CN114492906A - Training method of time sequence data prediction model, time sequence data prediction method and device - Google Patents

Training method of time sequence data prediction model, time sequence data prediction method and device Download PDF

Info

Publication number
CN114492906A
CN114492906A CN202011272716.9A CN202011272716A CN114492906A CN 114492906 A CN114492906 A CN 114492906A CN 202011272716 A CN202011272716 A CN 202011272716A CN 114492906 A CN114492906 A CN 114492906A
Authority
CN
China
Prior art keywords
time
series data
time series
training
prediction model
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.)
Pending
Application number
CN202011272716.9A
Other languages
Chinese (zh)
Inventor
王浩彬
李晨
黄芬芬
王田媛
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.)
China Telecom Corp Ltd
Original Assignee
China Telecom Corp Ltd
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 China Telecom Corp Ltd filed Critical China Telecom Corp Ltd
Priority to CN202011272716.9A priority Critical patent/CN114492906A/en
Publication of CN114492906A publication Critical patent/CN114492906A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The disclosure provides a training method of a time sequence data prediction model, a time sequence data prediction method and a time sequence data prediction device, and relates to the field of artificial intelligence. Inputting training time series data (X, z)t)，ztIs time series data corresponding to time t, X is time t, time before time t, and ztA related feature; is provided with
Figure DDA0002774548570000011
fiIs the ith function in the function library, alphaiIs the coefficient of the ith function, ytIs residual error time sequence data corresponding to the time t; is provided with
Figure DDA0002774548570000012
Figure DDA0002774548570000013
n and p are hyperparameters, yt‑jIs the residual time series data corresponding to the time (t-j) (. beta.)jRepresents yt‑jThe coefficient of (a) is determined,
Figure DDA0002774548570000014
is ytThe predicted value of (2); predicting models using time series data
Figure DDA0002774548570000015
Calculating the predicted value of the time series data corresponding to the t moment
Figure DDA0002774548570000016
According to ztAnd
Figure DDA0002774548570000017
using fitting method to alphaiAnd betajAnd training and outputting the trained time sequence data prediction model. Therefore, accurate prediction of non-stationary time series data is realized through a stationary transformation part and a stationary time series data prediction part of the time series data prediction model.

Description

Training method of time sequence data prediction model, time sequence data prediction method and device
Technical Field
The disclosure relates to the field of artificial intelligence, and in particular relates to a training method of a time sequence data prediction model, a time sequence data prediction method and a time sequence data prediction device.
Background
Accurate prediction of resources is an important means for guaranteeing normal and efficient operation of the cloud platform and the virtualization platform. Accurate resource prediction can warn operation and maintenance personnel to update and expand physical resources in time, expand and shrink virtualized resources and the like.
Disclosure of Invention
The embodiment of the disclosure provides a non-stable time sequence data prediction scheme capable of predicting resource data such as physical resources or virtualized resources of a cloud platform and a virtualized platform.
Some embodiments of the present disclosure provide a training method for a time series data prediction model, including:
inputting training time series data (X, z)t)，ztRepresents time series data corresponding to time t, X represents time t, time before time t, and ztA related feature;
is provided with
Figure BDA0002774548550000011
fiRepresenting the ith function, alpha, in a library of functionsiIs the coefficient of the ith function, n is a hyperparameter, ytRepresenting residual error time sequence data corresponding to the t moment;
is provided with
Figure BDA0002774548550000012
p is a hyperparameter, yt-jRepresents the residual time series data, beta, corresponding to the time (t-j)jRepresents yt-jThe coefficient of (a) is determined,
Figure BDA0002774548550000013
the predicted value of residual error time sequence data corresponding to t moment is represented;
predicting models using time series data
Figure BDA0002774548550000014
Calculating the predicted value of the time series data corresponding to the t moment
Figure BDA0002774548550000015
According to ztAnd
Figure BDA0002774548550000016
using fitting method to alphaiAnd betajTraining is carried out, and a trained time sequence data prediction model is output
Figure BDA0002774548550000017
In some embodiments, the residual timing data for each time instant constitutes stationary timing data.
In some embodiments, the timing data (X, z) for trainingt) For the resource data for training, ztShowing resource data corresponding to the time t, wherein X shows the time t, the time before the time t and the related characteristics of the resource data; and the time sequence data prediction model obtained by training is a resource data prediction model.
In some embodiments, the functions in the library of functions include exponential functions, logarithmic functions, linear functions, trigonometric functions.
In some embodiments, the training time series data (X, z) is input when the time series data prediction model is a memory resource growth amount prediction modelt) In (1)X is the time t, the time before the time t and the characteristic related to the corresponding memory resource growth amount, and the input time series data (X, z) for trainingt) Z intFunction f selected for increasing storage resources corresponding to time tiIncluding exponential and linear functions.
Some embodiments of the present disclosure provide a time series data prediction method, including:
inputting historical time sequence data X to be predicted into a time sequence data prediction model obtained based on training of any one of claims 1-4
Figure BDA0002774548550000021
Obtaining the predicted value of the time series data corresponding to the t moment output by the time series data prediction model
Figure BDA0002774548550000025
In some embodiments, the time series data prediction model utilizes
Figure BDA0002774548550000022
Performing first prediction on historical time series data X; time series data prediction model utilization
Figure BDA0002774548550000023
Figure BDA0002774548550000024
Performing second prediction on residual error time sequence data of the historical time sequence data; the time series data prediction model combines the first prediction result and the second prediction result as the final time series data prediction value corresponding to the t moment
Figure BDA0002774548550000026
In some embodiments, further comprising: predicted value of time series data corresponding to time t
Figure BDA0002774548550000027
Corresponding to the time tAnd when the source data is predicted, updating the corresponding resource based on the predicted value of the corresponding resource data at the time t.
Some embodiments of the present disclosure provide a time series data prediction apparatus, including: a memory; and a processor coupled to the memory, the processor configured to perform a training method of a time series data prediction model or a time series data prediction method based on instructions stored in the memory.
Some embodiments of the present disclosure provide a time series data prediction apparatus, including: one or more of a training module or a prediction module;
a training module configured to perform a training method of the time series data prediction model;
a prediction module configured to perform a time series data prediction method.
Some embodiments of the present disclosure provide a non-transitory computer-readable storage medium having stored thereon a computer program that, when executed by a processor, implements a training method of a time series data prediction model or a time series data prediction method.
Therefore, accurate prediction of non-stationary time series data is realized through a stationary transformation part and a stationary time series data prediction part of the time series data prediction model.
Drawings
The drawings that will be used in the description of the embodiments or the related art will be briefly described below. The present disclosure can be understood more clearly from the following detailed description, which proceeds with reference to the accompanying drawings.
It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those drawings by one of ordinary skill in the art without undue inventive faculty.
Fig. 1 illustrates a flow diagram of a method of training a time series data prediction model according to some embodiments of the present disclosure.
Fig. 2 illustrates a flow diagram of a method of temporal data prediction according to some embodiments of the present disclosure.
Fig. 3 is a schematic structural diagram of a time series data prediction apparatus according to some embodiments of the disclosure.
Fig. 4 is a schematic structural diagram of a time series data prediction apparatus according to another embodiment of the disclosure.
Detailed Description
The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure.
Unless otherwise specified, "first", "second", and the like in the present disclosure are described to distinguish different objects, and are not intended to mean size, timing, or the like.
According to the embodiment of the disclosure, the time sequence data prediction model is obtained through historical time sequence data training, and the time sequence data is predicted by using the time sequence data prediction model. The time sequence data includes resource data such as physical resources or virtualized resources of the cloud platform and the virtualized platform, for example. The resource data includes, for example, CPU, memory, bandwidth, and the like. The present disclosure is not limited to the enumerated examples of the time series data, the enumerated examples of the resource data.
In an embodiment of the disclosure, a time series data prediction model includes a stationary transformation portion and a stationary time series data prediction portion formed based on a priori experience. The stationary transformation part can perform a first prediction on non-stationary historical time series data based on a priori experience and transform the non-stationary historical time series data into stationary residual time series data. The stationary time series data prediction section can perform a second prediction on stationary residual time series data of the history time series data. And combining the first prediction result and the second prediction result to obtain a final prediction result of the time sequence data. Therefore, accurate prediction of non-stationary time series data is achieved through two parts of the time series data prediction model.
Fig. 1 illustrates a flow diagram of a method of training a time series data prediction model according to some embodiments of the present disclosure.
As shown in fig. 1, the training method of the time series data prediction model of this embodiment includes: step 110-.
In step 110, time series data (X, z) for training is inputt)，ztIndicating the time series data corresponding to the time t,x represents time t, time before time t, and ztRelated features.
The time series data for training is, for example, non-stationary time series data, such as the aforementioned resource data of physical resources or virtualized resources of the cloud platform and the virtualized platform. E.g. timing data (X, z) for trainingt) For the resource data for training, ztShowing resource data corresponding to the time t, wherein X shows the time t, the time before the time t and the related characteristics of the resource data; and the time sequence data prediction model obtained by training is a resource data prediction model.
X for example comprises a plurality of pieces of history data with time stamps before time t.
At step 120, set up
Figure BDA0002774548550000041
fiRepresenting the ith function, alpha, in a library of functionsiIs the coefficient of the ith function, n is a hyperparameter, ytAnd (3) residual time series data corresponding to the time t is shown.
The functions in the library of functions include, but are not limited to, exponential functions, logarithmic functions, linear functions, trigonometric functions, and the like. f. ofiThe particular function selected may be indicated.
αiAre coefficients that need to be trained. Before training begins, some initial values can be set, and alpha is continuously adjusted and optimized in the subsequent training processiThe value of (a).
The hyper-parameter n may be preset according to the service.
In step 130, set up
Figure BDA0002774548550000042
p is a hyperparameter, yt-jRepresents the residual time series data, beta, corresponding to the time (t-j)jRepresents yt-jThe coefficient of (a) is determined,
Figure BDA00027745485500000515
and (3) the predicted value of the residual time series data corresponding to the time t is shown.
βjIs also required to be trainedThe coefficient of refining. Before training begins, some initial values can be set, and beta is continuously adjusted and optimized in the subsequent training processjThe value of (a).
The hyper-parameter p may be pre-set according to the service.
If the first prediction is
Figure BDA0002774548550000051
Then y ist-1、yt-2、yt-pThe value of (A) can adopt a preset initial value; then based on
Figure BDA0002774548550000052
Making subsequent predictions, e.g. prediction
Figure BDA0002774548550000053
And the like.
At step 140, the model is predicted using the time series data
Figure BDA0002774548550000054
Calculating the predicted value of the time series data corresponding to the t moment
Figure BDA0002774548550000055
In step 150, according to ztAnd
Figure BDA00027745485500000516
using fitting method to alphaiAnd betajTraining is carried out, and a trained time sequence data prediction model is output
Figure BDA0002774548550000056
That is, by continuously training and fitting alphaiAnd betajSo that
Figure BDA0002774548550000057
Closer and closer to ztWhen is coming into contact with
Figure BDA0002774548550000058
And ztWhen the difference between the time sequence data and the time sequence data is stably smaller than a preset threshold value, the training can be stopped, and the model at the moment is used as a time sequence data prediction model after the training.
In the time-series data prediction model,
Figure BDA0002774548550000059
to smooth the transformed portion based on a priori experience,
Figure BDA00027745485500000510
a portion is predicted for stationary timing data. Residual time series data y of each timetAnd yt-jAnd forming stable time sequence data.
The trained time sequence data prediction model can be used for accurately predicting non-stationary time sequence data.
Fig. 2 illustrates a flow diagram of a method of temporal data prediction according to some embodiments of the present disclosure.
As shown in fig. 2, the time series data prediction method of this embodiment includes: step 210 and 220, predicting the time series data corresponding to the time t
Figure BDA00027745485500000517
If the predicted value of the resource data is corresponding to time t, the method further includes step 230.
In step 210, the historical time series data X to be predicted is input into the time series data prediction model obtained by the training
Figure BDA00027745485500000511
Time series data prediction model utilization
Figure BDA00027745485500000512
A first prediction is made (210a) of historical time series data X, the time series data prediction model using
Figure BDA00027745485500000513
Second prediction (210b) is performed on residual time series data of the historical time series data, and the time series data prediction model is combinedUsing the first prediction result and the second prediction result as the final prediction values of the time series data corresponding to the t time
Figure BDA00027745485500000514
(210c)。
Since the training time series data X and the historical time series data X to be predicted both belong to time series data, the symbols herein are not distinguished.
In step 220, the predicted value of the time series data corresponding to the time t outputted by the time series data prediction model is obtained
Figure BDA0002774548550000065
In step 230, the predicted value of the corresponding time series data at time t
Figure BDA0002774548550000066
And when the predicted value of the resource data corresponding to the time t is the predicted value of the resource data corresponding to the time t, updating the corresponding resource based on the predicted value of the resource data corresponding to the time t, for example, updating and expanding the physical resource, and expanding and reducing the virtualized resource for the cloud platform or the virtualized platform.
The following describes the scheme of the present disclosure by taking an example of predicting the increase of the cloud platform storage device.
Time series data (X, z) for training input when the time series data prediction model is a storage resource increase amount prediction modelt) Wherein X is a characteristic related to time t, time before time t, and corresponding increase of storage resource, and the input training time series data (X, z)t) Z intFunction f selected for increasing storage resources corresponding to time tiIncluding exponential and linear functions. The characteristics related to the increase of the storage resources may be, for example, time characteristics, or characteristics of a CPU, a memory, a network, and the like, that is, the characteristics are performed based on the time characteristics
Figure BDA0002774548550000061
Or based on CPU characteristics, memory characteristics, and network characteristics
Figure BDA0002774548550000062
The first prediction of (1). The following is performed based on the time of day characteristics
Figure BDA0002774548550000063
The first prediction of (2) is described as an example.
Suppose the increase of the storage device is shown in table one, time 1-4 is training data, and time 5 is test data. And setting the super parameter n-2 and p-2.
Watch 1
Time of day Amount of increase in storage devices
1 5.77
2 5.74
3 8.01
4 11.51
5 21.24
Through training, a storage resource growth prediction model is obtained, and the method comprises the following steps: smoothing transformation part formed based on prior experience
Figure BDA0002774548550000064
And stationary time series data prediction section
Figure BDA0002774548550000071
Wherein the coefficient alpha is determined by training1、α2Respectively 0.1 and 0.5, coefficient beta1、β2Respectively 0.7 and 0.3.
The data at time 3, time 4, and time 5 are predicted using the data at time 1 and time 2 in table one as the basic data and the storage resource increase amount prediction model, and the prediction results are shown in table two. Taking the data at the predicted time 3 as an example, the calculation is performed based on X ═ 2 (i.e., time 2)
Figure BDA0002774548550000072
Based on yt-1、yt-2Equal to 5.74 and 5.77 respectively
Figure BDA0002774548550000073
The sum of the two results is used as the predicted value of the storage device growth at time 3
Figure BDA0002774548550000074
Watch two
Time of day Amount of increase in storage devices
1 5.77
2 5.74
3 7.81
4 11.67
5 21.58
Fig. 3 is a schematic structural diagram of a time series data prediction apparatus according to some embodiments of the disclosure.
As shown in fig. 3, the time series data prediction apparatus 300 of this embodiment includes: a memory 310 and a processor 320 coupled to the memory 310, the processor 320 being configured to perform a training method of the time series data prediction model or a time series data prediction method in any of the foregoing embodiments based on instructions stored in the memory 310.
Memory 310 may include, for example, a system memory, a fixed non-volatile storage medium, and so on. The system memory stores, for example, an operating system, an application program, a Boot Loader (Boot Loader), and other programs.
FIG. 4 is a schematic diagram of a time series data prediction apparatus according to another embodiment of the disclosure
As shown in fig. 4, the time series data prediction apparatus 400 of this embodiment includes: one or more of the training module 410 or the prediction module 420.
A training module 410 configured to perform a training method of the time series data prediction model.
A prediction module 420 configured to perform a time series data prediction method.
In some embodiments, the apparatus 400 may further include an execution module 430 configured to predict the corresponding time series data at time t
Figure BDA0002774548550000081
When the predicted value of the resource data corresponding to the time t is used, the corresponding resource is updated based on the predicted value of the resource data corresponding to the time tFor example, for a cloud platform or a virtualization platform, physical resources are updated and expanded, and virtualized resources are expanded and reduced.
In some embodiments, apparatus 400 may also include a training data storage module 440 and a data to predict storage module 450.
A training data storage module 440 configured to store time series data (X, z) for trainingt)。
A data to be predicted storage module 450 configured to store historical time series data X to be predicted.
The disclosure also proposes a non-transitory computer-readable storage medium on which a computer program is stored, which when executed by a processor implements a training method of a time series data prediction model or a time series data prediction method.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more non-transitory computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only exemplary of the present disclosure and is not intended to limit the present disclosure, so that any modification, equivalent replacement, or improvement made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.

Claims (11)

1. A training method of a time series data prediction model is characterized by comprising the following steps:
inputting training time series data (X, z)t)，ztRepresents time series data corresponding to time t, X represents time t, time before time t, and ztA related feature;
is provided with
Figure FDA0002774548540000011
fiRepresenting the ith function, alpha, in a library of functionsiIs the coefficient of the ith function, n is a hyperparameter, ytRepresenting residual error time sequence data corresponding to the t moment;
is provided with
Figure FDA0002774548540000012
p is a hyperparameter, yt-jRepresents the residual time series data, beta, corresponding to the time (t-j)jRepresents yt-jThe coefficient of (a) is determined,
Figure FDA0002774548540000013
the predicted value of residual error time sequence data corresponding to t moment is represented;
predicting models using time series data
Figure FDA0002774548540000014
Calculating the predicted value of the time series data corresponding to the t moment
Figure FDA0002774548540000015
According to ztAnd
Figure FDA0002774548540000016
using fitting method to alphaiAnd betajTraining is carried out, and a trained time sequence data prediction model is output
Figure FDA0002774548540000017
2. The method of claim 1, wherein the residual timing data for each time instant constitutes stationary timing data.
3. The method of claim 1,
time series data (X, z) for trainingt) For the resource data for training, ztShowing resource data corresponding to the time t, wherein X shows the time t, the time before the time t and the related characteristics of the resource data;
and the time sequence data prediction model obtained by training is a resource data prediction model.
4. The method of claim 1,
the functions in the function library include exponential functions, logarithmic functions, linear functions, and trigonometric functions.
5. The method according to any one of claims 1 to 4,
time series data (X, z) for training input when the time series data prediction model is a storage resource increase amount prediction modelt) Wherein X is a characteristic related to time t, time before time t, and corresponding increase of storage resource, and the input training time series data (X, z)t) Z intFunction f selected for increasing storage resource corresponding to t timeiIncluding exponential and linear functions.
6. A method for predicting time series data, comprising:
inputting historical time sequence data X to be predicted into a time sequence data prediction model obtained based on training of any one of claims 1-4
Figure FDA0002774548540000021
Obtaining the predicted value of the time series data corresponding to the t moment output by the time series data prediction model
Figure FDA0002774548540000026
7. The method of claim 6,
time series data prediction model utilization
Figure FDA0002774548540000022
Performing first prediction on historical time series data X;
time series data prediction model utilization
Figure FDA0002774548540000023
Performing second prediction on residual error time sequence data of the historical time sequence data;
the time series data prediction model combines the first prediction result and the second prediction result as the final t timePredicted value of corresponding time series data
Figure FDA0002774548540000024
8. The method of claim 6, further comprising:
predicted value of time series data corresponding to time t
Figure FDA0002774548540000025
And when the predicted value of the resource data corresponding to the time t is the predicted value of the resource data corresponding to the time t, updating the corresponding resource based on the predicted value of the resource data corresponding to the time t.
9. A time series data prediction apparatus comprising:
a memory; and
a processor coupled to the memory, the processor configured to perform the method of training the temporal data prediction model of any of claims 1-5 or the method of temporal data prediction of any of claims 6-8 based on instructions stored in the memory.
10. A time series data prediction apparatus comprising: one or more of a training module or a prediction module;
a training module configured to perform a training method of the time series data prediction model of any one of claims 1-5;
a prediction module configured to perform the time series data prediction method of any one of claims 6-8.
11. A non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of training a time series data prediction model of any one of claims 1-5 or the method of time series data prediction of any one of claims 6-8.
CN202011272716.9A 2020-11-12 2020-11-12 Training method of time sequence data prediction model, time sequence data prediction method and device Pending CN114492906A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011272716.9A CN114492906A (en) 2020-11-12 2020-11-12 Training method of time sequence data prediction model, time sequence data prediction method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011272716.9A CN114492906A (en) 2020-11-12 2020-11-12 Training method of time sequence data prediction model, time sequence data prediction method and device

Publications (1)

Publication Number Publication Date
CN114492906A true CN114492906A (en) 2022-05-13

Family

ID=81489814

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011272716.9A Pending CN114492906A (en) 2020-11-12 2020-11-12 Training method of time sequence data prediction model, time sequence data prediction method and device

Country Status (1)

Country Link
CN (1) CN114492906A (en)

Similar Documents

Publication Publication Date Title
CN109754060B (en) Training method and device for neural network machine learning model
CN110379416B (en) Neural network language model training method, device, equipment and storage medium
KR102170105B1 (en) Method and apparatus for generating neural network structure, electronic device, storage medium
EP3497625A1 (en) Aggregate features for machine learning
US11449731B2 (en) Update of attenuation coefficient for a model corresponding to time-series input data
CN111950810B (en) Multi-variable time sequence prediction method and equipment based on self-evolution pre-training
WO2018201151A1 (en) Neural network optimizer search
US20200380356A1 (en) Information processing apparatus, information processing method, and program
US10635078B2 (en) Simulation system, simulation method, and simulation program
CN111752903B (en) Prediction method for usable time of data storage space
US20210374544A1 (en) Leveraging lagging gradients in machine-learning model training
EP4111373A1 (en) Robust artificial intelligence inference in edge computing devices
CN105335375A (en) Topic mining method and apparatus
CN111967917A (en) Method and device for predicting user loss
CN113568379A (en) Control assistance device, control assistance method, computer-readable medium, and control system
CN115759230A (en) Model training and task processing method, device, system, equipment and storage medium
CN110276081B (en) Text generation method, device and storage medium
CN113641525A (en) Variable exception recovery method, apparatus, medium, and computer program product
CN114492906A (en) Training method of time sequence data prediction model, time sequence data prediction method and device
CN114492251B (en) Low-speed flow field divergence processing method, device, equipment and medium in supercomputing environment
JP2016018230A (en) Control parameter adaptation method and control parameter adaptation assist device
CN115640359A (en) Message synchronization control method and device, electronic equipment and storage medium
CN111027693A (en) Neural network compression method and system based on weight-removing pruning
AU2021219995B2 (en) Learning with moment estimation using different time constants
JP7424373B2 (en) Analytical equipment, analytical methods and analytical programs

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