CN110267292A - Cellular network method for predicting based on Three dimensional convolution neural network - Google Patents
Cellular network method for predicting based on Three dimensional convolution neural network Download PDFInfo
- Publication number
- CN110267292A CN110267292A CN201910408711.5A CN201910408711A CN110267292A CN 110267292 A CN110267292 A CN 110267292A CN 201910408711 A CN201910408711 A CN 201910408711A CN 110267292 A CN110267292 A CN 110267292A
- Authority
- CN
- China
- Prior art keywords
- data
- network
- short
- training
- long
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- 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
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of cellular network method for predicting based on Three dimensional convolution neural network, comprising the following steps: network flow data is modeled as three-dimensional tensor input form and obtains three-dimensional network data on flows model；According to three-dimensional network data on flows, training set data and test set data are obtained；Construct basic Three dimensional convolution neural network；To in short-term rely on data carry out Three dimensional convolution neural network training obtain short-time characteristic, to it is long when rely on data carry out Three dimensional convolution neural network training obtain long Shi Tezheng；To short-time characteristic and it is long when feature carry out Fusion training, obtain eigenmatrix, as the output of basic Three dimensional convolution neural network, form training pattern；Network flow data to be predicted is predicted using training pattern, obtains predicting network flow result.Prediction technique provided by the invention simultaneously consider network flow data short-term correlation and it is long when tendency, capture network flow data timing on feature correlation.
Description
[technical field]
The present invention relates to computer time series forecasting application field more particularly to a kind of bees based on Three dimensional convolution neural network
Nest network flow prediction method.
[background technique]
In recent years, universal with mobile device and mobile application, radio network technique is in the world to people
Daily life play key effect, more and more people access cellular network, cellular network communication amount using mobile device
And the demand of network flow increases rapidly.Newest industry prediction shows, by 2021, the cellular network stream of international mobile equipment
It will be more than everyone 48.3EB that amount is estimated, it is 7 times of current usage amount, and smart phone flow will be more than PC flow in the same year.
For cellular service providers and infrastructure provider, in order to cope with growing demand, provide a user
Stable cellular network service and secure service quality (Qos), Accurate Prediction mobile communication demand is vital.Example
Such as, by the Accurate Prediction to cellular network flow demand, timely flow scheduling may be implemented, by portion requirements from busy
Launching tower, which is distributed to idle launching tower, influences user experience to avoid network congestion.Obviously, carrying out volume forecasting can be with
It optimizes allocation of resources, improves energy efficiency, to realize that intelligent cellular network lays a good foundation.
In the related technology, cellular network volume forecasting is modeled usually as general time series analysis problem.It is right
Problems study in the Linear Statistical Model to generate, and what is be most widely used is the comprehensive method of moving average of autoregression
(ARIMA) and support vector regression method (SVR).However ARIMA method is only intended merely to the average value of concern historical series data,
To which the quick variational procedure of bottom flow load can not be captured, and the non-linear relation in real system can not be built
Mould；Although and SVR method can be handled non-linear relation, need to carry out key parameter tuning, could obtain standard
True prediction result.Simultaneously, it is contemplated that cellular network subscriber mobility, to the factors such as expression patterns and user demand diversity
It influences, such method largely all has ignored the potentially relevant property between the flow sequence in cellular network.For example, cellular network
In spatial dependence, the movement of user obviously will drive flow demand and shifts, and the traffic between different base station is caused to be deposited
Also having in the basic transport need of significant spatial dependence, and each region is influenced by surrounding enviroment, bustling area
The transport need in domain is obviously greater than remote areas, and these dependences are all that conventional method can not capture.In recent years, depth
Learning model also provides new thinking in the latest developments of every field for the problems such as volume forecasting.
Therefore, it is necessary to provide a kind of cellular network method for predicting based on Three dimensional convolution neural network to solve
State problem.
[summary of the invention]
The technical problem to be solved in the present invention is to provide a kind of short-term correlations and length for considering network flow data simultaneously
When tendency, capture network flow data timing on feature correlation the cellular network stream based on Three dimensional convolution neural network
Measure prediction technique.
In order to solve the above technical problems, the present invention provides a kind of cellular network flows based on Three dimensional convolution neural network
Prediction technique, comprising the following steps:
S1: being modeled as three-dimensional tensor input form for network flow data, obtains three-dimensional network data on flows model, described
Data are relied on when three-dimensional network data on flows model includes long and rely on data in short-term；
S2: according to the three-dimensional network data on flows, training set data and test set data are obtained；
S3: basic Three dimensional convolution neural network is constructed；
S4: short-time characteristic is obtained to the training for relying on data progress Three dimensional convolution neural network in short-term, to the length
When rely on data carry out Three dimensional convolution neural network training obtain long Shi Tezheng；
S5: to the short-time characteristic and it is long when feature carry out Fusion training, obtain eigenmatrix, and by the feature square
Output of the battle array as the basic Three dimensional convolution neural network, forms training pattern；
S6: network flow data to be predicted is predicted using the training pattern, obtains predicting network flow knot
Fruit.
Preferably, the step S1 includes the following steps:
City: being divided into the grid chart of a H × W by S11, with 15 minutes for an interval, records in grid chart and owns
The network flow data in region, and it is merged into the network flow data of 1 hour, wherein the net region without network flow data
It is filled with numerical value 0；
S12: tensor X is enabledt∈RH×WRepresent the overall network flow value transmitted in all grids in entire city in t time slot；
Enable tensorIt represents in coordinate as generated network flow in the net region of (i, j), one of time slot
Represent the interval of a hour；
S13: modeling the temporal correlation of network flow from dependence when relying in short-term and is long, wherein in short-term according to
Rely the temporal correlation for referring to that network flow is embodied in a time slot interval；It is relied on when long and refers to 24 time slots
The temporal correlation that network flow is embodied in being spaced.
Preferably, the step S2 includes the following steps:
S21: definition test set data length is n, and m sample conduct is extracted from the three-dimensional network data on flows model
Test sample collection, using remaining n-m sample as training sample set；
S22: the data concentrated respectively to the training sample set and the test sample carry out min-max standardization,
The data vector value for finally entering the training sample set and the test sample collection is mapped in [0,1] range.
Preferably, in the step S22 data vector conversion process are as follows:
Wherein min is the minimum value of the data of the training sample set or test sample concentration,
Max is the maximum value of the data of the training sample set or test sample concentration.
Preferably, the step S4 includes the following steps:
S41: two structures identical basic Three dimensional convolution neural network c_3DCNN and p_ are constructed based on step S3
3DCNN, be respectively trained in short-term rely on data flow and it is long when rely on data flow.
S42: the parameter of initialization c_3DCNN and p_3DCNN network；
S43: the training the number of iterations of setting c_3DCNN and p_3DCNN is epochs, to the root mean square on test set
Pt numerical value is arranged as monitoring data in error val_RMSE；
S44: will rely in short-term tensor and it is long when rely on tensor as the input number of c_3DCNN and p_3DCNN
According to, respectively extract short-time characteristic VcAnd feature V when longp。
Preferably, epochs=50, pt=10.
Preferably, Fusion training in the step S5 specifically:
Wherein, VfusionRepresent the feature obtained after fusion, WcAnd WpGeneration respectively
The influence generated, V are relied on when table needs the weight matrix learnt to rely in short-term to be fitted and is longcAnd VpIt indicates to extract in step S4
The short-time characteristic and long Shi Tezheng arrived,Indicate the point multiplication operation between vector.
Compared with the relevant technologies, the cellular network method for predicting provided by the invention based on Three dimensional convolution neural network
Beneficial effect be:
1) it is different from the short-term correlation that traditional prediction method is only capable of capture data on flows, the present invention considers not only network
The short-term correlation of data on flows, tendency when equally also contemplating long, when can capture network flow data more perfectly
Feature correlation in sequence；
2) advantage of deep neural network model is made full use of, the spatial coherence between city grid is sufficiently excavated.It will
Network flow data Series Modeling is three-dimensional tensor model, regards the data on flows collected out of city as Zhang Quanjing's picture,
To which the spatial coherence between city grid can be excavated using convolutional neural networks, time series forecasting problem is converted into image
Identification problem gives full play to the advantage of convolutional neural networks；
3) traditional time series forecasting network LSTM (long short-term memory nerve net is substituted using basic Three dimensional convolution neural network
Network), greatly reduce network parameter, effectively improves net training time and precision of prediction.
[Detailed description of the invention]
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing, in which:
Fig. 1 is the step process of the cellular network method for predicting provided by the invention based on Three dimensional convolution neural network
Figure；
Fig. 2 is the step flow chart of step S1 shown in FIG. 1；
Fig. 3 is the step flow chart of step S2 shown in FIG. 1；
Fig. 4 is the step flow chart of step S4 shown in FIG. 1.
[specific embodiment]
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that the described embodiments are merely a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1 to Fig. 4 is please referred to, it is pre- that the present invention provides a kind of cellular network flow based on Three dimensional convolution neural network
Survey method, includes the following steps:
S1: being modeled as three-dimensional tensor input form for network flow data, obtains three-dimensional network data on flows model, described
Data are relied on when three-dimensional network data on flows model includes long and rely on data in short-term.
In the present embodiment, a city M is selected, the network flow data in the city M is modeled as three-dimensional
Measure input form.
Specifically, the step S1 includes the following steps:
City: being divided into the grid chart of a H × W by S11, with 15 minutes for an interval, records in grid chart and owns
The network flow data in region, and it is merged into the network flow data of 1 hour, wherein the net region without network flow data
It is filled with numerical value 0.
Specifically, H=W=100 in the present embodiment, in other embodiments, the value of the H and W can be with roots
It is adjusted according to actual needs, the present invention is without limitation.
S12: tensor X is enabledt∈RH×WRepresent the overall network flow value transmitted in all grids in entire city in t time slot；
Enable tensorIt represents in coordinate as generated network flow in the net region of (i, j), one of time slot
Represent the interval of a hour.
It therefore can be by each tensor XtIt is considered as a channel and is 1 image, to realizes that the space of network flow relies on
Property.
S13: modeling the temporal correlation of network flow from dependence when relying in short-term and is long, wherein in short-term according to
Rely the temporal correlation for referring to that network flow is embodied in a time slot interval；It is relied on when long and refers to 24 time slots
The temporal correlation that network flow is embodied in being spaced.
Specifically, defining c is that sequence-dependent length, building rely on three-dimensional tensor flow table in short-term and be shown as [X in short-termt-c,
Xt-(c-1)..., Xt-1].The network flow of each time slot is got up along first axis connection, similarly, when definition p is long
Sequence-dependent length relies on three-dimensional tensor flow table and is shown as [X when constructing longt-p*24, Xt-(p-1)*24..., Xt-24].By each time
The network flow of slot is got up along first axis connection, can be obtained as urban cellular network flow data model, difference table
It is shown as relying on tensor X in short-termc∈Rc×H×WAnd tensor X is relied on when longp∈Rp×H×W。
S2: according to the three-dimensional network data on flows, training set data and test set data are obtained.
Specifically, the step S2 includes the following steps:
S21: definition test set data length is n, and m sample conduct is extracted from the three-dimensional network data on flows model
Test sample collection, using remaining n-m sample as training sample set.
S22: the data concentrated respectively to the training sample set and the test sample carry out min-max standardization,
The data vector value for finally entering the training sample set and the test sample collection is mapped in [0,1] range.
Specifically, the conversion process of the data vector are as follows:
Wherein min is the minimum value of the data of the training sample set or test sample concentration,
Max is the maximum value of the data of the training sample set or test sample concentration.
Step S3: basic Three dimensional convolution neural network is constructed.
A Three dimensional convolution neural network is constructed, the three convolution kernel sizes that are linked in sequence are the convolutional layer of (3,3,3), are used
ReLU function f (x)=max { 0, X } is used as activation primitive, using the obtained output characteristic pattern after convolution as maximum pond layer
Input, and inhibit a part of network neural member at random using DropOut, prevent over-fitting.
S4: short-time characteristic is obtained to the training for relying on data progress Three dimensional convolution neural network in short-term, to the length
When rely on data carry out Three dimensional convolution neural network training obtain long Shi Tezheng.
Specifically, the step S4 includes the following steps:
S41: two structures identical basic Three dimensional convolution neural network c_3DCNN and p_ are constructed based on step S3
3DCNN, be respectively trained in short-term rely on data flow and it is long when rely on data flow.
S42: the parameter of initialization c_3DCNN and p_3DCNN network.
Weight matrix Wi and bi in two Three dimensional convolution neural network network input layers and hidden layer is initialized as
Range is the random number of [0,1], and the parameter in training process is facilitated to adjust.
S43: the training the number of iterations of setting c_3DCNN and p_3DCNN is epochs, to the root mean square on test set
Pt numerical value is arranged as monitoring data in error val_RMSE.
If val_RMSE is not promoted in pt the number of iterations, even if not reaching the number of iterations epochs of setting,
Also stop network training immediately, prevent the generation of over-fitting.
Specifically, in the present embodiment, epochs=50, pt=10, even in 10 the number of iterations, val_RMSE
It is not promoted, even if not reaching 10 iteration, also stops network training immediately, prevent the generation of over-fitting.
S44: will rely in short-term tensor and it is long when rely on tensor as the input number of c_3DCNN and p_3DCNN
According to, respectively extract short-time characteristic VcAnd feature V when longp。
S5: to the short-time characteristic and it is long when feature carry out Fusion training, obtain eigenmatrix, and by the feature square
Output of the battle array as the basic Three dimensional convolution neural network, forms training pattern.
Due to grid each in city by rely in short-term and it is long when rely on and influenced, and influence degree is different, therefore will step
The short-time characteristic V obtained in rapid S4cAnd feature V when longpFusion retraining is carried out, as follows:
Wherein, VfusionRepresent the feature obtained after fusion, WcAnd WpIt respectively represents and the weight matrix learnt is needed to intend
Close the influence for relying on and generating when relying in short-term and is long, VcAnd VpIndicate the short-time characteristic extracted in step S4 and long Shi Tezheng,Indicate the point multiplication operation between vector.
The eigenmatrix V obtained after fusionfusionIt is flattened into a feature vector Vout, output as whole network.
The cross entropy of whole network is minimized by back-propagation algorithm again, is that Adam optimizer is excellent to network progress with majorized function
Change, the learning rate that optimizer is arranged is lr.
Step 6: network flow data to be predicted being predicted using the training pattern, obtains predicting network flow
As a result.
Sample in test set is input in the network of trained completion, normalized predicting network flow knot is obtained
Fruit Ypre, and renormalization operation is carried out to it, to obtain predicting network flow test result.
Compared with the relevant technologies, the cellular network method for predicting provided by the invention based on Three dimensional convolution neural network
Beneficial effect be:
1) it is different from the short-term correlation that traditional prediction method is only capable of capture data on flows, the present invention considers not only network
The short-term correlation of data on flows, tendency when equally also contemplating long, when can capture network flow data more perfectly
Feature correlation in sequence；
2) advantage of deep neural network model is made full use of, the spatial coherence between city grid is sufficiently excavated.It will
Network flow data Series Modeling is three-dimensional tensor model, regards the data on flows collected out of city as Zhang Quanjing's picture,
To which the spatial coherence between city grid can be excavated using convolutional neural networks, time series forecasting problem is converted into image
Identification problem gives full play to the advantage of convolutional neural networks；
3) traditional time series forecasting network LSTM (long short-term memory nerve net is substituted using basic Three dimensional convolution neural network
Network), greatly reduce network parameter, effectively improves net training time and precision of prediction.
Above-described is only embodiments of the present invention, it should be noted here that for those of ordinary skill in the art
For, without departing from the concept of the premise of the invention, improvement can also be made, but these belong to protection model of the invention
It encloses.
Claims (7)
1. a kind of cellular network method for predicting based on Three dimensional convolution neural network, comprising the following steps:
S1: being modeled as three-dimensional tensor input form for network flow data, obtains three-dimensional network data on flows model, the three-dimensional
Data are relied on when network flow data model includes long and rely on data in short-term；
S2: according to the three-dimensional network data on flows, training set data and test set data are obtained；
S3: basic Three dimensional convolution neural network is constructed；
S4: short-time characteristic is obtained to the training for relying on data progress Three dimensional convolution neural network in short-term, to the long Shi Yi
The training for relying data to carry out Three dimensional convolution neural network obtains long Shi Tezheng；
S5: to the short-time characteristic and it is long when feature carry out Fusion training, obtain eigenmatrix, and the eigenmatrix is made
For the output of the basic Three dimensional convolution neural network, training pattern is formed；
S6: network flow data to be predicted is predicted using the training pattern, obtains predicting network flow result.
2. the method according to claim 1, wherein the step S1 includes the following steps:
City: being divided into the grid chart of a H × W by S11, with 15 minutes for an interval, records all areas in grid chart
Network flow data, and the network flow data of 1 hour is merged into, wherein the net region without network flow data is filled
For numerical value 0；
S12: tensor X is enabledt∈RH×WRepresent the overall network flow value transmitted in all grids in entire city in t time slot；It enables and opening
AmountIt represents in coordinate as generated network flow in the net region of (i, j), one of time slot represents
The interval of one hour；
S13: the temporal correlation of network flow is modeled from dependence when relying in short-term and is long, wherein relying in short-term is
Refer to the temporal correlation that network flow is embodied in a time slot interval；It is relied on when long and refers to 24 time slot intervals
The temporal correlation that interior network flow is embodied.
3. the method according to claim 1, wherein the step S2 includes the following steps:
S21: definition test set data length is n, and m sample is extracted from the three-dimensional network data on flows model as test
Sample set, using remaining n-m sample as training sample set；
S22: the data concentrated respectively to the training sample set and the test sample carry out min-max standardization, make institute
It states training sample set and data vector value that the test sample collection finally enters is mapped in [0,1] range.
4. according to the method described in claim 3, it is characterized in that, in the step S22 data vector conversion process are as follows:
Wherein min is the minimum value of the data of the training sample set or test sample concentration, max
For the maximum value for the data that the training sample set or the test sample are concentrated.
5. the method according to claim 1, wherein the step S4 includes the following steps:
S41: constructing two structures identical basic Three dimensional convolution neural network c_3DCNN and p_3DCNN based on step S3, point
Xun Lian not rely in short-term data flow and it is long when rely on data flow.
S42: the parameter of initialization c_3DCNN and p_3DCNN network；
S43: the training the number of iterations of setting c_3DCNN and p_3DCNN is epochs, to the root-mean-square error on test set
Pt numerical value is arranged as monitoring data in val_RMSE；
S44: will rely in short-term tensor and it is long when rely on tensor as the input data of c_3DCNN and p_3DCNN, point
Indescribably take short-time characteristic VcAnd feature V when longp。
6. according to the method described in claim 5, it is characterized in that, epochs=50, pt=10.
7. the method according to claim 1, wherein Fusion training in the step S5 specifically:
Wherein, VfusionRepresent the feature obtained after fusion, WcAnd WpIt is short to be fitted to respectively represent the weight matrix for needing to learn
When rely on it is long when rely on generate influence, VcAnd VpIndicate the short-time characteristic extracted in step S4 and long Shi Tezheng,Table
Show the point multiplication operation between vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910408711.5A CN110267292B (en) | 2019-05-16 | 2019-05-16 | Cellular network flow prediction method based on three-dimensional convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910408711.5A CN110267292B (en) | 2019-05-16 | 2019-05-16 | Cellular network flow prediction method based on three-dimensional convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110267292A true CN110267292A (en) | 2019-09-20 |
| CN110267292B CN110267292B (en) | 2022-07-08 |
Family
ID=67914727
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910408711.5A Active CN110267292B (en) | 2019-05-16 | 2019-05-16 | Cellular network flow prediction method based on three-dimensional convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110267292B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111159149A (en) * | 2019-12-13 | 2020-05-15 | 国网浙江省电力有限公司紧水滩水力发电厂 | River flow prediction method based on three-dimensional convolutional neural network |
| 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 |
| CN111404942A (en) * | 2020-03-18 | 2020-07-10 | 广东技术师范大学 | Vertical malicious crawler flow identification method based on deep learning |
| CN111866024A (en) * | 2020-08-05 | 2020-10-30 | 国家计算机网络与信息安全管理中心 | Network encryption traffic identification method and device |
| CN111949704A (en) * | 2020-07-17 | 2020-11-17 | 网络通信与安全紫金山实验室 | Interpretable multidimensional time sequence data analysis method |
| CN112291807A (en) * | 2020-10-15 | 2021-01-29 | 山东科技大学 | Wireless cellular network traffic prediction method based on deep migration learning and cross-domain data fusion |
| 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 |
| CN116566841A (en) * | 2023-05-09 | 2023-08-08 | 北京有元科技有限公司 | Flow trend prediction method and device based on network flow query |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| US20180033144A1 (en) * | 2016-09-21 | 2018-02-01 | Realize, Inc. | Anomaly detection in volumetric images |
| CN109508375A (en) * | 2018-11-19 | 2019-03-22 | 重庆邮电大学 | A kind of social affective classification method based on multi-modal fusion |
| CN109697852A (en) * | 2019-01-23 | 2019-04-30 | 吉林大学 | Urban road congestion degree prediction technique based on timing traffic events |
-
2019
- 2019-05-16 CN CN201910408711.5A patent/CN110267292B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180033144A1 (en) * | 2016-09-21 | 2018-02-01 | Realize, Inc. | Anomaly detection in volumetric images |
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| CN109508375A (en) * | 2018-11-19 | 2019-03-22 | 重庆邮电大学 | A kind of social affective classification method based on multi-modal fusion |
| CN109697852A (en) * | 2019-01-23 | 2019-04-30 | 吉林大学 | Urban road congestion degree prediction technique based on timing traffic events |
Non-Patent Citations (2)
| Title |
|---|
| CEN CHEN; KENLI LI; SIN G. TEO; GUIZI CHEN; XIAOFENG ZOU; XULEI: "Exploiting Spatio-Temporal Correlations with Multiple 3D Convolutional Neural Networks for Citywide Vehicle Flow Prediction", 《 2018 IEEE INTERNATIONAL CONFERENCE ON DATA MINING (ICDM)》 * |
| XI OUYANG; SHUANGJIE XU; CHAOYUN ZHANG; PAN ZHOU; YANG YANG; GUA: "A 3D-CNN and LSTM Based Multi-Task Learning Architecture for Action Recognition", 《IEEE ACCESS》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111159149A (en) * | 2019-12-13 | 2020-05-15 | 国网浙江省电力有限公司紧水滩水力发电厂 | River flow prediction method based on three-dimensional convolutional neural network |
| CN111159149B (en) * | 2019-12-13 | 2023-06-30 | 国网浙江省电力有限公司紧水滩水力发电厂 | River flow prediction method based on three-dimensional convolutional neural network |
| 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 |
| 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 |
| CN111404942A (en) * | 2020-03-18 | 2020-07-10 | 广东技术师范大学 | Vertical malicious crawler flow identification method based on deep learning |
| CN111949704A (en) * | 2020-07-17 | 2020-11-17 | 网络通信与安全紫金山实验室 | Interpretable multidimensional time sequence data analysis method |
| CN111866024A (en) * | 2020-08-05 | 2020-10-30 | 国家计算机网络与信息安全管理中心 | Network encryption traffic identification method and device |
| CN111866024B (en) * | 2020-08-05 | 2022-10-14 | 国家计算机网络与信息安全管理中心 | Network encryption traffic identification method and device |
| CN112291807A (en) * | 2020-10-15 | 2021-01-29 | 山东科技大学 | Wireless cellular network traffic prediction method based on deep migration learning and cross-domain data fusion |
| 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 |
| CN114726745B (en) * | 2021-01-05 | 2024-05-17 | 中国移动通信有限公司研究院 | Network traffic prediction method, device and computer readable storage medium |
| CN116566841A (en) * | 2023-05-09 | 2023-08-08 | 北京有元科技有限公司 | Flow trend prediction method and device based on network flow query |
| CN116566841B (en) * | 2023-05-09 | 2023-12-01 | 北京有元科技有限公司 | Flow trend prediction method based on network flow query |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110267292B (en) | 2022-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110267292A (en) | Cellular network method for predicting based on Three dimensional convolution neural network | |
| Fang et al. | Mobile demand forecasting via deep graph-sequence spatiotemporal modeling in cellular networks | |
| US20210133536A1 (en) | Load prediction method and apparatus based on neural network | |
| CN109376969A (en) | City fining population distribution dynamic prediction method and device based on deep learning | |
| Larios et al. | Energy efficient wireless sensor network communications based on computational intelligent data fusion for environmental monitoring | |
| CN110807230B (en) | Method for autonomously learning and optimizing topological structure robustness of Internet of things | |
| Dou et al. | Hybrid model for renewable energy and loads prediction based on data mining and variational mode decomposition | |
| Zhang et al. | LNTP: An end-to-end online prediction model for network traffic | |
| CN105340311B (en) | The network equipment, network service prediction meanss and method | |
| CN113469325B (en) | Hierarchical federation learning method for edge aggregation interval self-adaptive control, computer equipment and storage medium | |
| CN112512069B (en) | Network intelligent optimization method and device based on channel beam pattern | |
| CN112911626B (en) | Wireless network flow prediction method based on multi-graph convolution | |
| Kumari et al. | An energy efficient smart metering system using edge computing in LoRa network | |
| He et al. | Graph attention spatial-temporal network for deep learning based mobile traffic prediction | |
| CN110381515A (en) | Based on the method for closing merotype realization subzone network floating resources index prediction | |
| CN109816144A (en) | The short-term load forecasting method of distributed memory parallel computation optimization deepness belief network | |
| CN115775085A (en) | Smart city management method and system based on digital twin | |
| Fang et al. | Idle time window prediction in cellular networks with deep spatiotemporal modeling | |
| Muccini et al. | Leveraging machine learning techniques for architecting self-adaptive iot systems | |
| Zou et al. | Aiming in harsh environments: A new framework for flexible and adaptive resource management | |
| Nabi et al. | Deep learning based fusion model for multivariate LTE traffic forecasting and optimized radio parameter estimation | |
| Chiumento et al. | Energy efficient WSN: A cross-layer graph signal processing solution to information redundancy | |
| CN115860153B (en) | Wireless flow prediction method and system based on personalized packet federal learning | |
| CN116843069A (en) | Commuting flow estimation method and system based on crowd activity intensity characteristics | |
| Tripathi et al. | Data-driven optimizations in IoT: A new frontier of challenges and opportunities |
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 |