of moment in the sequence is flat
Sliding index,For first Smoothness Index at (i-1)-th moment in sequence;When i=1,For first value in the sequence
Or the average value of multiple values;
Second Smoothness Index is obtained by following formula:
Wherein,For second Smoothness Index at i-th of moment in the sequence,For sequence
In (i-1)-th moment the second Smoothness Index, as i=1,For the average value of a value or multiple values value in sequence;
The second predicted flow rate data Z at i-th of moment in sequence is obtained by following formulai:
Zi=Ai-BiT
Wherein, T is the issue of prediction.
Illustrated referring to Fig. 4, Fig. 4 is shown according to the acquisition process of the second predicted flow rate of one embodiment of the disclosure
Figure.
As shown in figure 4, the step of obtaining the second predicted flow rate may include:
Step S141, for (n+1)th moment in the second time period to any moment in the 2n moment
First predicted flow rate carries out bi-exponential smoothing processing;
Step S142 determines the second pre- of the moment according to the bi-exponential smoothing processing result of the predicted flow rate at the moment
Measurement of discharge data;
Step S143, the second predicted flow rate data include (n+1)th moment to the 2n moment in second time period
The second predicted flow rate.
The SES operation for carrying out the second predicted flow rate data is introduced below in conjunction with specific example, it is clear that
, once it is described as illustratively, being not limited to the disclosure.
The first predicted flow rate that prediction can be obtained obtains a sequence, is indicated in sequence with zi according to time-sequencing
The flow at i-th of moment.For example, the t predicted with step S1304、t5、t6For the predicted flow rate at moment, when with this 3
The flow at quarter forms a new sequence, includes the predicted flow rate at 3 moment in the sequence, is followed successively by z1、z2、z3, then:
Using can following formula to the first predicted flow rate data (z of the 1st moment (i.e. 2018/9/30 00:00)1)
Carry out bi-exponential smoothing processing:
Wherein, smoothing factor α value (0,1),It can be first value z in sequence in the initial stage1,
It can be the average value of all values in sequence;
The second predicted flow rate data can be obtained by following formula:
Z1=A1-B1T;
Wherein, T is the issue of prediction, for example, predicted according to No. 29 three groups of flows (flow of t1 to t3 moment) 30
Three group of first predicted flow rate (flow of t4 to t6 moment) at number corresponding moment, since t3 to the t4 moment is continuously, T can
1 is thought, if three groups of flows (flow of t1 to t3 moment) according to No. 29 predict three groups of flows at No. 31 corresponding moment, T
It can be 3, and so on.
Same as described above to the predicted flow rate progress bi-exponential smoothing processing at subsequent each moment, details are not described herein again.
For step S150:
In a kind of possible embodiment, the first predicted flow rate data and second predicted flow rate can be calculated
The arithmetic mean of instantaneous value of data;And using the arithmetic mean of instantaneous value as the predicted flow rate data.
For example, z is being obtained1、z2、z3Corresponding smoothing processing result is respectively as follows: Z1、Z2、Z3When, it can use prediction
The available final predicted flow rate data of flow smoothing processing result corresponding with its, such as the average value of the two can be taken,
z1And Z1Average value be sequence in the 1st moment final predicted flow rate data.It is to be understood that above-mentioned predicted flow rate
Data can be flow value.
It should be noted that obtaining second time period above according to the first predicted flow rate data and the second predicted flow rate data
The mode of interior predicted flow rate data is only an example of the disclosure, does not limit the disclosure in any way.This field skill
Art personnel also can according to need the predicted flow rate data obtained in second time period using other modes.
The prediction effect of the method for predicting described in the disclosure of illustrating below is illustrated.
It can be to (the second time included multiple periods more, each within 24 hours on the day of predicted time section (second time period)
A period when it is 1 hour a length of) data on flows predicted that such as predicted time section can be (example on March 31st, 2018
Such as, it can be 2018/3/31 0:00-2018/3/31 23:00).
The data on flows of first time period before acquiring predicted time section by network flow collector, wherein when first
Between section can be on March 30,17 days to 2018 March in 2018, referring to Fig. 5, Fig. 5 is shown according to one embodiment party of the disclosure
The schematic diagram of data on flows in the first time period of formula.
The data on flows that will acquire obtains pretreated data on flows by pretreatment, utilizes the first model and the second mould
Type and pretreated data on flows predict the flow on 24 hours same day of on March 31st, 2018.
Also referring to table 3 and Fig. 6, the case where table 3 shows predicted value and true value, Fig. 6 is shown according to the disclosure
The predicted value of one embodiment and the curve synoptic diagram of true value.
Table 3
It can be seen that from table 3 and Fig. 6 and differed by the data on flows that the method for predicting of the disclosure is predicted with true value
Less, for the data on flows of prediction close to true value, accuracy is very high.
The method for predicting of the disclosure can accurately predict the data on flows in future time.
Referring to Fig. 7, Fig. 7 shows the block diagram of the volume forecasting device according to one embodiment of the disclosure.
As shown in fig. 7, described device includes:
Module 10 is obtained, for obtaining data on flows of the target network in first time period;
Preprocessing module 20 is connected to the acquisition module 10, for pre-processing to the data on flows, obtains pre-
Data on flows that treated;
First computing module 30 is connected to the preprocessing module 20, for according to the pretreated data on flows
And first model, obtain the first predicted flow rate data in second time period;
Second computing module 40 is connected to first computing module 30, for according to the first predicted flow rate data
The second predicted flow rate data corresponding with the second model acquisition the first predicted flow rate data;
Third computing module 50 is connected to second computing module 40, for according to the first predicted flow rate data
And the predicted flow rate data in second time period described in the second predicted flow rate data acquisition.
Referring to Fig. 8, Fig. 8 shows the block diagram of the volume forecasting device according to one embodiment of the disclosure.
As shown in figure 8, described device includes obtaining module 10, preprocessing module 20, the first computing module 30, the second operation
Module 40, third computing module 50.
In a kind of possible embodiment, first model is shot and long term memory network model LSTM;
In a kind of possible embodiment, the pretreated data on flows includes the 1st moment in first time period
To the n flow at n-th of moment.
In a kind of possible embodiment, the first predicted flow rate of second period includes arriving at (n+1)th moment
First predicted flow rate at the 2n moment.
In a kind of possible embodiment, first computing module may include:
Training submodule 310, for i-th of moment in the flow according to the n continuous moment data on flows, the
The output of i-1 moment LSTM, (i-1)-th moment LSTM state parameter, obtain i-th of moment LSTM output, i-th when
Carve the state parameter and the predicted flow rate at i+1 moment of LSTM;Wherein, 1≤i≤n;
As the i=1, the output of (i-1)-th moment LSTM, (i-1)-th moment LSTM state parameter be respectively with
Machine value or 0;
As the i=n, the predicted flow rate at i+1 moment is first prediction at (n+1)th moment in second time period
Flow.
First operation submodule 320 is connected to the trained submodule 310, when for according to m-th in second time period
The predicted flow rate at quarter, the output of the m-1 moment LSTM, the m-1 moment LSTM state parameter, obtain m-th of moment
The output of LSTM, m-th moment LSTM state parameter and the predicted flow rate at the m+1 moment;Wherein, n+1≤m≤2n.
In a kind of possible embodiment, second model is SES model.
In a kind of possible embodiment, second computing module may include:
Smooth submodule 410, for the appointing into the 2n moment for (n+1)th moment in the second time period
First predicted flow rate at one moment carries out bi-exponential smoothing processing;
It determines submodule 420, is connected to the smooth submodule 410, for the second order according to the predicted flow rate at the moment
Exponential smoothing processing result determines the second predicted flow rate data at the moment;
Second operation submodule 430 is connected to the determining submodule 420, is used for the second predicted flow rate data packet
Include second predicted flow rate at (n+1)th moment to the 2n moment in second time period.
In a kind of possible embodiment, the third computing module may include:
Average submodule 510, for calculating the first predicted flow rate of any moment and the arithmetic average of the second predicted flow rate
Value;
Third operation submodule 520 is connected to the submodule 510 that is averaged, for using the arithmetic mean of instantaneous value as this
The predicted flow rate data at moment.
By the volume forecasting device of the disclosure, the flow in following a period of time can be carried out at prediction and analysis
Reason, to realize the actively monitoring and intelligent management to network flow.
For example, by predicting as a result, the approximate trend of available flow in future works as prediction if giving a threshold value
Result when exceeding given threshold value, issue alarm, so that network management personnel can preview network state, it can be found that
Potential attack and intrusion behavior, realize network invasion monitoring.In addition, passing through the prediction to data on flows, it will be appreciated that network
Between traffic conditions and trend, to be more effectively carried out the network optimization, preferably progress routing Design and load balancing
Design etc..
Referring to Fig. 9, Fig. 9 shows the block diagram of the volume forecasting system 900 according to one embodiment of the disclosure.
Referring to Fig. 9, which may include processor 901, machine readable storage Jie for being stored with machine-executable instruction
Matter 902.Processor 901 can be communicated with machine readable storage medium 902 via system bus 903.Also, processor 901 passes through
Machine-executable instruction corresponding with volume forecasting logic is in read machine readable storage medium storing program for executing 902 to execute stream described above
Measure prediction technique.
Machine readable storage medium 902 referred to herein can be any electronics, magnetism, optics or other physical stores
System may include or store information, such as executable instruction, data, etc..For example, machine readable storage medium may is that
RAM (Radom Access Memory, random access memory), volatile memory, nonvolatile memory, flash memory, storage are driven
Dynamic device (such as hard disk drive), solid state hard disk, any kind of storage dish (such as CD, dvd) or similar storage are situated between
Matter or their combination.
The presently disclosed embodiments is described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or technological improvement to the technology in market for best explaining each embodiment, or lead this technology
Other those of ordinary skill in domain can understand each embodiment disclosed herein.
Claims (12)
1. a kind of method for predicting, which is characterized in that the described method includes:
Obtain data on flows of the target network in first time period;
The data on flows is pre-processed, pretreated data on flows is obtained;
According to the pretreated data on flows and the first model, the first predicted flow rate data in second time period are obtained;
Corresponding second prediction of the first predicted flow rate data is obtained according to the first predicted flow rate data and the second model
Data on flows;
According to the prediction in second time period described in the first predicted flow rate data and the second predicted flow rate data acquisition
Data on flows.
2. the method according to claim 1, wherein the pretreated data on flows includes first time period
The n flow at interior 1st moment to n-th of moment;
First predicted flow rate of the second time period includes first predicted flow rate at (n+1)th moment to the 2n moment.
3. according to the method described in claim 2, it is characterized in that, first model is shot and long term memory network model
LSTM;
It is described according to the pretreated data on flows and the first model, obtain the first predicted flow rate number in second time period
According to, comprising:
According to the data on flows at i-th of moment in the flow at the n continuous moment, the output of (i-1)-th moment LSTM, the
The state parameter of i-1 moment LSTM, obtain the output of i-th of moment LSTM, the state parameter of i-th moment LSTM and i-th+
The predicted flow rate at 1 moment;Wherein, 1≤i≤n;
As the i=1, the output of (i-1)-th moment LSTM, (i-1)-th moment LSTM state parameter be respectively random value
Or 0;
As the i=n, the predicted flow rate at i+1 moment is the first pre- flow measurement at (n+1)th moment in second time period
Amount.
4. according to the method described in claim 3, it is characterized in that, described according to the pretreated data on flows and first
Model obtains the first predicted flow rate data in the second time period, further includes:
According to the predicted flow rate at m-th of moment in second time period, the output of the m-1 moment LSTM, the m-1 moment LSTM
State parameter, obtain the output of m-th of moment LSTM, the state parameter of m-th moment LSTM and the prediction at the m+1 moment
Flow;Wherein, n+1≤m≤2n.
5. according to the method described in claim 2, it is characterized in that, second model is SES model;
Corresponding second prediction of the first predicted flow rate data is obtained according to the first predicted flow rate data and the second model
Data on flows, comprising:
For (n+1)th moment in the second time period to any moment in the 2n moment the first predicted flow rate into
Row bi-exponential smoothing processing;
The second predicted flow rate data at the moment are determined according to the bi-exponential smoothing processing result of the predicted flow rate at the moment;
The second predicted flow rate data include the second pre- flow measurement at (n+1)th moment to the 2n moment in second time period
Amount.
6. according to the method described in claim 5, it is characterized in that, described according to the first predicted flow rate data and described
Predicted flow rate data in second time period described in two predicted flow rate data acquisitions, comprising:
Calculate the first predicted flow rate of any moment and the arithmetic mean of instantaneous value of the second predicted flow rate;
Using the arithmetic mean of instantaneous value as the predicted flow rate data at the moment.
7. a kind of volume forecasting device, which is characterized in that described device includes:
Module is obtained, for obtaining data on flows of the target network in first time period;
Preprocessing module is connected to the acquisition module, for pre-processing to the data on flows, obtains pretreated
Data on flows;
First computing module is connected to the preprocessing module, for according to the pretreated data on flows and the first mould
Type obtains the first predicted flow rate data in second time period;
Second computing module is connected to first computing module, for according to the first predicted flow rate data and the second mould
Type obtains the corresponding second predicted flow rate data of the first predicted flow rate data;
Third computing module is connected to second computing module, for according to the first predicted flow rate data and described the
Predicted flow rate data in second time period described in two predicted flow rate data acquisitions.
8. device according to claim 7, which is characterized in that the pretreated data on flows includes first time period
The n flow at interior 1st moment to n-th of moment;
First predicted flow rate of the second time period includes first predicted flow rate at (n+1)th moment to the 2n moment.
9. device according to claim 8, which is characterized in that first model is shot and long term memory network model
LSTM;
First computing module, comprising:
Training submodule, for i-th of moment in the flow according to the n continuous moment data on flows, (i-1)-th when
The output of LSTM, the state parameter of (i-1)-th moment LSTM are carved, the output of i-th of moment LSTM, i-th of moment LSTM are obtained
State parameter and the predicted flow rate at i+1 moment;Wherein, 1≤i≤n;
As the i=1, the output of (i-1)-th moment LSTM, (i-1)-th moment LSTM state parameter be respectively random value
Or 0;
As the i=n, the predicted flow rate at i+1 moment is the first pre- flow measurement at (n+1)th moment in second time period
Amount.
10. device according to claim 9, which is characterized in that first computing module, further includes:
First operation submodule is connected to the trained submodule, for the pre- flow measurement according to m-th of moment in second time period
Amount, the output of the m-1 moment LSTM, the m-1 moment LSTM state parameter, obtain m-th of moment LSTM output, the
The state parameter and the predicted flow rate at the m+1 moment of m moment LSTM;Wherein, n+1≤m≤2n.
11. device according to claim 8, which is characterized in that second model is SES model;
Second computing module, comprising:
Smooth submodule, for for (n+1)th moment in the second time period to any moment in the 2n moment
The first predicted flow rate carry out bi-exponential smoothing processing;
It determines submodule, is connected to the smooth submodule, the bi-exponential for the predicted flow rate according to the moment is smoothly located
Reason result determines the second predicted flow rate data at the moment;
Second operation submodule is connected to the determining submodule, includes the second time for the second predicted flow rate data
Second predicted flow rate at (n+1)th moment to the 2n moment in section.
12. device according to claim 11, which is characterized in that the third computing module, comprising:
Average submodule, for calculating the first predicted flow rate of any moment and the arithmetic mean of instantaneous value of the second predicted flow rate;
Third operation submodule is connected to the average submodule, for using the arithmetic mean of instantaneous value as the prediction at the moment
Data on flows.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811195940.5A CN109120463B (en) | 2018-10-15 | 2018-10-15 | Flow prediction method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811195940.5A CN109120463B (en) | 2018-10-15 | 2018-10-15 | Flow prediction method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109120463A true CN109120463A (en) | 2019-01-01 |
CN109120463B CN109120463B (en) | 2022-01-07 |
Family
ID=64854255
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811195940.5A Active CN109120463B (en) | 2018-10-15 | 2018-10-15 | Flow prediction method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109120463B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
CN110445645A (en) * | 2019-07-26 | 2019-11-12 | 新华三大数据技术有限公司 | Link flow prediction technique and device |
CN110839184A (en) * | 2019-10-15 | 2020-02-25 | 北京邮电大学 | Method and device for adjusting bandwidth of mobile fronthaul optical network based on flow prediction |
CN111327453A (en) * | 2020-01-19 | 2020-06-23 | 国网福建省电力有限公司经济技术研究院 | Communication bandwidth estimation method considering gridding dynamic and static components |
CN112087350A (en) * | 2020-09-17 | 2020-12-15 | 中国工商银行股份有限公司 | Method, device, system and medium for monitoring network access line flow |
CN113079033A (en) * | 2021-03-08 | 2021-07-06 | 南京苏宁软件技术有限公司 | Flow control method and device, electronic equipment and computer readable medium |
CN114726745A (en) * | 2021-01-05 | 2022-07-08 | 中国移动通信有限公司研究院 | Network flow prediction method and device and computer readable storage medium |
CN115633317A (en) * | 2022-12-21 | 2023-01-20 | 北京金楼世纪科技有限公司 | Message channel configuration method and system |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU5943199A (en) * | 1999-03-05 | 2000-09-07 | Inmarsat Ltd | Communication methods and apparatus |
CN101453747A (en) * | 2008-10-31 | 2009-06-10 | 中国移动通信集团北京有限公司 | Telephone traffic prediction method and apparatus |
CN102781081A (en) * | 2012-07-13 | 2012-11-14 | 浙江工业大学 | Energy-saving transmission for wireless sensor network based on secondary exponential smoothing forecasting |
CN103164742A (en) * | 2013-04-02 | 2013-06-19 | 南京邮电大学 | Server performance prediction method based on particle swarm optimization nerve network |
CN104219691A (en) * | 2013-05-29 | 2014-12-17 | 华为技术有限公司 | Traffic prediction method and system for cellular network |
CN104301895A (en) * | 2014-09-28 | 2015-01-21 | 北京邮电大学 | Double-layer trigger intrusion detection method based on flow prediction |
CN105447594A (en) * | 2015-11-17 | 2016-03-30 | 福州大学 | Electric power system grey load prediction method based on exponential smoothing |
GB2541511A (en) * | 2015-06-23 | 2017-02-22 | Ford Global Tech Llc | Rapid traffic parameter estimation |
CN107086935A (en) * | 2017-06-16 | 2017-08-22 | 重庆邮电大学 | Flow of the people distribution forecasting method based on WIFI AP |
CN107124320A (en) * | 2017-06-30 | 2017-09-01 | 北京金山安全软件有限公司 | Traffic data monitoring method and device and server |
CN107171848A (en) * | 2017-05-27 | 2017-09-15 | 华为技术有限公司 | A kind of method for predicting and device |
CN108062561A (en) * | 2017-12-05 | 2018-05-22 | 华南理工大学 | A kind of short time data stream Forecasting Methodology based on long memory network model in short-term |
CN108197739A (en) * | 2017-12-29 | 2018-06-22 | 中车工业研究院有限公司 | A kind of urban track traffic ridership Forecasting Methodology |
-
2018
- 2018-10-15 CN CN201811195940.5A patent/CN109120463B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU5943199A (en) * | 1999-03-05 | 2000-09-07 | Inmarsat Ltd | Communication methods and apparatus |
CN101453747A (en) * | 2008-10-31 | 2009-06-10 | 中国移动通信集团北京有限公司 | Telephone traffic prediction method and apparatus |
CN102781081A (en) * | 2012-07-13 | 2012-11-14 | 浙江工业大学 | Energy-saving transmission for wireless sensor network based on secondary exponential smoothing forecasting |
CN103164742A (en) * | 2013-04-02 | 2013-06-19 | 南京邮电大学 | Server performance prediction method based on particle swarm optimization nerve network |
CN104219691A (en) * | 2013-05-29 | 2014-12-17 | 华为技术有限公司 | Traffic prediction method and system for cellular network |
CN104301895A (en) * | 2014-09-28 | 2015-01-21 | 北京邮电大学 | Double-layer trigger intrusion detection method based on flow prediction |
GB2541511A (en) * | 2015-06-23 | 2017-02-22 | Ford Global Tech Llc | Rapid traffic parameter estimation |
CN105447594A (en) * | 2015-11-17 | 2016-03-30 | 福州大学 | Electric power system grey load prediction method based on exponential smoothing |
CN107171848A (en) * | 2017-05-27 | 2017-09-15 | 华为技术有限公司 | A kind of method for predicting and device |
CN107086935A (en) * | 2017-06-16 | 2017-08-22 | 重庆邮电大学 | Flow of the people distribution forecasting method based on WIFI AP |
CN107124320A (en) * | 2017-06-30 | 2017-09-01 | 北京金山安全软件有限公司 | Traffic data monitoring method and device and server |
CN108062561A (en) * | 2017-12-05 | 2018-05-22 | 华南理工大学 | A kind of short time data stream Forecasting Methodology based on long memory network model in short-term |
CN108197739A (en) * | 2017-12-29 | 2018-06-22 | 中车工业研究院有限公司 | A kind of urban track traffic ridership Forecasting Methodology |
Non-Patent Citations (1)
Title |
---|
郝占军: "网络流量分析与预测模型研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110381524B (en) * | 2019-07-15 | 2022-12-20 | 安徽理工大学 | Bi-LSTM-based large scene mobile flow online prediction method, system and storage medium |
CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
CN110445645A (en) * | 2019-07-26 | 2019-11-12 | 新华三大数据技术有限公司 | Link flow prediction technique and device |
CN110839184A (en) * | 2019-10-15 | 2020-02-25 | 北京邮电大学 | Method and device for adjusting bandwidth of mobile fronthaul optical network based on flow prediction |
CN110839184B (en) * | 2019-10-15 | 2021-06-15 | 北京邮电大学 | Method and device for adjusting bandwidth of mobile fronthaul optical network based on flow prediction |
CN111327453A (en) * | 2020-01-19 | 2020-06-23 | 国网福建省电力有限公司经济技术研究院 | Communication bandwidth estimation method considering gridding dynamic and static components |
CN111327453B (en) * | 2020-01-19 | 2023-04-07 | 国网福建省电力有限公司经济技术研究院 | Communication bandwidth estimation method considering gridding dynamic and static components |
CN112087350A (en) * | 2020-09-17 | 2020-12-15 | 中国工商银行股份有限公司 | Method, device, system and medium for monitoring network access line flow |
CN112087350B (en) * | 2020-09-17 | 2022-03-18 | 中国工商银行股份有限公司 | Method, device, system and medium for monitoring network access line flow |
CN114726745A (en) * | 2021-01-05 | 2022-07-08 | 中国移动通信有限公司研究院 | Network flow prediction method and device and computer readable storage medium |
CN113079033A (en) * | 2021-03-08 | 2021-07-06 | 南京苏宁软件技术有限公司 | Flow control method and device, electronic equipment and computer readable medium |
CN113079033B (en) * | 2021-03-08 | 2022-09-27 | 南京苏宁软件技术有限公司 | Flow control method and device, electronic equipment and computer readable medium |
CN115633317A (en) * | 2022-12-21 | 2023-01-20 | 北京金楼世纪科技有限公司 | Message channel configuration method and system |
Also Published As
Publication number | Publication date |
---|---|
CN109120463B (en) | 2022-01-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109120463A (en) | Method for predicting and device | |
CN107608862B (en) | Monitoring alarm method, monitoring alarm device and computer readable storage medium | |
CN109873712B (en) | Network traffic prediction method and device | |
JP5962190B2 (en) | Method and apparatus for predicting short-term power load | |
CN107480028B (en) | Method and device for acquiring usable residual time of disk | |
US20150278706A1 (en) | Method, Predictive Analytics System, and Computer Program Product for Performing Online and Offline Learning | |
CN106549772A (en) | Resource prediction method, system and capacity management device | |
CN103778474A (en) | Resource load capacity prediction method, analysis prediction system and service operation monitoring system | |
CN108335487B (en) | Road traffic state prediction system based on traffic state time sequence | |
US20150024367A1 (en) | Cost-aware non-stationary online learning | |
Yu et al. | Integrating clustering and learning for improved workload prediction in the cloud | |
WO2019025944A1 (en) | Integrated method for automating enforcement of service level agreements for cloud services | |
CN109471698B (en) | System and method for detecting abnormal behavior of virtual machine in cloud environment | |
WO2017071369A1 (en) | Method and device for predicting user unsubscription | |
US20150012255A1 (en) | Clustering based continuous performance prediction and monitoring for semiconductor manufacturing processes using nonparametric bayesian models | |
CN112183868B (en) | Traffic flow prediction model construction method and electronic equipment | |
CN111783356A (en) | Petroleum yield prediction method and device based on artificial intelligence | |
US8793106B2 (en) | Continuous prediction of expected chip performance throughout the production lifecycle | |
US20190104028A1 (en) | Performance monitoring at edge of communication networks using hybrid multi-granular computation with learning feedback | |
Cenggoro et al. | Dynamic bandwidth management based on traffic prediction using Deep Long Short Term Memory | |
Ponmalar et al. | Machine Learning Based Network Traffic Predictive Analysis | |
CN107590747A (en) | Power grid asset turnover rate computational methods based on the analysis of comprehensive energy big data | |
CN109542585B (en) | Virtual machine workload prediction method supporting irregular time intervals | |
CN116306030A (en) | New energy prediction dynamic scene generation method considering prediction error and fluctuation distribution | |
CN109978172B (en) | Resource pool utilization rate prediction method and device based on extreme learning machine |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |