CN112564945A - IP network flow estimation method based on time sequence prior and sparse representation - Google Patents

IP network flow estimation method based on time sequence prior and sparse representation Download PDF

Info

Publication number
CN112564945A
CN112564945A CN202011318745.4A CN202011318745A CN112564945A CN 112564945 A CN112564945 A CN 112564945A CN 202011318745 A CN202011318745 A CN 202011318745A CN 112564945 A CN112564945 A CN 112564945A
Authority
CN
China
Prior art keywords
matrix
flow
traffic
sparse representation
time sequence
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
Application number
CN202011318745.4A
Other languages
Chinese (zh)
Other versions
CN112564945B (en
Inventor
王传栋
张永
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202011318745.4A priority Critical patent/CN112564945B/en
Publication of CN112564945A publication Critical patent/CN112564945A/en
Application granted granted Critical
Publication of CN112564945B publication Critical patent/CN112564945B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/142—Network analysis or design using statistical or mathematical methods
    • 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Algebra (AREA)
  • Environmental & Geological Engineering (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses an IP network flow estimation method based on time sequence prior and sparse representation, which comprises the steps of firstly, acquiring flow values transmitted among all source-destination nodes in a network to construct an incomplete flow matrix; then modeling is carried out on the space-time correlation existing in the incomplete flow matrix by utilizing a sparse representation theory and a regularization technology to form a flow matrix estimation model; and then, converting the complex original problem into a plurality of sub-problems which are easy to solve by an alternating direction multiplier method, and finding out the local optimal solution of the original problem by iteratively optimizing the global optimal solution of the sub-problems. And finally, estimating a complete flow matrix. The invention can utilize the spatial correlation of the traffic matrix while considering the time sequence correlation of adjacent network nodes based on the time sequence and the spatial correlation in the traffic matrix, and provides theoretical support for the optimization of the traffic estimation method.

Description

IP network flow estimation method based on time sequence prior and sparse representation
Technical Field
The invention relates to a network traffic estimation method, in particular to an IP network traffic estimation method based on time sequence prior and sparse representation.
Background
Traffic Matrix (TM) is common full-network-level Traffic data, records Traffic values transmitted between all source-Destination (OD) node pairs of a measured network, and is widely applied to Traffic engineering, full-network anomaly detection and other application problems. However, since the traffic matrix needs to capture the global state information of the network traffic, the cost of directly measuring all the traffic matrix data is too high, and it is almost infeasible and impractical in practical application. Estimating the traffic matrix by indirect observation can reduce the cost and overhead of direct measurement, which has become a popular research field.
Many effective methods should be used for the traffic matrix estimation, but the estimation accuracy is not enough due to the fact that the inherent space-time correlation characteristic of the traffic matrix is not utilized
Disclosure of Invention
The purpose of the invention is as follows: aiming at the defects of the prior art, the invention provides an IP network flow estimation method based on time sequence prior and sparse representation, and the accuracy of flow matrix estimation is improved.
The technical scheme is as follows: the invention discloses an IP network flow estimation method based on time sequence prior and sparse representation, which is characterized by comprising the following steps of:
s1, acquiring traffic values transmitted among all source-destination nodes in the network to construct an incomplete traffic matrix;
s2, establishing a flow matrix estimation model aiming at space-time correlation existing in an incomplete flow matrix by utilizing a sparse representation theory and a regularization technology;
s3, converting the original flow matrix estimation problem into a plurality of sub-problems easy to solve by an alternating direction multiplier method;
and S4, iteratively optimizing the global optimal solution of the sub-problem to find the local optimal solution of the original problem, and estimating a complete flow matrix.
The incomplete traffic matrix in step S1 is constructed as follows:
assuming that the number of time intervals of one day is T and the total number of OD pairs is N, the traffic matrix can be represented as
Figure BDA0002792144430000011
Figure BDA0002792144430000012
Wherein m isijRepresents the flow value of the jth OD pair at the ith time interval, "# represents a known flow value,"? "represents missing flow values, and in the flow matrix, one column represents one sample, one OD pair for all time intervals of the day, and one row represents the flow values of all OD pairs for one time interval.
The step S2 includes the following steps:
firstly, a traffic estimation model based on a sparse representation theory is established according to the spatial correlation existing in an incomplete traffic matrix, and the expression is as follows:
Figure BDA0002792144430000021
wherein,
Figure BDA0002792144430000022
in order for the traffic matrix to be known to be incomplete,
Figure BDA0002792144430000023
for the complete traffic matrix to be solved,
Figure BDA0002792144430000024
the coefficient matrix is expressed for sparseness needing to be solved, the omega set expresses a known flow value element subscript set in the flow matrix, and lambda1Is an adjustable parameter. PΩ(. cndot.) is a projection operator, indicating that when the element index (i, j) ∈ Ω, the corresponding position sample element is obtained:
Figure BDA0002792144430000025
since the element values in the traffic matrix are not negative, X ≧ 0, and to avoid trivial solution, the diagonal elements of the constraint sparse representation matrix W are all 0, i.e., diag (W) ═ 0;
aiming at the time sequence correlation existing in the incomplete flow matrix, a flow estimation model based on time sequence prior and sparse representation is established by combining a flow estimation model based on a sparse representation theory, and the expression of the flow estimation model is as follows:
Figure BDA0002792144430000026
wherein λ is2In order to be able to adjust the parameters,
Figure BDA0002792144430000027
is a Toeplitz (0,1, -1) matrix.
In step S3, the sub-problem includes the traffic matrix X, the sparse representation coefficient matrix W, and the error variable C representing the error of the incomplete traffic matrix M and the complete traffic matrix X outside the set Ω in the iterative process,
Figure BDA0002792144430000028
step S4 includes the following specific steps:
s41, introducing an error variable C for solving conveniently, and rewriting the flow estimation model into the following form:
Figure BDA0002792144430000029
s.t.X≥0,diag(W)＝0，M＝X+C，PΩ(C)＝0
s42, putting constraint terms into an objective function, defining indicating functions g (X) and f (C), and converting the optimization problem into an equivalent penalty function form, wherein the expression of the indicating functions is as follows:
Figure BDA00027921444300000210
the penalty function is expressed as follows:
Figure BDA00027921444300000211
s43, obtaining the expression of X, W, C and beta to be solved according to the penalty function as follows:
Figure BDA0002792144430000031
where ρ is a fixed parameter, preferably 1.1 or 1.2, β is a parameter of a penalty function, β isk，βk-1Beta values, beta, in the k-th iteration and the k-1 th iteration, respectivelymaxIs a fixed parameter representing the maximum value of beta, preferably 106F is the F norm; and performing alternate optimization solution on the formula, wherein the flow matrix X when the preset maximum iteration times are reached is the estimated complete flow matrix, the maximum iteration times are constants within 50, and the specific numerical values are determined according to the experimental effect.
In the solving process of the sparse representation matrix W, each column is solved respectively, the solving process of each column is regarded as a LASSO problem, and W is setiI-th column, X, representing WiI-th column of X, X-iThe matrix obtained by removing the ith column from X is shown, and W is solved so as to satisfy the constraint diag (W) of 0iWhen X is not involved in the calculation, WiThe expression of (a) is as follows:
Figure BDA0002792144430000032
has the advantages that: compared with the prior art, the invention has the following remarkable advantages: the method effectively makes up the defect of inaccurate estimation caused by the traditional KNN estimation algorithm and the like, can better embody the advantages of the method under the high-dimensional condition, considers the time sequence correlation of adjacent network nodes and simultaneously utilizes the spatial correlation of the traffic matrix, so that the obtained traffic matrix is more accurate, and provides theoretical support for the optimization of the traffic estimation method.
Drawings
FIG. 1 is a schematic view of a model structure according to the present invention;
FIG. 2 is a flow chart of the method of the present invention.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
As shown in fig. 1, the basic idea of the IP network traffic estimation method based on time sequence prior and sparse representation of the present invention is to collect traffic values at t times in the network, construct an incomplete traffic matrix according to the known traffic values, construct a time sequence prior and sparse representation model based on time sequence correlation and spatial correlation, solve the model by an alternating direction multiplier method, and estimate an complete traffic matrix. The method comprises the following specific steps:
step 1), constructing an incomplete flow matrix M:
the method comprises the steps of collecting traffic values transmitted among all source-Destination (OD) nodes in a network to obtain an incomplete traffic matrix M. In the flow matrix M, one column represents one sample, which is one OD versus the flow value for all time intervals during the day. One row represents the flow values of all OD pairs in a time interval, and if the time interval number of one day is T and the total number of the OD pairs is N, the flow matrix can be represented as
Figure BDA0002792144430000041
Figure BDA0002792144430000042
Wherein m isijRepresents the flow value of the jth OD pair at the ith time interval, "# represents a known flow value,"? "represents the missing flow value. Due to the influence of the communication behavior of the user, the traffic values of the adjacent nodes in the traffic matrix are related, i.e. spatially related, and a large number of missing values in the traffic matrix are a sparse matrix.
Step 2), establishing a flow estimation model based on a sparse representation theory:
in a sparse traffic matrix, each sample can be represented as a linear combination of other samples, and the closer to the sample the higher the weight coefficient, the farther away the sample the lower or close to 0 the weight coefficient. Therefore, the traffic matrix estimation problem can be estimated using sparse representation theory. Representing the weight coefficient matrix by W, the traffic matrix estimation problem can be modeled as:
Figure BDA0002792144430000043
wherein,
Figure BDA0002792144430000044
in order for the traffic matrix to be known to be incomplete,
Figure BDA0002792144430000045
for the complete traffic matrix to be solved,
Figure BDA0002792144430000046
the method comprises the steps of representing a coefficient matrix for sparseness needing to be solved, representing a known flow value element subscript set in a flow matrix by an omega set, and representing a known flow value element subscript set in the flow matrix by a lambda1Is an adjustable parameter. PΩ(. cndot.) is a projection operator, indicating that when the element index (i, j) ∈ Ω, the corresponding position sample element is obtained:
Figure BDA0002792144430000047
since the element values in the traffic matrix are not negative, there is X ≧ 0, and to avoid trivial solution, the diagonal elements of the constraint sparse representation matrix W are all 0, i.e., diag (W) > 0.
Step 3), establishing a flow estimation model based on time sequence prior and sparse representation:
the flow data was found to be time-ordered. For each OD stream, its own time series characteristics can be described in terms of time correlation. To characterize the time-dependent nature of the flow matrix, the flow matrix is characterized by minimizing an objective function | RX |1The effect of temporal stabilization of elements in the traffic matrix in the time dimension can be obtained, so that the correlation in the time sequence in the traffic matrix is well described. At this time, the flow estimation model based on the time sequence prior and sparse representation is as follows:
Figure BDA0002792144430000048
wherein λ is2In order to be able to adjust the parameters,
Figure BDA0002792144430000049
is a Toeplitz (0,1, -1) matrix.
Step 4), converting a more complex model solution problem into three sub-problems which are easy to solve by an Alternating Direction Method of Multiprocessors (ADMM) according to the flow estimation model which is established in the step 3) and is based on time sequence prior and sparse representation, namely: a flow matrix X, a sparse representation matrix W and an error variable C in an iteration process. And (4) finding a local optimal solution of the original problem by iteratively solving the global optimal solution of the subproblem, and finally estimating a complete flow matrix X.
Further, the specific steps of the step 4) are as follows:
step 4.1), in order to conveniently solve the optimization problem, firstly introducing an error variable C and rewriting the error variable C into:
Figure BDA0002792144430000051
s.t.X≥0,diag(W)＝0,M＝X+C,PΩ(C)＝0
(5)
step 4.2), in order to solve the optimization problem of step 4.1), two indicator functions are defined as follows:
Figure BDA0002792144430000052
wherein g (X) and f (C) have the meaning that constraint terms are put into the objective function in order for the variables to satisfy the constraint. Thus, the above optimization problem can be converted into an equivalent penalty function form:
Figure BDA0002792144430000053
step 4.3), alternate optimization solution:
Figure BDA0002792144430000054
step 4.3.1), to solve X efficiently, without introducing variables D and S, making D ═ X and S ═ RX, the subproblems can be converted into an equivalent constraint optimization problem as follows (for simplicity, W is used here without affecting understandingk-1Abbreviated as W, betakAbbreviated β):
Figure BDA0002792144430000055
the sub-problem, the corresponding Lagrangian function, can be defined as:
Figure BDA0002792144430000061
if order
Figure BDA0002792144430000062
The corresponding optimization steps are:
updating X:
Figure BDA0002792144430000063
updating D:
Figure BDA0002792144430000064
updating S:
Figure BDA0002792144430000065
fourthly-fifth updating U1、U2：
Figure BDA0002792144430000066
Figure BDA0002792144430000067
New mut：
μt＝min(ρμt-1,μmax) (16)
Step 4.3.2), solving W:
Figure BDA0002792144430000068
it can be seen that each row of the W matrix is independent, each row can be separated, each subproblem can be regarded as a LASSO problem, and W is setiI-th column, X, representing WiI-th column of X, X-iThe matrix obtained by removing the ith column from X is shown. To satisfy constraint diag (W) 0, we find WiAt this time, the ith column of X does not participate in the calculation.
Figure BDA0002792144430000069
Step 4.3.3), solving C:
Figure BDA00027921444300000610
step 4.3.4), update β
βk＝min(ρβk-1,βmax) (20)
And 4.4) reaching the maximum iteration times, and obtaining the estimated complete flow matrix X after the solution is finished.
The algorithm steps for sorting out the traffic estimation based on the time sequence prior and the sparse representation are shown in fig. 2:
Figure BDA0002792144430000071
in the field of flow estimation, the estimation missing value algorithm has advantages and disadvantages, and the IP network flow estimation method based on time sequence prior and sparse representation can effectively utilize time sequence prior information and sparse representation theory, excavate the inherent space-time correlation characteristic in the flow matrix and improve the accuracy of flow matrix estimation.

Claims (6)

1. An IP network flow estimation method based on time sequence prior and sparse representation is characterized by comprising the following steps:
s1, acquiring traffic values transmitted among all source-destination nodes in the network to construct an incomplete traffic matrix;
s2, establishing a flow matrix estimation model aiming at space-time correlation existing in an incomplete flow matrix by utilizing a sparse representation theory and a regularization technology;
s3, converting the original flow matrix estimation problem into a plurality of sub-problems easy to solve by an alternating direction multiplier method;
and S4, iteratively optimizing the global optimal solution of the sub-problem to find the local optimal solution of the original problem, and estimating a complete flow matrix.
2. The method for estimating IP network traffic based on timing priors and sparse representation according to claim 1, wherein the incomplete traffic matrix in step S1 is constructed by the following steps:
assuming that the number of time intervals of one day is T and the total number of OD pairs is N, the traffic matrix can be represented as
Figure FDA0002792144420000011
Figure FDA0002792144420000012
Wherein m isijRepresents the flow value of the jth OD pair at the ith time interval, "# represents a known flow value,"? "represents missing flow values, and in the flow matrix, one column represents one sample, one OD pair for all time intervals of the day, and one row represents the flow values of all OD pairs for one time interval.
3. The method for estimating IP network traffic based on timing prior and sparse representation according to claim 1, wherein the step S2 comprises the following steps:
firstly, establishing a flow estimation model based on a sparse representation theory, wherein the expression is as follows:
Figure FDA0002792144420000013
wherein,
Figure FDA0002792144420000014
in order for the traffic matrix to be known to be incomplete,
Figure FDA0002792144420000015
for the complete traffic matrix to be solved,
Figure FDA0002792144420000016
the coefficient matrix is expressed for sparseness needing to be solved, the omega set expresses a known flow value element subscript set in the flow matrix, and lambda1Is an adjustable parameter; pΩ(. cndot.) is a projection operator, indicating that when the element index (i, j) ∈ Ω, the corresponding position sample element is obtained:
Figure FDA0002792144420000017
since the element values in the traffic matrix are not negative, X ≧ 0, and to avoid trivial solution, the diagonal elements of the constraint sparse representation matrix W are all 0, i.e., diag (W) ═ 0;
establishing a flow estimation model based on time sequence prior and sparse representation, wherein the expression is as follows:
Figure FDA0002792144420000018
wherein λ is2In order to be able to adjust the parameters,
Figure FDA0002792144420000021
is a Toeplitz (0,1, -1) matrix.
4. The IP network traffic estimation method based on time series prior and sparse representation according to claim 1, wherein in step S3, the sub-problems include a traffic matrix X in an iterative process, a sparse representation coefficient matrix W and an error variable C representing the error of the incomplete traffic matrix M and the complete traffic matrix X outside the set Ω,
Figure FDA0002792144420000022
5. the method for estimating the IP network traffic based on timing prior and sparse representation according to claim 4, wherein the step S4 comprises the following specific steps:
s41, introducing an error variable C for solving conveniently, and rewriting the flow estimation model into the following form:
Figure FDA0002792144420000023
s.t.X≥0，diag(W)＝0，M＝X+C，PΩ(C)＝0
s42, putting constraint terms into an objective function, defining indicating functions g (X) and f (C), and converting the optimization problem into an equivalent penalty function form, wherein the expression of the indicating functions is as follows:
Figure FDA0002792144420000024
the penalty function is expressed as follows:
Figure FDA0002792144420000025
s43, obtaining the expression of X, W, C and beta to be solved according to the penalty function as follows:
Figure FDA0002792144420000026
where ρ is a fixed parameter, β is a parameter of a penalty function, βk，βk-1Beta values, beta, in the k-th iteration and the k-1 th iteration, respectivelymaxIs a fixed parameter representing the maximum value of β, F is the F norm; and carrying out alternate optimization solution on the formula, wherein the flow matrix X when the preset maximum iteration times is reached is the estimated complete flow matrix.
6. The method as claimed in claim 5, wherein in the solving process of the sparse representation matrix W, each column is solved separately, the solving process of each column is regarded as a LASSO problem, and W is setiI-th column, X, representing WiI-th column of X, X-iThe matrix obtained by removing the ith column from X is shown, and W is solved so as to satisfy the constraint diag (W) of 0iWhen X is not involved in the calculation, WiThe expression of (a) is as follows:
Figure FDA0002792144420000031
CN202011318745.4A 2020-11-23 2020-11-23 IP network flow estimation method based on time sequence prior and sparse representation Active CN112564945B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011318745.4A CN112564945B (en) 2020-11-23 2020-11-23 IP network flow estimation method based on time sequence prior and sparse representation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011318745.4A CN112564945B (en) 2020-11-23 2020-11-23 IP network flow estimation method based on time sequence prior and sparse representation

Publications (2)

Publication Number Publication Date
CN112564945A true CN112564945A (en) 2021-03-26
CN112564945B CN112564945B (en) 2023-03-24

Family

ID=75044689

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011318745.4A Active CN112564945B (en) 2020-11-23 2020-11-23 IP network flow estimation method based on time sequence prior and sparse representation

Country Status (1)

Country Link
CN (1) CN112564945B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115452081A (en) * 2022-09-22 2022-12-09 西安交通大学 Two-phase flow gas phase mass flow measuring method and system based on differential pressure throttling device

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106844518A (en) * 2016-12-29 2017-06-13 天津中科智能识别产业技术研究院有限公司 A kind of imperfect cross-module state search method based on sub-space learning
CN107133930A (en) * 2017-04-30 2017-09-05 天津大学 Ranks missing image fill method with rarefaction representation is rebuild based on low-rank matrix

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106844518A (en) * 2016-12-29 2017-06-13 天津中科智能识别产业技术研究院有限公司 A kind of imperfect cross-module state search method based on sub-space learning
CN107133930A (en) * 2017-04-30 2017-09-05 天津大学 Ranks missing image fill method with rarefaction representation is rebuild based on low-rank matrix

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Y. GUO, X. WANG, M. WANG AND H. ZHANG: "An_Improved_Low_Rank_Matrix_Completion_Method_for_Traffic_Data", 《2018 11TH INTERNATIONAL CONFERENCE ON INTELLIGENT COMPUTATION TECHNOLOGY AND AUTOMATION (ICICTA), 2018, PP. 255-260, DOI: 10.1109/ICICTA.2018.00064.》 *
何燕文等: "基于矩阵分解的电量数据恢复研究", 《电力信息与通信技术》 *
杨京礼等: "一种稀疏度自适应的网络流量矩阵测量方法", 《哈尔滨工业大学学报》 *
牛艳宁: "网络流量的新模型研究", 《中国优秀硕士学位论文全文数据库》 *
蒋定德等: "大尺度IP骨干网络流量矩阵估计方法研究", 《电子学报》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115452081A (en) * 2022-09-22 2022-12-09 西安交通大学 Two-phase flow gas phase mass flow measuring method and system based on differential pressure throttling device

Also Published As

Publication number Publication date
CN112564945B (en) 2023-03-24

Similar Documents

Publication Publication Date Title
CN105825230B (en) Forecast of Soil Moisture Content method and system based on depth confidence network model
CN108010320A (en) A kind of complementing method of the road grid traffic data based on adaptive space-time constraint low-rank algorithm
CN112950924B (en) Complex traffic network traffic speed prediction method based on deep learning
CN107220764A (en) A kind of electricity sales amount Forecasting Methodology compensated based on preamble analysis and factor and device
CN108304685A (en) A kind of non-linear degradation equipment method for predicting residual useful life and system
CN112132430B (en) Reliability evaluation method and system for distributed state sensor of power distribution main equipment
CN115186488A (en) Time-space modeling method and system for lithium battery temperature field
CN115495991A (en) Rainfall interval prediction method based on time convolution network
CN110401978A (en) Indoor orientation method based on neural network and particle filter multi-source fusion
CN113094860A (en) Industrial control network flow modeling method based on attention mechanism
CN110490366A (en) Runoff forestry method based on variation mode decomposition and iteration decision tree
CN112564945B (en) IP network flow estimation method based on time sequence prior and sparse representation
CN114861928B (en) Quantum measurement method and device and computing equipment
CN116259172A (en) Urban road speed prediction method considering space-time characteristics of traffic network
CN110807508B (en) Bus peak load prediction method considering complex weather influence
CN116106761A (en) Real-time lithium ion battery electric quantity estimation method based on typical correlation analysis
CN112215495B (en) Pollution source contribution calculation method based on long-time and short-time memory neural network
CN114048546A (en) Graph convolution network and unsupervised domain self-adaptive prediction method for residual service life of aircraft engine
CN116319378B (en) Network traffic matrix estimation and model training method and system based on deep learning
CN110909492B (en) Sewage treatment process soft measurement method based on extreme gradient lifting algorithm
CN109061544B (en) Electric energy metering error estimation method
CN114598627A (en) Abnormal network information detection method based on knowledge graph
CN116306832A (en) Multimode generation antagonistic neural network modeling method and device for multidimensional sequence data
CN116937559A (en) Power system load prediction system and method based on cyclic neural network and tensor decomposition
CN117113068A (en) Simulation method and system for longitudinal displacement of bridge support

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