CN110460463B - Service quality prediction method and system based on deep neural network - Google Patents

Service quality prediction method and system based on deep neural network Download PDF

Info

Publication number
CN110460463B
CN110460463B CN201910671422.4A CN201910671422A CN110460463B CN 110460463 B CN110460463 B CN 110460463B CN 201910671422 A CN201910671422 A CN 201910671422A CN 110460463 B CN110460463 B CN 110460463B
Authority
CN
China
Prior art keywords
information
context variable
data
module
matrix
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910671422.4A
Other languages
Chinese (zh)
Other versions
CN110460463A (en
Inventor
李秉卓
叶春杨
周辉
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.)
Hainan University
Original Assignee
Hainan University
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 Hainan University filed Critical Hainan University
Priority to CN201910671422.4A priority Critical patent/CN110460463B/en
Publication of CN110460463A publication Critical patent/CN110460463A/en
Application granted granted Critical
Publication of CN110460463B publication Critical patent/CN110460463B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—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
    • 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
    • H04L43/50—Testing arrangements
    • H04L43/55—Testing of service level quality, e.g. simulating service usage

Abstract

The invention discloses a service quality prediction method and a service quality prediction system based on a deep neural network, and relates to the technical field of network services. The method comprises the following steps: inputting request context variable information, and coding the request context variable information in a coding module through an entity expression matrix to obtain an embedded request matrix; compressing the coded context variable information and extracting characteristics to obtain an embedded request vector of the request context variable information according to time sequence; sequentially inputting the embedded request vectors to an LSTM network module according to the time sequence so as to output the predicted data information of the deep neural network for the context variable information of the next input request; and sensing the predicted data information to obtain decoded predicted data information, and restoring the decoded predicted data information into service quality data and outputting the service quality data.

Description

Service quality prediction method and system based on deep neural network
Technical Field
The invention relates to the technical field of network service, in particular to a service quality prediction method and a service quality prediction system based on a deep neural network.
Background
The number of webservices in the internet has increased dramatically over the last decade, and with the increase in the number of public Web services, the application of the SOA (Service Oriented architecture) architecture has become very widespread. However, the complicated network environment of the internet makes the state of the service vary, so that effective service quality prediction is very important at this time.
Most of the conventional service quality prediction methods are based on a CF (Collaborative Filtering) model and an MF (Matrix factorization) model, and the available context data is relatively limited. Meanwhile, it is difficult to simply extend the two methods to support the prediction of multi-type QoS (Quality of Service) data. It also lacks corresponding modeling means for implicit variables that affect the QoS response sequence.
Disclosure of Invention
The invention mainly aims to provide a service quality prediction method and a service quality prediction system based on a deep neural network, aiming at fitting invisible factors influencing QoS data in the network and enabling the invisible factors to predict QoS data of the real world.
In order to achieve the above object, the present invention provides a service quality prediction method based on a deep neural network, comprising the following steps:
inputting request context variable information, and coding the request context variable information in a coding module through an entity expression matrix to obtain an embedded request matrix;
compressing the coded context variable information and extracting characteristics to obtain an embedded request vector of the request context variable information according to time sequence;
sequentially inputting the embedded request vectors to an LSTM (Long Short-Term Memory) network module according to the time sequence so as to output the predicted data information of the deep neural network for the context variable information of the next input request;
and sensing the predicted data information to obtain decoded predicted data information, and restoring the decoded predicted data information into service quality data and outputting the service quality data.
Preferably, the context variable information includes a context and its corresponding quality of service data information.
Preferably, the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
Preferably, the step of performing dimension compression and feature extraction on the encoded context variable information further includes:
and compressing the coded context variable information, and inputting the compressed context variable information into a feature extractor consisting of a multilayer full-connection network for feature extraction.
Preferably, sequentially inputting the embedded request vector to the LSTM network module according to the time sequence to output the predicted data information of the deep neural network for the next input request context variable information further includes:
and sequentially obtaining corresponding calculation results through calculation according to the embedded request vectors sequentially input according to the time sequence, and taking the last calculation result in the calculation results as the predicted data information of the deep neural network for requesting context variable information for the next time.
Preferably, the restoring the decoded prediction data information into quality of service data and outputting further includes:
and respectively inputting the predicted data information into a restoration module formed by two fully-connected networks for calculation to obtain service quality data comprising response time information and network throughput information.
The invention also provides a service quality prediction system based on the deep neural network, which comprises an encoding module, a compression module, an LSTM network module, a perception module and a restoration module;
the encoding module is used for receiving the request context variable information, encoding and outputting an embedded request matrix to the compression module;
the compression module is used for compressing the coded context variable information, extracting features and outputting an embedded request vector to the LSTM network module;
the LSTM network module calculates the embedded request vector and outputs the predicted data information of the deep neural network for the next input request context variable information to the perception module;
the perception module perceives the predicted data information, and sends the decoded predicted data information to a restoration module, and the restoration module restores the decoded predicted data information into service quality data and outputs the service quality data.
Preferably, the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
Preferably, the restoring module includes a first restoring module and a second restoring module, which are used for restoring the response time information and the network throughput information respectively.
According to the technical scheme, the collected context variable information is input, invisible factors influencing QoS data in the network are fitted, and the QoS data in the real world can be predicted. The LSTM network module can predict complex unknown data, has certain correctness, and can be used in specific engineering practices such as service composition and the like.
Drawings
FIG. 1 is a schematic flow chart of a deep neural network-based QoS prediction method according to the present invention;
fig. 2 is a schematic diagram of the service quality prediction system based on the deep neural network according to the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention is further described below with reference to the accompanying drawings.
A service quality prediction method based on a deep neural network, as shown in fig. 1, includes the following steps:
and inputting request context variable information, and coding the request context variable information in a coding module through an entity expression matrix to obtain an embedded request matrix.
In an embodiment, the requested context variable information includes numeric and non-numeric features, which are preprocessed before being input to the encoding module. For continuous data, sampling is needed to convert the continuous data into discretization data so as to fuse the service quality data information and the request context information. For non-numeric variables, namely, the first six dimensions (u1, uas1, ur1, s1, sas1, sr1) in the octave obtained for each request context variable information, in the case that the total entity number N is 13, we can write it as a six-element group of (1, 4, 5, 6, 9, 11), and encode it to obtain N1(010011100101), i.e. for the first six-membered X of the octave1＝(x1，x2，x3，x4，x5，x6) Is mixing X1Corresponding to N1X ofiThe position code is 1. Defining an n x k dimensional element expression matrix Mn×kWherein M isiRepresenting the vectorized expression of the i-th entity, k representing the dimension of the expression vector, and for six tuples (1, 4, 5, 6, 9, 11), we select M1，M4，M6，M9，M11Six k-dimensional vectors form a 6 x k-dimensional matrix G1。G1Is a (1, 4, 5, 6, 9, 11) six-tuple matrixed expression, and the expression matrix can store a vector expression of each corresponding element. The operation is defined as follows:
Figure BDA0002141842390000051
s.t→len(N1) N, wherein the symbol
Figure BDA0002141842390000052
Indicating that all rows M are selected from the matrix MiIf and only if (X)1)iN is the number of all entities.
For response time and throughput of the numerical type, we use two matrices R that are independentrs×k，Tts×kAs its expression matrix. Where rs, ts are the number of response time and throughput expression vectors, respectively. In the specific embodiment, the step size is set to be 0.01 in discretization, and the step size is set to be 0.1 in throughput, so that the response time and the throughput of any value can be mapped to a corresponding integer expression, and the integer expression is used as the element identification of the integer expression. Specifically, corresponding to (0.13, 23.6), it is expressed as (13, 23), while R is selected from the R matrix13Selecting T from T matrix26Forming a 2 x k matrix G2. Namely for X2＝(x7，x8) Then N corresponding to the two can be obtained2、N3Selecting Rn7，Tn8Form a new matrix G2. The operation is determinedIt is defined as follows:
Figure BDA0002141842390000053
s.t→len(N2)＝max(ResponseTime)*100
s.t→len(N3) Max (throughput) × 10, where max (responsetime) max (throughput) represents the maximum of response time and throughput in all data, respectively. Directly splicing the matrix expressions obtained in the two steps, namely obtaining the matrix expression of one octave (1, 4, 5, 6, 9, 11, 13, 26), wherein G ═ concat (G)1，G2) The 8 × k matrix is the matrix representation of the request.
And compressing the coded context variable information and extracting the characteristics to obtain an embedded request vector of the request context variable information according to the time sequence.
In a specific embodiment, the input of the predetermined network is sd ═ sd1，sd2，sd3]Wherein sdiRepresenting the input of the ith step of the LSTM network module. Wherein each sdiCan be represented as sdi＝[sdi1，sdi2，…，sdik]The context variable information i is a matrix expression composed of k vectors in the context variable information i, and the sign extraction is to express the matrix expression sdiCompressed as vector expression Pi. When the preset i is 3, i.e. the first three requests are used to predict the fourth request, the goal is to put the three-dimensional tensor [ sd [ ]1，sd2，sd3]Compressed into two-dimensional tensor P ═ P1，P2，P3]。
And sequentially inputting the embedded request vectors to an LSTM network module according to the time sequence so as to output the predicted data information of the deep neural network for the context variable information of the next input request.
In a specific embodiment, [ P ] is input sequentially according to a time sequence1，P2，P3]To LSTM network module to obtain three outputs h1，h2，h3]Taking the third output h3As the final output of the network, we can consider here h3The predicted outcome of the network for the next requested data is shown. The calculation process is defined as follows:
Figure BDA0002141842390000061
ft＝σ(Wf*[Ct-1,ht-1,Pt]+bf)
it＝σ(Wi*[Ct-1,ht-1,Pt]+bi)
ot＝σ(Wo*[Ct-1,ht-1,Pt]+bo)
Figure BDA0002141842390000063
wherein h istFor the output of the LSTM network module in each step, ftTo forget the output of the gate, itTo the output of the input gate, otWhich represents the output of the output gate(s),
Figure BDA0002141842390000062
is a new cell state that has not been processed by the input gate.
And sensing the predicted data information to obtain decoded predicted data information, and restoring the decoded predicted data information into service quality data and outputting the service quality data. The restored service quality data can be conveniently identified and used by the user.
Preferably, the context variable information includes a context and its corresponding quality of service data information.
Preferably, the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
Preferably, the step of performing dimension compression and feature extraction on the encoded context variable information further includes: and compressing the coded context variable information, and inputting the compressed context variable information into a feature extractor consisting of a multilayer full-connection network for feature extraction.
Preferably, sequentially inputting the embedded request vector to the LSTM network module according to the time sequence to output the predicted data information of the deep neural network for the next input request context variable information further includes: and sequentially obtaining corresponding calculation results through calculation according to the embedded request vectors sequentially input according to the time sequence, and taking the last calculation result in the calculation results as the predicted data information of the deep neural network for requesting context variable information for the next time.
Preferably, the restoring the decoded prediction data information into quality of service data and outputting further includes: and respectively inputting the predicted data information into a restoration module formed by two fully-connected networks for calculation to obtain service quality data comprising response time information and network throughput information.
In a particular embodiment, Loss of absolute error function Loss is selected1The parameters were trained using the Adam optimization algorithm as a loss function of the LSTM network module. Wherein:
Figure BDA0002141842390000071
where m is the number of variables to be predicted, n is the total number of samples, abs () represents the absolute value of the number,
Figure BDA0002141842390000072
representing the value of the ith sample predicted for the kth prediction task,
Figure BDA0002141842390000073
representing the true values of the corresponding samples in the dataset. λ is the penalty coefficient, L2And (theta) is a regularization term.
As shown in fig. 2, an embodiment of the present invention further provides a deep neural network-based service quality prediction system, where the system includes an encoding module, a compression module, an LSTM network module, a sensing module, and a restoration module;
the encoding module is used for receiving the request context variable information, encoding and outputting an embedded request matrix to the compression module;
the compression module is used for compressing the coded context variable information, extracting features and outputting an embedded request vector to the LSTM network module;
the LSTM network module calculates the embedded request vector and outputs the predicted data information of the deep neural network for the next input request context variable information to the perception module;
the perception module perceives the predicted data information, and sends the decoded predicted data information to a restoration module, and the restoration module restores the decoded predicted data information into service quality data and outputs the service quality data.
Preferably, the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
Preferably, the restoring module includes a first restoring module and a second restoring module, which are used for restoring the response time information and the network throughput information respectively.
It should be understood that the above is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent structures or equivalent flow transformations made by the present specification and drawings, or applied directly or indirectly to other related technical fields, are included in the scope of the present invention.

Claims (8)

1. A service quality prediction method based on a deep neural network is characterized by comprising the following steps:
inputting request context variable information, and coding the request context variable information in a coding module through an entity expression matrix to obtain an embedded request matrix;
compressing the coded context variable information and extracting characteristics to obtain an embedded request vector of the request context variable information according to time sequence;
sequentially inputting the embedded request vectors to an LSTM network module according to the time sequence so as to output the predicted data information of the deep neural network for the context variable information of the next input request;
sensing the predicted data information to obtain decoded predicted data information, restoring the decoded predicted data information into service quality data and outputting the service quality data;
the context variable information comprises context and service quality data information corresponding to the context; the embedded request matrix is G ═ concat (G)1，G2)，G1The operation is defined as follows:
Figure FDA0003211465580000011
s.t→len(N1) N, wherein the symbol
Figure FDA0003211465580000012
Indicating that all rows M are selected from the matrix MiIf and only if (X)1)i1, n is the number of all entities; n is a radical of1(010011100101); the M matrix is an entity expression matrix, and the expression is as follows: mn×kWherein M isiRepresenting the vectorized expression of the ith entity, and k represents the dimension of the expression vector;
G2the operation is defined as follows:
Figure FDA0003211465580000013
s.t→len(N2)＝max(ResponseTime)*100
s.t→len(N3) Max (throughput) × 10, where max (responsetime) max (throughput) respectively represents the maximum value of response time and throughput in all data, and N2 and N3 are matrices corresponding to the 7 th and eighth dimensions of the octave group respectively;
wherein the context variable information comprises non-numerical data, the non-linear numerical data comprises octave, the octave refers to a matrix with eight-dimensional data, X1 represents the first six-element of the octave, X1＝(x1，x2，x3，x4，x5，x6)，(X1)i1 means that X is1Corresponding to N1X ofiThe position code is 1;
r is Rrs×kIs an expression matrix of response time of numerical data, T is Tts×kAn expression matrix that refers to the throughput of numerical data; rs, ts are the number of response time and throughput expression vectors, respectively.
2. The deep neural network-based service quality prediction method of claim 1, wherein the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
3. The method of claim 2, wherein the step of performing dimension compression and feature extraction on the encoded context variable information further comprises:
and compressing the coded context variable information, and inputting the compressed context variable information into a feature extractor consisting of a multilayer full-connection network for feature extraction.
4. The method of claim 3, wherein the sequentially inputting the embedded request vectors to the LSTM network module according to the time sequence to output the predicted data information of the next input request context variable information by the deep neural network further comprises:
and sequentially obtaining corresponding calculation results through calculation according to the embedded request vectors sequentially input according to the time sequence, and taking the last calculation result in the calculation results as the predicted data information of the deep neural network for requesting context variable information for the next time.
5. The method of claim 4, wherein the restoring the decoded prediction data information into quality of service data and outputting further comprises:
and respectively inputting the predicted data information into a restoration module formed by two fully-connected networks for calculation to obtain service quality data comprising response time information and network throughput information.
6. A service quality prediction system based on a deep neural network is characterized by comprising an encoding module, a compression module, an LSTM network module, a perception module and a restoration module;
the encoding module is used for receiving the request context variable information, encoding and outputting an embedded request matrix to the compression module;
the compression module is used for compressing the coded context variable information, extracting features and outputting an embedded request vector to the LSTM network module;
the LSTM network module calculates the embedded request vector and outputs the predicted data information of the deep neural network for the next input request context variable information to the perception module;
the perception module perceives the predicted data information and sends the decoded predicted data information to a restoration module, and the restoration module restores the decoded predicted data information into service quality data and outputs the service quality data;
the context variable informationThe information comprises context and service quality data information corresponding to the context; the embedded request matrix is G ═ concat (G)1，G2)，G1The operation is defined as follows:
Figure FDA0003211465580000031
s.t→len(N1) N, wherein the symbol
Figure FDA0003211465580000032
Indicating that all rows M are selected from the matrix MiIf and only if (X)1)i1, n is the number of all entities; n is a radical of1(010011100101); the M matrix is an entity expression matrix, and the expression is as follows: mn×kWherein M isiRepresenting the vectorized expression of the ith entity, and k represents the dimension of the expression vector;
G2the operation is defined as follows:
Figure FDA0003211465580000033
s.t→len(N2)＝max(ResponseTime)*100
s.t→len(N3) Max (throughput) × 10, where max (responsetime) max (throughput) respectively represents the maximum value of response time and throughput in all data, and N2 and N3 are matrices corresponding to the 7 th and eighth dimensions of the octave group respectively;
wherein the context variable information comprises non-numerical data, the non-linear numerical data comprises octave, the octave refers to a matrix with eight-dimensional data, X1 represents the first six-element of the octave, X1＝(x1，x2，x3，x4，x5，x6)，(X1)i1 means that X is1Corresponding to N1X ofiThe position code is 1;
r is Rrs×kIs a numerical typeAn expression matrix of response times of the data, T being Tts×kAn expression matrix that refers to the throughput of numerical data; rs, ts are the number of response time and throughput expression vectors, respectively.
7. The deep neural network-based quality of service prediction system of claim 6, wherein the context variable information includes response time information and network throughput information; the restored quality of service data includes response time information and network throughput information.
8. The deep neural network-based quality of service prediction system of claim 7, wherein the restoration module comprises a first restoration module and a second restoration module to restore response time information and network throughput information, respectively.
CN201910671422.4A 2019-07-24 2019-07-24 Service quality prediction method and system based on deep neural network Active CN110460463B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910671422.4A CN110460463B (en) 2019-07-24 2019-07-24 Service quality prediction method and system based on deep neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910671422.4A CN110460463B (en) 2019-07-24 2019-07-24 Service quality prediction method and system based on deep neural network

Publications (2)

Publication Number Publication Date
CN110460463A CN110460463A (en) 2019-11-15
CN110460463B true CN110460463B (en) 2021-12-14

Family

ID=68483279

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910671422.4A Active CN110460463B (en) 2019-07-24 2019-07-24 Service quality prediction method and system based on deep neural network

Country Status (1)

Country Link
CN (1) CN110460463B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111541570B (en) * 2020-04-22 2021-05-07 北京交通大学 Cloud service QoS prediction method based on multi-source feature learning
CN114124554B (en) * 2021-11-29 2022-08-30 燕山大学 Virtual network service chain throughput prediction method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108024158A (en) * 2017-11-30 2018-05-11 天津大学 There is supervision video abstraction extraction method using visual attention mechanism
CN109409499A (en) * 2018-09-20 2019-03-01 北京航空航天大学 One kind being based on deep learning and the modified track restoration methods of Kalman filtering
CN109858972A (en) * 2019-02-13 2019-06-07 重庆金窝窝网络科技有限公司 The prediction technique and device of ad click rate

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10521697B2 (en) * 2017-09-29 2019-12-31 Konica Minolta Laboratory U.S.A., Inc. Local connectivity feature transform of binary images containing text characters for optical character/word recognition

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108024158A (en) * 2017-11-30 2018-05-11 天津大学 There is supervision video abstraction extraction method using visual attention mechanism
CN109409499A (en) * 2018-09-20 2019-03-01 北京航空航天大学 One kind being based on deep learning and the modified track restoration methods of Kalman filtering
CN109858972A (en) * 2019-02-13 2019-06-07 重庆金窝窝网络科技有限公司 The prediction technique and device of ad click rate

Also Published As

Publication number Publication date
CN110460463A (en) 2019-11-15

Similar Documents

Publication Publication Date Title
CN110659744A (en) Training event prediction model, and method and device for evaluating operation event
CN111079532A (en) Video content description method based on text self-encoder
CN110460463B (en) Service quality prediction method and system based on deep neural network
CN113905391A (en) Ensemble learning network traffic prediction method, system, device, terminal, and medium
CN112884230B (en) Power load prediction method and device based on multivariate time sequence and related components
CN113537456B (en) Depth feature compression method
CN110599766A (en) Road traffic jam propagation prediction method based on SAE-LSTM-SAD
CN110719289B (en) Industrial control network intrusion detection method based on multilayer feature fusion neural network
CN109902808B (en) Method for optimizing convolutional neural network based on floating point digital variation genetic algorithm
CN116645668B (en) Image generation method, device, equipment and storage medium
WO2023279694A1 (en) Vehicle trade-in prediction method, apparatus, device, and storage medium
CN113255908A (en) Method, neural network model and device for service prediction based on event sequence
CN112131888A (en) Method, device and equipment for analyzing semantic emotion and storage medium
CN109993109A (en) Image character recognition method
Dumouchelle et al. Neur2sp: Neural two-stage stochastic programming
CN110175338B (en) Data processing method and device
CN114445121A (en) Advertisement click rate prediction model construction and advertisement click rate prediction method
CN111597401B (en) Data processing method, device, equipment and medium based on graph relation network
CN116600119B (en) Video encoding method, video decoding method, video encoding device, video decoding device, computer equipment and storage medium
CN116798536A (en) Training method and device for molecular multi-conformation prediction model
CN113342868B (en) Information recommendation method, device, equipment and computer readable storage medium
CN111222722B (en) Method, neural network model and device for business prediction for business object
CN110659962B (en) Commodity information output method and related device
Sentz et al. Reduced basis approximations of parameterized dynamical partial differential equations via neural networks
CN113255891B (en) Method, neural network model and device for processing event characteristics

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