WO2021212753A1 - Computer performance data determining method and apparatus, computer device, and storage medium - Google Patents

Computer performance data determining method and apparatus, computer device, and storage medium Download PDF

Info

Publication number
WO2021212753A1
WO2021212753A1 PCT/CN2020/118939 CN2020118939W WO2021212753A1 WO 2021212753 A1 WO2021212753 A1 WO 2021212753A1 CN 2020118939 W CN2020118939 W CN 2020118939W WO 2021212753 A1 WO2021212753 A1 WO 2021212753A1
Authority
WO
WIPO (PCT)
Prior art keywords
performance data
sequence
computer
historical performance
historical
Prior art date
Application number
PCT/CN2020/118939
Other languages
French (fr)
Chinese (zh)
Inventor
徐锐杰
Original Assignee
平安科技(深圳)有限公司
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 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021212753A1 publication Critical patent/WO2021212753A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00Error detection; Error correction; Monitoring
    • G06F11/30Monitoring
    • G06F11/34Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3447Performance evaluation by modeling
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/049Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent

Definitions

  • This application relates to blockchain technology, in particular to a method, device, computer equipment, and computer-readable storage medium for determining computer performance data.
  • performance (such as capacity) prediction is an important application scenario of AIOPS (AI-based IT operations).
  • AIOPS AI-based IT operations.
  • the existing performance prediction methods do not have high accuracy for cloud computing platform service vendors to perform performance predictions, and are not effective in long-term performance predictions.
  • a method for determining computer performance data comprising:
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • a device for determining computer performance data comprising:
  • the acquisition module is used to acquire multiple historical performance data sequences of the computer
  • the first training module is configured to use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain the pre-trained long and short-term memory network;
  • the classification module is configured to classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
  • the second training module is used to use each historical performance data sequence subset to train a performance prediction composed of the pre-trained long short-term memory network and the fully connected layer behind the pre-trained long short-term memory network. Judgment model to obtain N trained performance prediction models corresponding to the N sequence types one-to-one;
  • the prediction module is used to predict the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • a computer device includes a processor, and the processor implements the following steps when the processor is used to execute a computer program stored in a memory:
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • the fourth aspect of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • This application realizes the prediction of computer performance based on the historical performance data of the computer.
  • Fig. 1 is a flowchart of a method for determining computer performance data provided by an embodiment of the present application.
  • Fig. 2 is a structural diagram of a computer performance data determining apparatus provided by an embodiment of the present application.
  • Fig. 3 is a schematic diagram of a computer device provided by an embodiment of the present application.
  • the method for determining computer performance data of the present application is applied to one or more computer devices.
  • the computer device is a device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions.
  • Its hardware includes, but is not limited to, a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit). Specific Integrated Circuit, ASIC), Programmable Gate Array (Field-Programmable Gate Array, FPGA), Digital Signal Processor (DSP), embedded devices, etc.
  • the computer device may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the computer device can interact with the user through a keyboard, a mouse, a remote control, a touch panel, or a voice control device.
  • FIG. 1 is a flowchart of a method for determining computer performance data provided in Embodiment 1 of the present application.
  • the computer performance data determination method is applied to computer equipment to predict computer performance data based on historical computer performance data.
  • the computer performance data may include computer CPU data, computer GPU data, computer memory data, or computer storage data.
  • the method for determining computer performance data includes:
  • the acquiring multiple historical performance data sequences of the computer includes:
  • the next element of the historical performance data sequence in the main historical performance data sequence where each historical performance data sequence is located may be marked as the label of the historical performance data sequence.
  • the main sequence of historical performance data of a computer is ⁇ X1,X2,...,Xi,...,Xn ⁇ , where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length, and the Intercept multiple historical performance data sequences from each main sequence of historical performance data for the step size as ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,... Xh+1 ⁇ , ⁇ X3,X4,... Xh+2 ⁇ Etc.; historical performance data sequence ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,...
  • the tags of Xh+1 ⁇ , ⁇ X3, X4,... Xh+2 ⁇ are Xh+1, Xh+2, Xh+3, respectively.
  • Each historical performance data series is a time series.
  • the parameters in the long and short-term memory network can be initialized, each historical performance data sequence is input into the long- and short-term memory network, and the output value calculated by the long and short-term memory network according to the input value and the historical performance data can be calculated
  • the distance between the tags of the sequence is optimized according to the distance in the parameters in the long- and short-term memory network.
  • the method before using the multiple historical performance data sequences to pre-train the long- and short-term memory network, the method further includes:
  • Missing value preprocessing, abnormal data preprocessing, etc. may be performed on the multiple historical performance data sequences.
  • Performing missing value preprocessing on the historical performance data sequence set can prevent missing values in the multiple historical performance sequences and reduce the impact on training the long and short-term memory network.
  • Performing abnormal data preprocessing on the multiple historical performance data sequences increases the training convergence speed of the long- and short-term memory network, and reduces the time for training the long- and short-term memory network.
  • the historical performance data sequence can be preprocessed with missing values using interpolation or replacement methods.
  • the abnormal data When there is abnormal data in a historical performance data sequence (that is, the historical performance data sequence includes values that are not within the preset range), if the abnormal data is greater than the maximum value of the preset range, the abnormal data is set to the maximum value of the preset range Value; if the abnormal data is less than the minimum value of the preset range, the abnormal data is set to the minimum value of the preset range.
  • the classifying the multiple historical performance data sequences includes:
  • the multiple historical performance data sequences in the historical performance data sequence set may be clustered according to the DBScan clustering algorithm, where the distance function in the DBScan clustering algorithm uses one-dimensional NCC (normalized cross correlation) algorithm; when the clustering result is abnormal, delete the historical performance data sequence in the clustering result that is beyond the preset range from the cluster center according to the received delete instruction.
  • NCC normalized cross correlation
  • the clustering of the multiple historical performance data sequences includes:
  • the historical performance data sequence is divided into the cluster to which the center point sequence closest to the historical performance data sequence belongs.
  • the clustering of the multiple historical performance data sequences includes:
  • W-1 is greater than N, calculate the distance between every two clusters, merge the two clusters with the smallest distance, and get W-2 clusters; and so on, when W-R is equal to N, get N clusters.
  • the preset convolutional neural network model is backpropagated, and the parameters of the preset convolutional neural network model are optimized.
  • each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long and short-term memory network.
  • Performance prediction models include:
  • the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
  • the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network Performance prediction models include:
  • the method Before predicting the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, the method further includes:
  • the to-be-predicted performance data sequence of the computer also includes computer GPU performance, computer memory performance, or computer storage performance.
  • the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences includes:
  • the historical performance data sequence When training the preset neural network, the historical performance data sequence may be used as the input of the preset neural network, and the sequence type of the historical performance data sequence may be used as the label of the historical performance data sequence;
  • the labels of the output and historical performance data sequences are optimized through the back-propagation algorithm to optimize the parameters of the preset neural network.
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • the computer CPU performance sequence corresponds to the performance prediction model A, and the performance prediction model A can more accurately extract the characteristics of the computer CPU performance sequence (because the performance prediction model A is also trained according to the sequence type during the training process).
  • the computer performance data determination method of the first embodiment obtains a computer's historical performance data sequence set, the historical performance data sequence set includes a plurality of historical performance data sequences and a label of each historical performance data sequence; the historical performance data sequence set is used Perform pre-training on the long and short-term memory network to obtain the pre-trained long and short-term memory network; classify the multiple historical performance data sequences to obtain N sequence types and N sequence types corresponding to the N sequence types one-to-one Historical performance data sequence subsets; each historical performance data sequence subset is used to train a performance prediction composed of the pre-trained long-short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network To obtain the N performance prediction models corresponding to the N sequence types one-to-one; obtain the performance data sequence to be predicted of the computer; determine the sequence type of the performance data sequence to be predicted; The performance prediction model corresponding to the sequence type of the performance data sequence to be predicted predicts the performance data of the computer.
  • the first embodiment predicts the
  • the above-mentioned performance data may also be stored in a node of a blockchain.
  • the method further includes:
  • the method further includes:
  • the main sequence of historical CPU usage of a computer is ⁇ X1,X2,...,Xi,...,Xn ⁇ , where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length and the 1 is the step size to intercept multiple historical CPU usage sequence from each main sequence of historical CPU usage as ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,... Xh+1 ⁇ , ⁇ X3,X4,... Xh +2 ⁇ etc.; historical CPU usage rate sequence ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,...
  • the tags of Xh+1 ⁇ , ⁇ X3, X4,... Xh+2 ⁇ are Xh+1, Xh+2, Xh+3, respectively.
  • Each historical CPU usage sequence is a time sequence.
  • the parameters in the long and short-term memory network may be initialized, each historical CPU usage rate sequence is input into the long-term short-term memory network, and the output value calculated by the long-term short-term memory network according to the input value and the historical CPU Use the distance of the tags of the rate sequence to optimize the parameters in the long and short-term memory network according to the distance.
  • the classifying the multiple historical CPU usage rate sequences includes:
  • the sequence type of the CPU usage rate sequence to be predicted is determined according to the N historical CPU usage rate sequence subsets.
  • the computer CPU performance sequence corresponds to the prediction model A of the CPU usage rate.
  • Fig. 2 is a structural diagram of a computer performance data determining device provided in the second embodiment of the present application.
  • the computer performance data determining device 20 is applied to computer equipment.
  • the computer performance data determining device 20 is used to predict the computer performance data based on the historical performance data of the computer.
  • the computer performance data may include computer CPU data, computer GPU data, computer memory data, or computer storage data.
  • the device 20 for determining computer performance data may include an acquisition module 201, a first training module 202, a classification module 203, a second training module 204, and a pre-judgment module 205.
  • the obtaining module 201 is used to obtain multiple historical performance data sequences of the computer.
  • the acquiring multiple historical performance data sequences of the computer includes:
  • the next element of the historical performance data sequence in the main historical performance data sequence where each historical performance data sequence is located may be marked as the label of the historical performance data sequence.
  • the main sequence of historical performance data of a computer is ⁇ X1,X2,...,Xi,...,Xn ⁇ , where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length, and 1 Intercept multiple historical performance data sequences from each main sequence of historical performance data for the step size as ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,... Xh+1 ⁇ , ⁇ X3,X4,... Xh+2 ⁇ Etc.; historical performance data sequence ⁇ X1,X2,...Xh ⁇ , ⁇ X2,X3,...
  • the tags of Xh+1 ⁇ , ⁇ X3, X4,... Xh+2 ⁇ are Xh+1, Xh+2, Xh+3, respectively.
  • Each historical performance data series is a time series.
  • the first training module 202 is configured to use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain a pre-trained long- and short-term memory network.
  • the parameters in the long and short-term memory network can be initialized, each historical performance data sequence is input into the long- and short-term memory network, and the output value calculated by the long and short-term memory network according to the input value and the historical performance data can be calculated
  • the distance between the tags of the sequence is optimized according to the distance in the parameters in the long- and short-term memory network.
  • the method before using the multiple historical performance data sequences to pre-train the long- and short-term memory network, the method further includes:
  • Missing value preprocessing, abnormal data preprocessing, etc. may be performed on the multiple historical performance data sequences.
  • Performing missing value preprocessing on the historical performance data sequence set can prevent missing values in the multiple historical performance sequences and reduce the impact on training the long and short-term memory network.
  • Performing abnormal data preprocessing on the multiple historical performance data sequences increases the training convergence speed of the long- and short-term memory network, and reduces the time for training the long- and short-term memory network.
  • interpolation or replacement methods can be used to preprocess the missing values of the historical performance data sequence.
  • the abnormal data When there is abnormal data in a historical performance data sequence (that is, the historical performance data sequence includes values that are not within the preset range), if the abnormal data is greater than the maximum value of the preset range, the abnormal data is set to the maximum value of the preset range Value; if the abnormal data is less than the minimum value of the preset range, the abnormal data is set to the minimum value of the preset range.
  • the classification module 203 is configured to classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one.
  • the classifying the multiple historical performance data sequences includes:
  • the multiple historical performance data sequences in the historical performance data sequence set may be clustered according to the DBScan clustering algorithm, where the distance function in the DBScan clustering algorithm uses one-dimensional NCC (normalized cross correlation) algorithm; when the clustering result is abnormal, delete the historical performance data sequence in the clustering result that is beyond the preset range from the cluster center according to the received delete instruction.
  • NCC normalized cross correlation
  • the clustering of the multiple historical performance data sequences includes:
  • the historical performance data sequence is divided into the cluster to which the center point sequence closest to the historical performance data sequence belongs.
  • the clustering of the multiple historical performance data sequences includes:
  • W-1 is greater than N, calculate the distance between every two clusters, merge the two clusters with the smallest distance, and get W-2 clusters; and so on, when W-R is equal to N, get N clusters.
  • the preset convolutional neural network model is backpropagated, and the parameters of the preset convolutional neural network model are optimized.
  • the second training module 204 is configured to use each subset of historical performance data sequences to train a performance consisting of the pre-trained long-short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network
  • the predictive model obtains N trained performance predictive models corresponding to the N sequence types in a one-to-one manner.
  • each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long and short-term memory network.
  • Performance prediction models include:
  • the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
  • the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network Performance prediction models include:
  • the predicting module 205 is configured to predict the performance data of the computer through the trained performance predicting model corresponding to the sequence type of the to-be-predicted performance data sequence of the computer.
  • the method Before predicting the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, the method further includes:
  • the to-be-predicted performance data sequence of the computer also includes computer GPU performance, computer memory performance, or computer storage performance.
  • the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences includes:
  • the historical performance data sequence When training the preset neural network, the historical performance data sequence may be used as the input of the preset neural network, and the sequence type of the historical performance data sequence may be used as the label of the historical performance data sequence;
  • the labels of the output and historical performance data sequences are optimized through the back-propagation algorithm to optimize the parameters of the preset neural network.
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • the computer CPU performance sequence corresponds to the performance prediction model A, and the performance prediction model A can more accurately extract the characteristics of the computer CPU performance sequence (because the performance prediction model A is also trained according to the sequence type during the training process).
  • the computer performance data determining device 20 of the second embodiment predicts the computer performance based on the historical performance data of the computer.
  • the computer performance data determining device 20 further includes a returning module for returning a computer performance abnormality reminder if the computer performance data prediction result is not within the preset normal state range.
  • This embodiment provides a computer-readable storage medium.
  • the computer-readable storage medium may be volatile or non-volatile.
  • a computer program is stored on the computer-readable storage medium, and the computer program is processed. The following steps are implemented when the device is executed:
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • each module in the above-mentioned device embodiment is realized, for example, the modules 201-205 in FIG. 2.
  • FIG. 3 is a schematic diagram of a computer device provided in Embodiment 3 of this application.
  • the computer device 30 includes a memory 301, a processor 302, and a computer program 303 that is stored in the memory 301 and can run on the processor 302, such as a computer performance data determination program.
  • the processor 302 executes the computer program 303, the following steps are implemented:
  • the performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  • each module in the above-mentioned device embodiment is realized, for example, the modules 201-205 in FIG. 2.
  • the computer program 303 may be divided into one or more modules, and the one or more modules are stored in the memory 301 and executed by the processor 302 to complete the method.
  • the one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 303 in the computer device 30.
  • the computer program 303 may be divided into an acquisition module 201, a first training module 202, a classification module 203, a second training module 204, and a pre-judgment module 205 in FIG. 2. For specific functions of each module, refer to the second embodiment.
  • the schematic diagram 3 is only an example of the computer device 30, and does not constitute a limitation on the computer device 30. It may include more or less components than those shown in the figure, or combine certain components, or different components.
  • the computer device 30 may also include input and output devices, network access devices, buses, and so on.
  • the so-called processor 302 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor can be a microprocessor or the processor 302 can also be any conventional processor, etc.
  • the processor 302 is the control center of the computer device 30, which uses various interfaces and lines to connect the entire computer device 30. Various parts.
  • the memory 301 may be used to store the computer program 303, and the processor 302 implements the computer device by running or executing the computer program or module stored in the memory 301 and calling data stored in the memory 301 30 various functions.
  • the memory 301 may mainly include a storage program area and a storage data area.
  • the storage program area may store an operating system, an application program required by at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may Data and the like created in accordance with the use of the computer device 30 are stored.
  • the memory 301 may include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, a flash memory card (Flash Card), At least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.
  • non-volatile memory such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, a flash memory card (Flash Card), At least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.
  • the integrated module of the computer device 30 is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the computer program can be stored in a computer-readable storage medium.
  • the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) .
  • the computer usable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function, etc.; the storage data area may store a block chain node Use the created data, etc.
  • the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module.
  • the above-mentioned integrated modules can be implemented in the form of hardware, or in the form of hardware plus software functional modules.
  • the above-mentioned integrated modules implemented in the form of software functional modules may be stored in a computer readable storage medium.
  • the above-mentioned software function module is stored in a storage medium and includes several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor execute the method described in the various embodiments of this application. Part of the steps.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • Quality & Reliability (AREA)
  • Debugging And Monitoring (AREA)

Abstract

A computer performance data determining method and a related apparatus. The method comprises: obtaining a plurality of historical performance data sequences of a computer (101); pre-training a long short term memory network by using the plurality of historical performance data sequences (102); classifying the plurality of historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets having one-to-one correspondence to the N sequence types (103); training a performance pre-determination model consisting of a pre-trained long short term memory network and a full connection layer located behind the pre-trained long short term memory network by using each historical performance data sequence subset, so as to obtain N trained performance pre-determination models having one-to-one correspondence to the N sequence types; and pre-determining the performance data of the computer by means of the trained performance pre-determination model corresponding to the sequence type of a performance data sequence to be pre-determined of the computer.

Description

计算机性能数据确定方法、装置、计算机设备及存储介质Method and device for determining computer performance data, computer equipment and storage medium
本申请要求于2020年4月23日提交中国专利局、申请号为CN202010328656.1,发明名称为“计算机性能数据确定方法、装置、计算机设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。This application claims the priority of the Chinese patent application filed with the Chinese Patent Office on April 23, 2020, the application number is CN202010328656.1, and the invention title is "Computer performance data determination method, device, computer equipment and storage medium", all of which The content is incorporated in this application by reference.
技术领域Technical field
本申请涉及区块链技术,具体涉及一种计算机性能数据确定方法、装置、计算机设备及计算机可读存储介质。This application relates to blockchain technology, in particular to a method, device, computer equipment, and computer-readable storage medium for determining computer performance data.
背景技术Background technique
在云计算平台服务厂商,性能(如容量)预测是一个重要的AIOPS(基于人工智能的IT运营)应用场景。现有的性能预测方法对云计算平台服务厂商进行性能预测的精准度不高,对长期的性能预测效果不好。In cloud computing platform service providers, performance (such as capacity) prediction is an important application scenario of AIOPS (AI-based IT operations). The existing performance prediction methods do not have high accuracy for cloud computing platform service vendors to perform performance predictions, and are not effective in long-term performance predictions.
技术问题technical problem
发明人意识到当存在大规模性能指标需要进行预测时,如果对每个性能指标使用一个模型进行预测,同时不同的性能指标需要调节模型的参数大不想同,导致需要的人力投入成本巨大。另一方面,当某一个性能指标的训练数据较少时,对性能指标的预测可能出现过拟合。The inventor realized that when there are large-scale performance indicators that need to be predicted, if one model is used for each performance indicator, and the parameters of the model need to be adjusted for different performance indicators, the required labor input costs are huge. On the other hand, when the training data of a certain performance index is small, the prediction of the performance index may be over-fitting.
技术解决方案Technical solutions
一种计算机性能数据确定方法,所述方法包括:A method for determining computer performance data, the method comprising:
获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
一种计算机性能数据确定装置,所述装置包括:A device for determining computer performance data, the device comprising:
获取模块,用于获取计算机的多个历史性能数据序列;The acquisition module is used to acquire multiple historical performance data sequences of the computer;
第一训练模块,用于使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;The first training module is configured to use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain the pre-trained long and short-term memory network;
分类模块,用于对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;The classification module is configured to classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
第二训练模块,用于用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;The second training module is used to use each historical performance data sequence subset to train a performance prediction composed of the pre-trained long short-term memory network and the fully connected layer behind the pre-trained long short-term memory network. Judgment model to obtain N trained performance prediction models corresponding to the N sequence types one-to-one;
预判模块,用于通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The prediction module is used to predict the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
一种计算机设备,所述计算机设备包括处理器,所述处理器用于执行存储器中存储的计算机程序时实现如下步骤:A computer device includes a processor, and the processor implements the following steps when the processor is used to execute a computer program stored in a memory:
获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
本申请的第四方面提供一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被处理器执行时实现如下步骤:The fourth aspect of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model consisting of the pre-trained long short-term memory network and the fully connected layer behind the pre-trained long short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
有益效果Beneficial effect
本申请实现了根据计算机的历史性能数据预判计算机性能。This application realizes the prediction of computer performance based on the historical performance data of the computer.
附图说明Description of the drawings
图1是本申请实施例提供的计算机性能数据确定方法的流程图。Fig. 1 is a flowchart of a method for determining computer performance data provided by an embodiment of the present application.
图2是本申请实施例提供的计算机性能数据确定装置的结构图。Fig. 2 is a structural diagram of a computer performance data determining apparatus provided by an embodiment of the present application.
图3是本申请实施例提供的计算机设备的示意图。Fig. 3 is a schematic diagram of a computer device provided by an embodiment of the present application.
本发明的实施方式Embodiments of the present invention
为了能够更清楚地理解本申请的上述目的、特征和优点,下面结合附图和具体实施例对本申请进行详细描述。需要说明的是,在不冲突的情况下,本申请的实施例及实施例中的特征可以相互组合。In order to be able to understand the above objectives, features and advantages of the application more clearly, the application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the application and the features in the embodiments can be combined with each other if there is no conflict.
在下面的描述中阐述了很多具体细节以便于充分理解本申请,所描述的实施例仅仅是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。In the following description, many specific details are set forth in order to fully understand the present application. The described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.
除非另有定义,本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中在本申请的说明书中所使用的术语只是为了描述具体的实施例的目的,不是旨在于限制本申请。Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of this application. The terms used in the specification of the application herein are only for the purpose of describing specific embodiments, and are not intended to limit the application.
优选地,本申请的计算机性能数据确定方法应用在一个或者多个计算机设备中。所述计算机设备是一种能够按照事先设定或存储的指令,自动进行数值计算和/或信息处理的设备,其硬件包括但不限于微处理器、专用集成电路(Application Specific Integrated Circuit,ASIC)、可编程门阵列(Field-Programmable Gate Array,FPGA)、数字处理器(Digital Signal Processor,DSP)、嵌入式设备等。Preferably, the method for determining computer performance data of the present application is applied to one or more computer devices. The computer device is a device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions. Its hardware includes, but is not limited to, a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit). Specific Integrated Circuit, ASIC), Programmable Gate Array (Field-Programmable Gate Array, FPGA), Digital Signal Processor (DSP), embedded devices, etc.
所述计算机设备可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。所述计算机设备可以与用户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互。The computer device may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The computer device can interact with the user through a keyboard, a mouse, a remote control, a touch panel, or a voice control device.
实施例一Example one
图1是本申请实施例一提供的计算机性能数据确定方法的流程图。所述计算机性能数据确定方法应用于计算机设备,用于根据计算机的历史性能数据预判计算机性能数据,计算机性能数据可以包括计算机CPU数据、计算机GPU数据、计算机内存数据或计算机存储数据。FIG. 1 is a flowchart of a method for determining computer performance data provided in Embodiment 1 of the present application. The computer performance data determination method is applied to computer equipment to predict computer performance data based on historical computer performance data. The computer performance data may include computer CPU data, computer GPU data, computer memory data, or computer storage data.
如图1所示,所述计算机性能数据确定方法包括:As shown in Figure 1, the method for determining computer performance data includes:
101,获取计算机的多个历史性能数据序列。101. Acquire multiple historical performance data sequences of the computer.
在一具体实施例中,所述获取计算机的多个历史性能数据序列包括:In a specific embodiment, the acquiring multiple historical performance data sequences of the computer includes:
获取多个计算机的历史性能数据主序列,历史性能数据主序列中的每个元素为计算机性能的状态向量;Obtain the main sequence of historical performance data of multiple computers, and each element in the main sequence of historical performance data is a state vector of computer performance;
以预设长度h为滑窗长度,以1为步长,从每个历史性能数据主序列中截取多个历史性能数据序列。Taking the preset length h as the sliding window length and 1 as the step length, multiple historical performance data sequences are intercepted from each main sequence of historical performance data.
可以将每个历史性能数据序列所在的历史性能数据主序列中该历史性能数据序列的下一元素标注为该历史性能数据序列的标签。例如,一个计算机的历史性能数据主序列为{X1,X2,…,Xi,…,Xn},其中,Xi为时间点i的CPU的状态向量;以预设长度h为滑窗长度、以1为步长从每个历史性能数据主序列中截取多个历史性能数据序列为{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}等;历史性能数据序列{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}的标签分别为Xh+1、Xh+2、Xh+3。The next element of the historical performance data sequence in the main historical performance data sequence where each historical performance data sequence is located may be marked as the label of the historical performance data sequence. For example, the main sequence of historical performance data of a computer is {X1,X2,...,Xi,...,Xn}, where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length, and the Intercept multiple historical performance data sequences from each main sequence of historical performance data for the step size as {X1,X2,…Xh}, {X2,X3,… Xh+1}, {X3,X4,… Xh+2} Etc.; historical performance data sequence {X1,X2,...Xh}, {X2,X3,... The tags of Xh+1}, {X3, X4,... Xh+2} are Xh+1, Xh+2, Xh+3, respectively.
每个历史性能数据序列都是一个时间序列。Each historical performance data series is a time series.
102,使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络。102. Use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain a pre-trained long and short-term memory network.
具体地,可以初始化所述长短期记忆网络中的参数,将每个历史性能数据序列输入所述长短期记忆网络,计算所述长短期记忆网络根据输入值计算出的输出值与该历史性能数据序列的标签的距离,根据距离优化所述长短期记忆网络中的参数。Specifically, the parameters in the long and short-term memory network can be initialized, each historical performance data sequence is input into the long- and short-term memory network, and the output value calculated by the long and short-term memory network according to the input value and the historical performance data can be calculated The distance between the tags of the sequence is optimized according to the distance in the parameters in the long- and short-term memory network.
在另一实施例中,在使用所述多个历史性能数据序列对长短期记忆网络进行预训练之前,所述方法还包括:In another embodiment, before using the multiple historical performance data sequences to pre-train the long- and short-term memory network, the method further includes:
对所述多个历史性能数据序列进行预处理。Preprocessing the multiple historical performance data sequences.
可以对所述多个历史性能数据序列进行缺失值预处理、异常数据预处理等。对所述历史性能数据序列集进行缺失值预处理可以使所述多个历史性能序列中不存在缺失值,减小对训练长短期记忆网络的影响。对所述多个历史性能数据序列进行异常数据预处理,增加长短期记忆网络的训练收敛速度,减少训练长短期记忆网络的时间。当一个历史性能数据序列存在缺失值时(由于数据来源错误或软件设计错误等导致),可以采用插值法或替换法对该历史性能数据序列进行缺失值预处理。当一个历史性能数据序列存在异常数据时(即该历史性能数据序列中包括不在预设范围内的值),若异常数据大于预设范围的最大值,则将异常数据设置为预设范围的最大值;若异常数据小于预设范围的最小值,则将异常数据设置为预设范围的最小值。Missing value preprocessing, abnormal data preprocessing, etc. may be performed on the multiple historical performance data sequences. Performing missing value preprocessing on the historical performance data sequence set can prevent missing values in the multiple historical performance sequences and reduce the impact on training the long and short-term memory network. Performing abnormal data preprocessing on the multiple historical performance data sequences increases the training convergence speed of the long- and short-term memory network, and reduces the time for training the long- and short-term memory network. When there are missing values in a historical performance data sequence (due to data source errors or software design errors, etc.), the historical performance data sequence can be preprocessed with missing values using interpolation or replacement methods. When there is abnormal data in a historical performance data sequence (that is, the historical performance data sequence includes values that are not within the preset range), if the abnormal data is greater than the maximum value of the preset range, the abnormal data is set to the maximum value of the preset range Value; if the abnormal data is less than the minimum value of the preset range, the abnormal data is set to the minimum value of the preset range.
103,对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集。103. Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types in a one-to-one manner.
在一具体实施例中,所述对所述多个历史性能数据序列进行分类包括:In a specific embodiment, the classifying the multiple historical performance data sequences includes:
对所述多个历史性能数据序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical performance data sequences, and modifying the clustering result according to the received modification instruction; or
用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类。Classify the multiple historical performance data sequences by using the trained preset convolutional neural network model.
可以根据DBScan聚类算法对所述历史性能数据序列集中的多个历史性能数据序列进行聚类,其中,DBScan聚类算法中的距离函数使用一维NCC(normalized cross correlation)算法;当聚类结果出现异常时,根据接收的删除指令删除聚类结果中距离聚类中心超出预设范围的历史性能数据序列。The multiple historical performance data sequences in the historical performance data sequence set may be clustered according to the DBScan clustering algorithm, where the distance function in the DBScan clustering algorithm uses one-dimensional NCC (normalized cross correlation) algorithm; when the clustering result is abnormal, delete the historical performance data sequence in the clustering result that is beyond the preset range from the cluster center according to the received delete instruction.
具体地,所述对所述多个历史性能数据序列进行聚类包括:Specifically, the clustering of the multiple historical performance data sequences includes:
从所述多个历史性能数据序列中选择N个中心点序列;Selecting N central point sequences from the multiple historical performance data sequences;
计算每个历史性能数据序列与所述N个中心点序列的距离,得到与该历史性能数据序列距离最近的中心点序列;Calculate the distance between each historical performance data sequence and the N central point sequences, and obtain the central point sequence closest to the historical performance data sequence;
将该历史性能数据序列分划至,与该历史性能数据序列距离最近的中心点序列所属的簇。The historical performance data sequence is divided into the cluster to which the center point sequence closest to the historical performance data sequence belongs.
具体地,所述对所述多个历史性能数据序列进行聚类包括:Specifically, the clustering of the multiple historical performance data sequences includes:
以每个历史性能数据序列为中心确定一个簇,得到W个簇;Determine a cluster with each historical performance data sequence as the center, and obtain W clusters;
若W大于N,计算每两个簇间的距离,合并距离最小的两个簇,得到W-1个簇;If W is greater than N, calculate the distance between every two clusters and merge the two clusters with the smallest distance to obtain W-1 clusters;
若W-1大于N,计算每两个簇间的距离,合并距离最小的两个簇,得到W-2个簇;依次类推,当W-R等于N时,得到N个簇。If W-1 is greater than N, calculate the distance between every two clusters, merge the two clusters with the smallest distance, and get W-2 clusters; and so on, when W-R is equal to N, get N clusters.
在所述用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类之前,训练预设卷积神经网络模型:Before classifying the multiple historical performance data sequences with the trained preset convolutional neural network model, training the preset convolutional neural network model:
将一个历史性能数据序列输入预设卷积神经网络模型,每个历史性能数据序列对应一个类型标签;Input a historical performance data sequence into the preset convolutional neural network model, and each historical performance data sequence corresponds to a type label;
基于预设卷积神经网络模型的输出和该历史性能数据序列的类型标签,对预设卷积神经网络模型进行反向传播,优化预设卷积神经网络模型的参数。Based on the output of the preset convolutional neural network model and the type label of the historical performance data sequence, the preset convolutional neural network model is backpropagated, and the parameters of the preset convolutional neural network model are optimized.
104,用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型。104. Use each subset of historical performance data sequence to train a performance prediction model consisting of the pre-trained long-term and short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network, to obtain and The N training performance prediction models corresponding to the N sequence types one-to-one.
在一具体实施例中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:In a specific embodiment, the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long and short-term memory network. Performance prediction models include:
判断该历史性能数据序列的数量是否大于预设阈值;Determine whether the number of historical performance data sequences is greater than a preset threshold;
若该历史性能数据序列的数量大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数;If the number of historical performance data sequences is greater than the preset threshold, use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function;
若该历史性能数据序列的数量不大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数。If the number of historical performance data sequences is not greater than the preset threshold, the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
当该历史性能数据序列的数量不大于预设阈值时,只优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数可以减少出现过拟合的情况,且由于所述性能预判模型经过预训练,不会导致预测准确率低的情况发生。When the number of historical performance data sequences is not greater than the preset threshold, only optimizing the parameters of the fully connected layer of the performance prediction model corresponding to the subset of historical performance data sequences can reduce the occurrence of overfitting, and due to the performance The prediction model is pre-trained and will not lead to low prediction accuracy.
在另一实施例中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:In another embodiment, the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network Performance prediction models include:
根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数。Use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function.
105,通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。105. Prejudge the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
在所述通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据之前,所述方法还包括:Before predicting the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, the method further includes:
(1)获取所述计算机的待预判性能数据序列。(1) Obtain the to-be-judged performance data sequence of the computer.
例如,计算机CPU性能序列为{X1,X2,… X15},其中,X1=(2019,06,12,13,15,43),前5维为获取时间,第6维为所述计算机在获取时间的CPU使用率;X2=(2019,06,12,13,16,80);X15=(2019,06,12,13,30,90)。类似地,所述计算机的待预判性能数据序列还包括计算机GPU性能、计算机内存性能或计算机存储性能。For example, the computer’s CPU performance sequence is {X1, X2,... X15}, where X1=(2019, 06, 12, 13, 15, 43), the first 5 dimensions are the acquisition time, and the 6th dimension is the computer’s acquisition time CPU usage by time; X2=(2019,06,12,13,16,80); X15=(2019,06,12,13,30,90). Similarly, the to-be-predicted performance data sequence of the computer also includes computer GPU performance, computer memory performance, or computer storage performance.
(2)根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型。(2) Determine the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences.
在一具体实施例中,所述根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型包括:In a specific embodiment, the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences includes:
计算N个历史性能数据序列子集的N个中心序列,将距离所述待预判性能数据序列最近的中心序列对应的序列类型确定为所述目标序列类型;或Calculate N central sequences of N historical performance data sequence subsets, and determine the sequence type corresponding to the central sequence closest to the performance data sequence to be predicted as the target sequence type; or
用N个历史性能数据序列子集训练预设神经网络,其中每个历史性能数据序列子集中的每个历史性能数据序列的标签为该历史性能数据序列的序列类型,将所述待预判性能数据序列输入训练后的所述预设神经网络,根据训练后的所述预设神经网络的输出确定所述待预判性能数据序列的序列类型。Use N historical performance data sequence subsets to train the preset neural network, where the label of each historical performance data sequence in each historical performance data sequence subset is the sequence type of the historical performance data sequence, and the performance to be predicted The data sequence is input to the preset neural network after training, and the sequence type of the performance data sequence to be predicted is determined according to the output of the preset neural network after training.
在训练所述预设神经网络时,可以将历史性能数据序列作为所述预设神经网络的输入,将历史性能数据序列的序列类型作为历史性能数据序列的标签;根据所述预设神经网络的输出和历史性能数据序列的标签通过反向传播算法优化所述预设神经网络的参数。When training the preset neural network, the historical performance data sequence may be used as the input of the preset neural network, and the sequence type of the historical performance data sequence may be used as the label of the historical performance data sequence; The labels of the output and historical performance data sequences are optimized through the back-propagation algorithm to optimize the parameters of the preset neural network.
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据,例如,计算机CPU性能序列为{X1,X2,… X15},其中,X1=(2019,06,12,13,15,43),前5维为获取时间,第6维为所述计算机在获取时间的CPU使用率;X2=(2019,06,12,13,16,80);X15=(2019,06,12,13,30,90)。计算机CPU性能序列对应性能预判模型A,性能预判模型A更能准确地提取计算机CPU性能序列的特征(因为训练过程中也根据序列类型对性能预判模型A进行训练)。将计算机CPU性能序列输入性能预判模型A,通过性能预测模型A预判所述计算机的CPU性能,得到X16=(2019,06,12,13,31,95)。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer. For example, the computer CPU performance sequence is {X1, X2,... X15}, Among them, X1=(2019,06,12,13,15,43), the first 5 dimensions are the acquisition time, and the 6th dimension is the CPU usage rate of the computer at the acquisition time; X2=(2019,06,12,13 , 16, 80); X15 = (2019, 06, 12, 13, 30, 90). The computer CPU performance sequence corresponds to the performance prediction model A, and the performance prediction model A can more accurately extract the characteristics of the computer CPU performance sequence (because the performance prediction model A is also trained according to the sequence type during the training process). The computer CPU performance sequence is input into the performance prediction model A, and the CPU performance of the computer is predicted through the performance prediction model A to obtain X16=(2019, 06, 12, 13, 31, 95).
实施例一的计算机性能数据确定方法获取计算机的历史性能数据序列集,所述历史性能数据序列集包括多个历史性能数据序列和每个历史性能数据序列的标签;使用所述历史性能数据序列集对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个性能预判模型;获取所述计算机的待预判性能数据序列;确定所述待预判性能数据序列的序列类型;通过与所述待预判性能数据序列的序列类型对应的性能预判模型预判所述计算机的性能数据。实施例一根据计算机的历史性能数据预判计算机性能。The computer performance data determination method of the first embodiment obtains a computer's historical performance data sequence set, the historical performance data sequence set includes a plurality of historical performance data sequences and a label of each historical performance data sequence; the historical performance data sequence set is used Perform pre-training on the long and short-term memory network to obtain the pre-trained long and short-term memory network; classify the multiple historical performance data sequences to obtain N sequence types and N sequence types corresponding to the N sequence types one-to-one Historical performance data sequence subsets; each historical performance data sequence subset is used to train a performance prediction composed of the pre-trained long-short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network To obtain the N performance prediction models corresponding to the N sequence types one-to-one; obtain the performance data sequence to be predicted of the computer; determine the sequence type of the performance data sequence to be predicted; The performance prediction model corresponding to the sequence type of the performance data sequence to be predicted predicts the performance data of the computer. The first embodiment predicts the computer performance based on the historical performance data of the computer.
需要强调的是,为进一步保证上述性能数据的私密和安全性,上述性能数据还可以存储于一区块链的节点中。It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned performance data, the above-mentioned performance data may also be stored in a node of a blockchain.
在另一实施例中,所述方法还包括:In another embodiment, the method further includes:
若所述计算机的性能数据预测结果不在预设正常状态范围,返回计算机性能异常提醒。在异常发生时,停止运行所述计算机中的新任务。If the prediction result of the computer's performance data is not within the preset normal state range, a notification of abnormal computer performance is returned. When an abnormality occurs, stop running the new task in the computer.
例如,所述计算机的性能数据预测结果为X16=(2019,06,12,13,31,95),95为预测的所述计算机的CPU使用率,预设正常状态范围为0-90,返回计算机CPU异常提醒给用户;并停止运行所述计算机中的新任务。For example, the performance data prediction result of the computer is X16=(2019, 06, 12, 13, 31, 95), 95 is the predicted CPU usage rate of the computer, the preset normal state range is 0-90, return The computer CPU is abnormally reminded to the user; and the new task in the computer is stopped running.
在另一实施例中,所述方法还包括:In another embodiment, the method further includes:
(1)获取计算机的多个历史CPU使用率序列。(1) Obtain multiple historical CPU usage sequences of the computer.
例如,一个计算机的历史CPU使用率主序列为{X1,X2,…,Xi,…,Xn},其中,Xi为时间点i的CPU的状态向量;以预设长度h为滑窗长度、以1为步长从每个历史CPU使用率主序列中截取多个历史CPU使用率序列为{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}等;历史CPU使用率序列{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}的标签分别为Xh+1、Xh+2、Xh+3。For example, the main sequence of historical CPU usage of a computer is {X1,X2,...,Xi,...,Xn}, where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length and the 1 is the step size to intercept multiple historical CPU usage sequence from each main sequence of historical CPU usage as {X1,X2,...Xh}, {X2,X3,... Xh+1}, {X3,X4,... Xh +2} etc.; historical CPU usage rate sequence {X1,X2,...Xh}, {X2,X3,... The tags of Xh+1}, {X3, X4,... Xh+2} are Xh+1, Xh+2, Xh+3, respectively.
每个历史CPU使用率序列都是一个时间序列。Each historical CPU usage sequence is a time sequence.
(2)使用所述多个历史CPU使用率序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络。(2) Pre-training the long and short-term memory network using the multiple historical CPU usage rate sequences to obtain a pre-trained long and short-term memory network.
具体地,可以初始化所述长短期记忆网络中的参数,将每个历史CPU使用率序列输入所述长短期记忆网络,计算所述长短期记忆网络根据输入值计算出的输出值与该历史CPU使用率序列的标签的距离,根据距离优化所述长短期记忆网络中的参数。Specifically, the parameters in the long and short-term memory network may be initialized, each historical CPU usage rate sequence is input into the long-term short-term memory network, and the output value calculated by the long-term short-term memory network according to the input value and the historical CPU Use the distance of the tags of the rate sequence to optimize the parameters in the long and short-term memory network according to the distance.
(3)对所述多个历史CPU使用率序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史CPU使用率序列子集。(3) Classify the multiple historical CPU usage rate sequences to obtain N sequence types and N historical CPU usage rate sequence subsets corresponding to the N sequence types one-to-one.
在一具体实施例中,所述对所述多个历史CPU使用率序列进行分类包括:In a specific embodiment, the classifying the multiple historical CPU usage rate sequences includes:
对所述多个历史CPU使用率序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical CPU usage rate sequences, and modifying the clustering result according to the received modification instruction; or
用训练好的预设卷积神经网络模型对所述多个历史CPU使用率序列进行分类。Use the trained preset convolutional neural network model to classify the multiple historical CPU usage rate sequences.
(4)用每个历史CPU使用率序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的CPU使用率预判模型,得到与所述N个序列类型一一对应的N个训练后的CPU使用率预判模型。(4) Use each historical CPU usage sequence subset to train a CPU usage prediction composed of the pre-trained long and short-term memory network and the fully connected layer after the pre-trained long and short-term memory network Model to obtain N trained CPU usage prediction models corresponding to the N sequence types one-to-one.
(5)通过与所述计算机的待预判CPU使用率序列的序列类型对应的训练后的CPU使用率预判模型,预判所述计算机的CPU性能。(5) Prejudge the CPU performance of the computer through the trained CPU usage prediction model corresponding to the sequence type of the CPU usage sequence to be predicted for the computer.
获取所述计算机的待预判CPU使用率序列。Obtain the CPU usage rate sequence to be predicted of the computer.
例如,计算机CPU性能序列为{X1,X2,… X15},其中,X1=(2019,06,12,13,15,43),前5维为获取时间,第6维为所述计算机在获取时间的CPU使用率;X2=(2019,06,12,13,16,80);X15=(2019,06,12,13,30,90)。For example, the computer’s CPU performance sequence is {X1, X2,... X15}, where X1=(2019, 06, 12, 13, 15, 43), the first 5 dimensions are the acquisition time, and the 6th dimension is the computer’s acquisition time CPU usage by time; X2=(2019,06,12,13,16,80); X15=(2019,06,12,13,30,90).
根据N个历史CPU使用率序列子集确定所述待预判CPU使用率序列的序列类型。The sequence type of the CPU usage rate sequence to be predicted is determined according to the N historical CPU usage rate sequence subsets.
计算机CPU性能序列对应CPU使用率预判模型A。将计算机CPU性能序列输入CPU使用率预判模型A,通过CPU使用率预判模型A预判所述计算机的CPU性能,得到X16=(2019,06,12,13,31,95)。The computer CPU performance sequence corresponds to the prediction model A of the CPU usage rate. The computer CPU performance sequence is input into the CPU usage prediction model A, and the CPU performance of the computer is predicted through the CPU usage prediction model A to obtain X16=(2019, 06, 12, 13, 31, 95).
实施例二Example two
图2是本申请实施例二提供的计算机性能数据确定装置的结构图。所述计算机性能数据确定装置20应用于计算机设备。所述计算机性能数据确定装置20用于根据计算机的历史性能数据预判计算机性能数据,计算机性能数据可以包括计算机CPU数据、计算机GPU数据、计算机内存数据或计算机存储数据。Fig. 2 is a structural diagram of a computer performance data determining device provided in the second embodiment of the present application. The computer performance data determining device 20 is applied to computer equipment. The computer performance data determining device 20 is used to predict the computer performance data based on the historical performance data of the computer. The computer performance data may include computer CPU data, computer GPU data, computer memory data, or computer storage data.
如图2所示,所述计算机性能数据确定装置20可以包括获取模块201、第一训练模块202、分类模块203、第二训练模块204、预判模块205。As shown in FIG. 2, the device 20 for determining computer performance data may include an acquisition module 201, a first training module 202, a classification module 203, a second training module 204, and a pre-judgment module 205.
获取模块201,用于获取计算机的多个历史性能数据序列。The obtaining module 201 is used to obtain multiple historical performance data sequences of the computer.
在一具体实施例中,所述获取计算机的多个历史性能数据序列包括:In a specific embodiment, the acquiring multiple historical performance data sequences of the computer includes:
获取多个计算机的历史性能数据主序列,历史性能数据主序列中的每个元素为计算机性能的状态向量;Obtain the main sequence of historical performance data of multiple computers, and each element in the main sequence of historical performance data is a state vector of computer performance;
以预设长度h为滑窗长度,以1为步长,从每个历史性能数据主序列中截取多个历史性能数据序列。Taking the preset length h as the sliding window length and 1 as the step length, multiple historical performance data sequences are intercepted from each main sequence of historical performance data.
可以将每个历史性能数据序列所在的历史性能数据主序列中该历史性能数据序列的下一元素标注为该历史性能数据序列的标签。例如,一个计算机的历史性能数据主序列为{X1,X2,…,Xi,…,Xn},其中,Xi为时间点i的CPU的状态向量;以预设长度h为滑窗长度、以1为步长从每个历史性能数据主序列中截取多个历史性能数据序列为{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}等;历史性能数据序列{X1,X2,…Xh}、{X2,X3,… Xh+1}、{ X3,X4,… Xh+2}的标签分别为Xh+1、Xh+2、Xh+3。The next element of the historical performance data sequence in the main historical performance data sequence where each historical performance data sequence is located may be marked as the label of the historical performance data sequence. For example, the main sequence of historical performance data of a computer is {X1,X2,...,Xi,...,Xn}, where Xi is the state vector of the CPU at time i; the preset length h is the sliding window length, and 1 Intercept multiple historical performance data sequences from each main sequence of historical performance data for the step size as {X1,X2,…Xh}, {X2,X3,… Xh+1}, {X3,X4,… Xh+2} Etc.; historical performance data sequence {X1,X2,...Xh}, {X2,X3,... The tags of Xh+1}, {X3, X4,... Xh+2} are Xh+1, Xh+2, Xh+3, respectively.
每个历史性能数据序列都是一个时间序列。Each historical performance data series is a time series.
第一训练模块202,用于使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络。The first training module 202 is configured to use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain a pre-trained long- and short-term memory network.
具体地,可以初始化所述长短期记忆网络中的参数,将每个历史性能数据序列输入所述长短期记忆网络,计算所述长短期记忆网络根据输入值计算出的输出值与该历史性能数据序列的标签的距离,根据距离优化所述长短期记忆网络中的参数。Specifically, the parameters in the long and short-term memory network can be initialized, each historical performance data sequence is input into the long- and short-term memory network, and the output value calculated by the long and short-term memory network according to the input value and the historical performance data can be calculated The distance between the tags of the sequence is optimized according to the distance in the parameters in the long- and short-term memory network.
在另一实施例中,在使用所述多个历史性能数据序列对长短期记忆网络进行预训练之前,所述方法还包括:In another embodiment, before using the multiple historical performance data sequences to pre-train the long- and short-term memory network, the method further includes:
对所述多个历史性能数据序列进行预处理。Preprocessing the multiple historical performance data sequences.
可以对所述多个历史性能数据序列进行缺失值预处理、异常数据预处理等。对所述历史性能数据序列集进行缺失值预处理可以使所述多个历史性能序列中不存在缺失值,减小对训练长短期记忆网络的影响。对所述多个历史性能数据序列进行异常数据预处理,增加长短期记忆网络的训练收敛速度,减少训练长短期记忆网络的时间。当一个历史性能数据序列存在缺失值时(由于数据来源错误或软件设计错误等导致),可以采用插值法或替换法对该历史性能数据序列进行缺失值预处理。当一个历史性能数据序列存在异常数据时(即该历史性能数据序列中包括不在预设范围内的值),若异常数据大于预设范围的最大值,则将异常数据设置为预设范围的最大值;若异常数据小于预设范围的最小值,则将异常数据设置为预设范围的最小值。Missing value preprocessing, abnormal data preprocessing, etc. may be performed on the multiple historical performance data sequences. Performing missing value preprocessing on the historical performance data sequence set can prevent missing values in the multiple historical performance sequences and reduce the impact on training the long and short-term memory network. Performing abnormal data preprocessing on the multiple historical performance data sequences increases the training convergence speed of the long- and short-term memory network, and reduces the time for training the long- and short-term memory network. When there are missing values in a historical performance data sequence (due to data source errors or software design errors, etc.), interpolation or replacement methods can be used to preprocess the missing values of the historical performance data sequence. When there is abnormal data in a historical performance data sequence (that is, the historical performance data sequence includes values that are not within the preset range), if the abnormal data is greater than the maximum value of the preset range, the abnormal data is set to the maximum value of the preset range Value; if the abnormal data is less than the minimum value of the preset range, the abnormal data is set to the minimum value of the preset range.
分类模块203,用于对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集。The classification module 203 is configured to classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one.
在一具体实施例中,所述对所述多个历史性能数据序列进行分类包括:In a specific embodiment, the classifying the multiple historical performance data sequences includes:
对所述多个历史性能数据序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical performance data sequences, and modifying the clustering result according to the received modification instruction; or
用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类。Classify the multiple historical performance data sequences by using the trained preset convolutional neural network model.
可以根据DBScan聚类算法对所述历史性能数据序列集中的多个历史性能数据序列进行聚类,其中,DBScan聚类算法中的距离函数使用一维NCC(normalized cross correlation)算法;当聚类结果出现异常时,根据接收的删除指令删除聚类结果中距离聚类中心超出预设范围的历史性能数据序列。The multiple historical performance data sequences in the historical performance data sequence set may be clustered according to the DBScan clustering algorithm, where the distance function in the DBScan clustering algorithm uses one-dimensional NCC (normalized cross correlation) algorithm; when the clustering result is abnormal, delete the historical performance data sequence in the clustering result that is beyond the preset range from the cluster center according to the received delete instruction.
具体地,所述对所述多个历史性能数据序列进行聚类包括:Specifically, the clustering of the multiple historical performance data sequences includes:
从所述多个历史性能数据序列中选择N个中心点序列;Selecting N central point sequences from the multiple historical performance data sequences;
计算每个历史性能数据序列与所述N个中心点序列的距离,得到与该历史性能数据序列距离最近的中心点序列;Calculate the distance between each historical performance data sequence and the N central point sequences, and obtain the central point sequence closest to the historical performance data sequence;
将该历史性能数据序列分划至,与该历史性能数据序列距离最近的中心点序列所属的簇。The historical performance data sequence is divided into the cluster to which the center point sequence closest to the historical performance data sequence belongs.
具体地,所述对所述多个历史性能数据序列进行聚类包括:Specifically, the clustering of the multiple historical performance data sequences includes:
以每个历史性能数据序列为中心确定一个簇,得到W个簇;Determine a cluster with each historical performance data sequence as the center, and obtain W clusters;
若W大于N,计算每两个簇间的距离,合并距离最小的两个簇,得到W-1个簇;If W is greater than N, calculate the distance between every two clusters and merge the two clusters with the smallest distance to obtain W-1 clusters;
若W-1大于N,计算每两个簇间的距离,合并距离最小的两个簇,得到W-2个簇;依次类推,当W-R等于N时,得到N个簇。If W-1 is greater than N, calculate the distance between every two clusters, merge the two clusters with the smallest distance, and get W-2 clusters; and so on, when W-R is equal to N, get N clusters.
在所述用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类之前,训练预设卷积神经网络模型:Before classifying the multiple historical performance data sequences with the trained preset convolutional neural network model, training the preset convolutional neural network model:
将一个历史性能数据序列输入预设卷积神经网络模型,每个历史性能数据序列对应一个类型标签;Input a historical performance data sequence into the preset convolutional neural network model, and each historical performance data sequence corresponds to a type label;
基于预设卷积神经网络模型的输出和该历史性能数据序列的类型标签,对预设卷积神经网络模型进行反向传播,优化预设卷积神经网络模型的参数。Based on the output of the preset convolutional neural network model and the type label of the historical performance data sequence, the preset convolutional neural network model is backpropagated, and the parameters of the preset convolutional neural network model are optimized.
第二训练模块204,用于用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型。The second training module 204 is configured to use each subset of historical performance data sequences to train a performance consisting of the pre-trained long-short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network The predictive model obtains N trained performance predictive models corresponding to the N sequence types in a one-to-one manner.
在一具体实施例中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:In a specific embodiment, the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long and short-term memory network. Performance prediction models include:
判断该历史性能数据序列的数量是否大于预设阈值;Determine whether the number of historical performance data sequences is greater than a preset threshold;
若该历史性能数据序列的数量大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数;If the number of historical performance data sequences is greater than the preset threshold, use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function;
若该历史性能数据序列的数量不大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数。If the number of historical performance data sequences is not greater than the preset threshold, the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
当该历史性能数据序列的数量不大于预设阈值时,只优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数可以减少出现过拟合的情况,且由于所述性能预判模型经过预训练,不会导致预测准确率低的情况发生。When the number of historical performance data sequences is not greater than the preset threshold, only optimizing the parameters of the fully connected layer of the performance prediction model corresponding to the subset of historical performance data sequences can reduce the occurrence of overfitting, and due to the performance The prediction model is pre-trained and will not lead to low prediction accuracy.
在另一实施例中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:In another embodiment, the training of each historical performance data sequence subset is composed of the pre-trained long and short-term memory network and the fully connected layer behind the pre-trained long- and short-term memory network Performance prediction models include:
根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数。Use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function.
预判模块205,用于通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The predicting module 205 is configured to predict the performance data of the computer through the trained performance predicting model corresponding to the sequence type of the to-be-predicted performance data sequence of the computer.
在所述通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据之前,所述方法还包括:Before predicting the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, the method further includes:
(1)获取所述计算机的待预判性能数据序列。(1) Obtain the to-be-judged performance data sequence of the computer.
例如,计算机CPU性能序列为{X1,X2,… X15},其中,X1=(2019,06,12,13,15,43),前5维为获取时间,第6维为所述计算机在获取时间的CPU使用率;X2=(2019,06,12,13,16,80);X15=(2019,06,12,13,30,90)。类似地,所述计算机的待预判性能数据序列还包括计算机GPU性能、计算机内存性能或计算机存储性能。For example, the computer’s CPU performance sequence is {X1, X2,... X15}, where X1=(2019, 06, 12, 13, 15, 43), the first 5 dimensions are the acquisition time, and the 6th dimension is the computer’s acquisition time CPU usage by time; X2=(2019,06,12,13,16,80); X15=(2019,06,12,13,30,90). Similarly, the to-be-predicted performance data sequence of the computer also includes computer GPU performance, computer memory performance, or computer storage performance.
(2)根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型。(2) Determine the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences.
在一具体实施例中,所述根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型包括:In a specific embodiment, the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences includes:
计算N个历史性能数据序列子集的N个中心序列,将距离所述待预判性能数据序列最近的中心序列对应的序列类型确定为所述目标序列类型;或Calculate N central sequences of N historical performance data sequence subsets, and determine the sequence type corresponding to the central sequence closest to the performance data sequence to be predicted as the target sequence type; or
用N个历史性能数据序列子集训练预设神经网络,其中每个历史性能数据序列子集中的每个历史性能数据序列的标签为该历史性能数据序列的序列类型,将所述待预判性能数据序列输入训练后的所述预设神经网络,根据训练后的所述预设神经网络的输出确定所述待预判性能数据序列的序列类型。Use N historical performance data sequence subsets to train the preset neural network, where the label of each historical performance data sequence in each historical performance data sequence subset is the sequence type of the historical performance data sequence, and the performance to be predicted The data sequence is input to the preset neural network after training, and the sequence type of the performance data sequence to be predicted is determined according to the output of the preset neural network after training.
在训练所述预设神经网络时,可以将历史性能数据序列作为所述预设神经网络的输入,将历史性能数据序列的序列类型作为历史性能数据序列的标签;根据所述预设神经网络的输出和历史性能数据序列的标签通过反向传播算法优化所述预设神经网络的参数。When training the preset neural network, the historical performance data sequence may be used as the input of the preset neural network, and the sequence type of the historical performance data sequence may be used as the label of the historical performance data sequence; The labels of the output and historical performance data sequences are optimized through the back-propagation algorithm to optimize the parameters of the preset neural network.
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据,例如,计算机CPU性能序列为{X1,X2,… X15},其中,X1=(2019,06,12,13,15,43),前5维为获取时间,第6维为所述计算机在获取时间的CPU使用率;X2=(2019,06,12,13,16,80);X15=(2019,06,12,13,30,90)。计算机CPU性能序列对应性能预判模型A,性能预判模型A更能准确地提取计算机CPU性能序列的特征(因为训练过程中也根据序列类型对性能预判模型A进行训练)。将计算机CPU性能序列输入性能预判模型A,通过性能预测模型A预判所述计算机的CPU性能,得到X16=(2019,06,12,13,31,95)。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer. For example, the computer CPU performance sequence is {X1, X2,... X15}, Among them, X1=(2019,06,12,13,15,43), the first 5 dimensions are the acquisition time, and the 6th dimension is the CPU usage rate of the computer at the acquisition time; X2=(2019,06,12,13 , 16, 80); X15 = (2019, 06, 12, 13, 30, 90). The computer CPU performance sequence corresponds to the performance prediction model A, and the performance prediction model A can more accurately extract the characteristics of the computer CPU performance sequence (because the performance prediction model A is also trained according to the sequence type during the training process). The computer CPU performance sequence is input into the performance prediction model A, and the CPU performance of the computer is predicted through the performance prediction model A to obtain X16=(2019, 06, 12, 13, 31, 95).
实施例二的计算机性能数据确定装置20根据计算机的历史性能数据预判计算机性能。The computer performance data determining device 20 of the second embodiment predicts the computer performance based on the historical performance data of the computer.
在另一实施例中,所述计算机性能数据确定装置20还包括返回模块,用于若所述计算机的性能数据预测结果不在预设正常状态范围,返回计算机性能异常提醒。In another embodiment, the computer performance data determining device 20 further includes a returning module for returning a computer performance abnormality reminder if the computer performance data prediction result is not within the preset normal state range.
在异常发生时,停止运行所述计算机中的新任务。When an abnormality occurs, stop running the new task in the computer.
例如,所述计算机的性能数据预测结果为X16=(2019,06,12,13,31,95),95为预测的所述计算机的CPU使用率,预设正常状态范围为0-90,返回计算机CPU异常提醒给用户;并停止运行所述计算机中的新任务。For example, the performance data prediction result of the computer is X16=(2019, 06, 12, 13, 31, 95), 95 is the predicted CPU usage rate of the computer, the preset normal state range is 0-90, return The computer CPU is abnormally reminded to the user; and the new task in the computer is stopped running.
实施例三Example three
本实施例提供一种计算机可读存储介质,该计算机可读存储介质可以是易失性的,也可以是非易失性的,该计算机可读存储介质上存储有计算机程序,该计算机程序被处理器执行时实现如下步骤:This embodiment provides a computer-readable storage medium. The computer-readable storage medium may be volatile or non-volatile. A computer program is stored on the computer-readable storage medium, and the computer program is processed. The following steps are implemented when the device is executed:
获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
或者,该计算机程序被处理器执行时实现上述装置实施例中各模块的功能,例如图2中的模块201-205。Or, when the computer program is executed by the processor, the function of each module in the above-mentioned device embodiment is realized, for example, the modules 201-205 in FIG. 2.
实施例四Example four
图3为本申请实施例三提供的计算机设备的示意图。所述计算机设备30包括存储器301、处理器302以及存储在所述存储器301中并可在所述处理器302上运行的计算机程序303,例如计算机性能数据确定程序。所述处理器302执行所述计算机程序303时实现如下步骤:FIG. 3 is a schematic diagram of a computer device provided in Embodiment 3 of this application. The computer device 30 includes a memory 301, a processor 302, and a computer program 303 that is stored in the memory 301 and can run on the processor 302, such as a computer performance data determination program. When the processor 302 executes the computer program 303, the following steps are implemented:
获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
或者,该计算机程序被处理器执行时实现上述装置实施例中各模块的功能,例如图2中的模块201-205。Or, when the computer program is executed by the processor, the function of each module in the above-mentioned device embodiment is realized, for example, the modules 201-205 in FIG. 2.
示例性的,所述计算机程序303可以被分割成一个或多个模块,所述一个或者多个模块被存储在所述存储器301中,并由所述处理器302执行,以完成本方法。所述一个或多个模块可以是能够完成特定功能的一系列计算机程序指令段,该指令段用于描述所述计算机程序303在所述计算机设备30中的执行过程。例如,所述计算机程序303可以被分割成图2中的获取模块201、第一训练模块202、分类模块203、第二训练模块204、预判模块205,各模块具体功能参见实施例二。Exemplarily, the computer program 303 may be divided into one or more modules, and the one or more modules are stored in the memory 301 and executed by the processor 302 to complete the method. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 303 in the computer device 30. For example, the computer program 303 may be divided into an acquisition module 201, a first training module 202, a classification module 203, a second training module 204, and a pre-judgment module 205 in FIG. 2. For specific functions of each module, refer to the second embodiment.
本领域技术人员可以理解,所述示意图3仅仅是计算机设备30的示例,并不构成对计算机设备30的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件,例如所述计算机设备30还可以包括输入输出设备、网络接入设备、总线等。Those skilled in the art can understand that the schematic diagram 3 is only an example of the computer device 30, and does not constitute a limitation on the computer device 30. It may include more or less components than those shown in the figure, or combine certain components, or different components. For example, the computer device 30 may also include input and output devices, network access devices, buses, and so on.
所称处理器302可以是中央处理单元(Central Processing Unit,CPU),还可以是其他通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器302也可以是任何常规的处理器等,所述处理器302是所述计算机设备30的控制中心,利用各种接口和线路连接整个计算机设备30的各个部分。The so-called processor 302 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor 302 can also be any conventional processor, etc. The processor 302 is the control center of the computer device 30, which uses various interfaces and lines to connect the entire computer device 30. Various parts.
所述存储器301可用于存储所述计算机程序303,所述处理器302通过运行或执行存储在所述存储器301内的计算机程序或模块,以及调用存储在存储器301内的数据,实现所述计算机设备30的各种功能。所述存储器301可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等;存储数据区可存储根据计算机设备30的使用所创建的数据等。此外,存储器301可以包括非易失性存储器,例如硬盘、内存、插接式硬盘,智能存储卡(Smart Media Card, SMC),安全数字(Secure Digital, SD)卡,闪存卡(Flash Card)、至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。The memory 301 may be used to store the computer program 303, and the processor 302 implements the computer device by running or executing the computer program or module stored in the memory 301 and calling data stored in the memory 301 30 various functions. The memory 301 may mainly include a storage program area and a storage data area. The storage program area may store an operating system, an application program required by at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may Data and the like created in accordance with the use of the computer device 30 are stored. In addition, the memory 301 may include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, a flash memory card (Flash Card), At least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.
所述计算机设备30集成的模块如果以软件功能模块的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请实现上述实施例方法中的全部或部分流程,也可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一计算机可读存储介质中,该计算机程序在被处理器执行时,可实现上述各个方法实施例的步骤。其中,所述计算机程序包括计算机程序代码,所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质可以包括:能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)。If the integrated module of the computer device 30 is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiments and methods, and can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, it can implement the steps of the foregoing method embodiments. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) .
进一步地,所述计算机可用存储介质可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需的应用程序等;存储数据区可存储根据区块链节点的使用所创建的数据等。Further, the computer usable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function, etc.; the storage data area may store a block chain node Use the created data, etc.
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。The blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
在本申请所提供的几个实施例中,应该理解到,所揭露的系统,装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述模块的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式。In the several embodiments provided in this application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的,作为模块显示的部件可以是或者也可以不是物理模块,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
另外,在本申请各个实施例中的各功能模块可以集成在一个处理模块中,也可以是各个模块单独物理存在,也可以两个或两个以上模块集成在一个模块中。上述集成的模块既可以采用硬件的形式实现,也可以采用硬件加软件功能模块的形式实现。In addition, the functional modules in the various embodiments of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware, or in the form of hardware plus software functional modules.
上述以软件功能模块的形式实现的集成的模块,可以存储在一个计算机可读取存储介质中。上述软件功能模块存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)或处理器(processor)执行本申请各个实施例所述方法的部分步骤。The above-mentioned integrated modules implemented in the form of software functional modules may be stored in a computer readable storage medium. The above-mentioned software function module is stored in a storage medium and includes several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor execute the method described in the various embodiments of this application. Part of the steps.
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其他的具体形式实现本申请。因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。此外,显然“包括”一词不排除其他模块或步骤,单数不排除复数。系统权利要求中陈述的多个模块或装置也可以由一个模块或装置通过软件或者硬件来实现。第一,第二等词语用来表示名称,而并不表示任何特定的顺序。For those skilled in the art, it is obvious that the present application is not limited to the details of the foregoing exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or basic characteristics of the application. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of this application is defined by the appended claims rather than the above description, and therefore it is intended to fall into the claims. All changes in the meaning and scope of the equivalent elements of are included in this application. Any associated diagram marks in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "including" does not exclude other modules or steps, and the singular does not exclude the plural. Multiple modules or devices stated in the system claims can also be implemented by one module or device through software or hardware. Words such as first and second are used to denote names, but do not denote any specific order.
最后应说明的是,以上实施例仅用以说明本申请的技术方案而非限制,尽管参照较佳实施例对本申请进行了详细说明,本领域的普通技术人员应当理解,可以对本申请的技术方案进行修改或等同替换,而不脱离本申请技术方案的精神和范围。Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the application and not to limit them. Although the application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the application can be Make modifications or equivalent replacements without departing from the spirit and scope of the technical solution of the present application.

Claims (20)

  1. 一种计算机性能数据确定方法,其中,所述方法包括:A method for determining computer performance data, wherein the method includes:
    获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
    使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
    对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
    用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
    通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  2. 如权利要求1所述的方法,其中,所述获取计算机的多个历史性能数据序列包括:The method according to claim 1, wherein said acquiring a plurality of historical performance data sequences of the computer comprises:
    获取多个计算机的历史性能数据主序列,历史性能数据主序列中的每个元素为计算机性能的状态向量;Obtain the main sequence of historical performance data of multiple computers, and each element in the main sequence of historical performance data is a state vector of computer performance;
    以预设长度h为滑窗长度,以1为步长,从每个历史性能数据主序列中截取多个历史性能数据序列。Taking the preset length h as the sliding window length and 1 as the step length, multiple historical performance data sequences are intercepted from each main sequence of historical performance data.
  3. 如权利要求1所述的方法,其中,所述对所述多个历史性能数据序列进行分类包括:The method of claim 1, wherein the classifying the plurality of historical performance data sequences comprises:
    对所述多个历史性能数据序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical performance data sequences, and modifying the clustering result according to the received modification instruction; or
    用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类。Classify the multiple historical performance data sequences by using the trained preset convolutional neural network model.
  4. 如权利要求1所述的方法,其中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:The method according to claim 1, wherein the training of each historical performance data sequence subset is composed of the pre-trained long short-term memory network and the pre-trained long short-term memory network. The performance prediction model composed of the connection layer includes:
    判断该历史性能数据序列的数量是否大于预设阈值;Determine whether the number of historical performance data sequences is greater than a preset threshold;
    若该历史性能数据序列的数量大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数;If the number of historical performance data sequences is greater than the preset threshold, use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function;
    若该历史性能数据序列的数量不大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数。If the number of historical performance data sequences is not greater than the preset threshold, the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
  5. 如权利要求1所述的方法,其中,在所述通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据之前,所述方法还包括:The method according to claim 1, wherein, before the performance data of the computer is predicted by the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, the The method also includes:
    根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型。The sequence type of the performance data sequence to be predicted is determined according to the subset of N historical performance data sequences.
  6. 如权利要求5所述的方法,其中,所述根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型包括:The method of claim 5, wherein the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences comprises:
    计算N个历史性能数据序列子集的N个中心序列,将距离所述待预判性能数据序列最近的中心序列对应的序列类型确定为所述目标序列类型;或Calculate N central sequences of N historical performance data sequence subsets, and determine the sequence type corresponding to the central sequence closest to the performance data sequence to be predicted as the target sequence type; or
    用N个历史性能数据序列子集训练预设神经网络,其中每个历史性能数据序列子集中的每个历史性能数据序列的标签为该历史性能数据序列的序列类型,将所述待预判性能数据序列输入训练后的所述预设神经网络,根据训练后的所述预设神经网络的输出确定所述待预判性能数据序列的序列类型。Use N historical performance data sequence subsets to train the preset neural network, where the label of each historical performance data sequence in each historical performance data sequence subset is the sequence type of the historical performance data sequence, and the performance to be predicted The data sequence is input to the preset neural network after training, and the sequence type of the performance data sequence to be predicted is determined according to the output of the preset neural network after training.
  7. 如权利要求1-5中任一项所述的方法,其中,所述方法还包括:The method according to any one of claims 1-5, wherein the method further comprises:
    若所述计算机的性能数据预测结果不在预设正常状态范围,返回计算机性能异常提醒。If the prediction result of the computer's performance data is not within the preset normal state range, a notification of abnormal computer performance is returned.
  8. 一种计算机性能数据确定装置,其中,所述装置包括:A device for determining computer performance data, wherein the device includes:
    获取模块,用于获取计算机的多个历史性能数据序列;The acquisition module is used to acquire multiple historical performance data sequences of the computer;
    第一训练模块,用于使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;The first training module is configured to use the multiple historical performance data sequences to pre-train the long and short-term memory network to obtain the pre-trained long and short-term memory network;
    分类模块,用于对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;The classification module is configured to classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
    第二训练模块,用于用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;The second training module is used to use each historical performance data sequence subset to train a performance prediction composed of the pre-trained long short-term memory network and the fully connected layer behind the pre-trained long short-term memory network. Judgment model to obtain N trained performance prediction models corresponding to the N sequence types one-to-one;
    预判模块,用于通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The prediction module is used to predict the performance data of the computer through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  9. 一种计算机设备,其中,所述计算机设备包括处理器,所述处理器用于执行存储器中存储的计算机程序实现如下步骤:A computer device, wherein the computer device includes a processor, and the processor is configured to execute a computer program stored in a memory to implement the following steps:
    获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
    使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
    对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
    用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
    通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  10. 如权利要求9所述的计算机设备,其中,所述获取计算机的多个历史性能数据序列包括:9. The computer device according to claim 9, wherein said acquiring a plurality of historical performance data sequences of the computer comprises:
    获取多个计算机的历史性能数据主序列,历史性能数据主序列中的每个元素为计算机性能的状态向量;Obtain the main sequence of historical performance data of multiple computers, and each element in the main sequence of historical performance data is a state vector of computer performance;
    以预设长度h为滑窗长度,以1为步长,从每个历史性能数据主序列中截取多个历史性能数据序列。Taking the preset length h as the sliding window length and 1 as the step length, multiple historical performance data sequences are intercepted from each main sequence of historical performance data.
  11. 如权利要求9所述的计算机设备,其中,所述对所述多个历史性能数据序列进行分类包括:9. The computer device of claim 9, wherein said classifying said plurality of historical performance data sequences comprises:
    对所述多个历史性能数据序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical performance data sequences, and modifying the clustering result according to the received modification instruction; or
    用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类。Classify the multiple historical performance data sequences by using the trained preset convolutional neural network model.
  12. 如权利要求9所述的计算机设备,其中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:The computer device according to claim 9, wherein said training a long short-term memory network after the pre-training and a long-short-term memory network after the pre-training is performed with each subset of historical performance data sequences. The performance prediction model composed of the fully connected layer includes:
    判断该历史性能数据序列的数量是否大于预设阈值;Determine whether the number of historical performance data sequences is greater than a preset threshold;
    若该历史性能数据序列的数量大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数;If the number of historical performance data sequences is greater than the preset threshold, use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function;
    若该历史性能数据序列的数量不大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数。If the number of historical performance data sequences is not greater than the preset threshold, the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
  13. 如权利要求9所述的计算机设备,其中,在所述通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据之前,所述处理器用于执行存储器中存储的计算机程序还实现如下步骤:The computer device according to claim 9, wherein, before the performance data of the computer is predicted by the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer, The processor is used to execute the computer program stored in the memory and further implements the following steps:
    根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型。The sequence type of the performance data sequence to be predicted is determined according to the subset of N historical performance data sequences.
  14. 如权利要求13所述的计算机设备,其中,所述根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型包括:The computer device according to claim 13, wherein the determining the sequence type of the performance data sequence to be predicted according to the subset of N historical performance data sequences comprises:
    计算N个历史性能数据序列子集的N个中心序列,将距离所述待预判性能数据序列最近的中心序列对应的序列类型确定为所述目标序列类型;或Calculate N central sequences of N historical performance data sequence subsets, and determine the sequence type corresponding to the central sequence closest to the performance data sequence to be predicted as the target sequence type; or
    用N个历史性能数据序列子集训练预设神经网络,其中每个历史性能数据序列子集中的每个历史性能数据序列的标签为该历史性能数据序列的序列类型,将所述待预判性能数据序列输入训练后的所述预设神经网络,根据训练后的所述预设神经网络的输出确定所述待预判性能数据序列的序列类型。Use N historical performance data sequence subsets to train the preset neural network, where the label of each historical performance data sequence in each historical performance data sequence subset is the sequence type of the historical performance data sequence, and the performance to be predicted The data sequence is input to the preset neural network after training, and the sequence type of the performance data sequence to be predicted is determined according to the output of the preset neural network after training.
  15. 如权利要求9-13中任一项所述的计算机设备,其中,所述处理器用于执行存储器中存储的计算机程序还实现如下步骤:The computer device according to any one of claims 9-13, wherein the processor is configured to execute the computer program stored in the memory and further implements the following steps:
    若所述计算机的性能数据预测结果不在预设正常状态范围,返回计算机性能异常提醒。If the prediction result of the computer's performance data is not within the preset normal state range, a notification of abnormal computer performance is returned.
  16. 一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序,其中,所述计算机程序被处理器执行时实现如下步骤:A computer-readable storage medium having a computer program stored on the computer-readable storage medium, wherein, when the computer program is executed by a processor, the following steps are implemented:
    获取计算机的多个历史性能数据序列;Obtain multiple historical performance data sequences of the computer;
    使用所述多个历史性能数据序列对长短期记忆网络进行预训练,得到预训练后的长短期记忆网络;Pre-training the long and short-term memory network using the multiple historical performance data sequences to obtain a pre-trained long and short-term memory network;
    对所述多个历史性能数据序列进行分类,得到N个序列类型和与所述N个序列类型一一对应的N个历史性能数据序列子集;Classify the multiple historical performance data sequences to obtain N sequence types and N historical performance data sequence subsets corresponding to the N sequence types one-to-one;
    用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型,得到与所述N个序列类型一一对应的N个训练后的性能预判模型;Use each subset of historical performance data sequence to train a performance prediction model composed of the pre-trained long-term short-term memory network and the fully connected layer behind the pre-trained long-term short-term memory network to obtain N training performance prediction models corresponding to N sequence types one to one;
    通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据。The performance data of the computer is predicted through the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer.
  17. 如权利要求16所述的计算机可读存储介质,其中,所述获取计算机的多个历史性能数据序列包括:15. The computer-readable storage medium of claim 16, wherein said acquiring a plurality of historical performance data sequences of the computer comprises:
    获取多个计算机的历史性能数据主序列,历史性能数据主序列中的每个元素为计算机性能的状态向量;Obtain the main sequence of historical performance data of multiple computers, and each element in the main sequence of historical performance data is a state vector of computer performance;
    以预设长度h为滑窗长度,以1为步长,从每个历史性能数据主序列中截取多个历史性能数据序列。Taking the preset length h as the sliding window length and 1 as the step length, multiple historical performance data sequences are intercepted from each main sequence of historical performance data.
  18. 如权利要求16所述的计算机可读存储介质,其中,所述对所述多个历史性能数据序列进行分类包括:15. The computer-readable storage medium of claim 16, wherein the classifying the plurality of historical performance data sequences comprises:
    对所述多个历史性能数据序列进行聚类,根据接收的修改指令对聚类结果进行修改;或Clustering the multiple historical performance data sequences, and modifying the clustering result according to the received modification instruction; or
    用训练好的预设卷积神经网络模型对所述多个历史性能数据序列进行分类。Classify the multiple historical performance data sequences by using the trained preset convolutional neural network model.
  19. 如权利要求16所述的计算机可读存储介质,其中,所述用每个历史性能数据序列子集训练一个由所述预训练后的长短期记忆网络和位于所述预训练后的长短期记忆网络后的全连接层构成的性能预判模型包括:The computer-readable storage medium according to claim 16, wherein said training a long and short-term memory network after said pre-training and said long- and short-term memory after said pre-training is performed with each subset of historical performance data sequences. The performance prediction model composed of the fully connected layer after the network includes:
    判断该历史性能数据序列的数量是否大于预设阈值;Determine whether the number of historical performance data sequences is greater than a preset threshold;
    若该历史性能数据序列的数量大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的参数;If the number of historical performance data sequences is greater than the preset threshold, use the historical performance data sequence subset to optimize the parameters of the performance prediction model corresponding to the historical performance data sequence subset according to the loss function;
    若该历史性能数据序列的数量不大于预设阈值,根据损失函数用该历史性能数据序列子集优化该历史性能数据序列子集对应的性能预判模型的全连接层的参数。If the number of historical performance data sequences is not greater than the preset threshold, the historical performance data sequence subset is used according to the loss function to optimize the parameters of the fully connected layer of the performance prediction model corresponding to the historical performance data sequence subset.
  20. 如权利要求16所述的计算机可读存储介质,其中,在所述通过与所述计算机的待预判性能数据序列的序列类型对应的训练后的性能预判模型,预判所述计算机的性能数据之前,所述计算机程序被处理器执行时还实现如下步骤:The computer-readable storage medium of claim 16, wherein the performance of the computer is predicted by the trained performance prediction model corresponding to the sequence type of the performance data sequence to be predicted of the computer Before data, when the computer program is executed by the processor, the following steps are also implemented:
    根据N个历史性能数据序列子集确定所述待预判性能数据序列的序列类型。The sequence type of the performance data sequence to be predicted is determined according to the subset of N historical performance data sequences.
PCT/CN2020/118939 2020-04-23 2020-09-29 Computer performance data determining method and apparatus, computer device, and storage medium WO2021212753A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010328656.1A CN111679959A (en) 2020-04-23 2020-04-23 Computer performance data determination method and device, computer equipment and storage medium
CN202010328656.1 2020-04-23

Publications (1)

Publication Number Publication Date
WO2021212753A1 true WO2021212753A1 (en) 2021-10-28

Family

ID=72433824

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/118939 WO2021212753A1 (en) 2020-04-23 2020-09-29 Computer performance data determining method and apparatus, computer device, and storage medium

Country Status (2)

Country Link
CN (1) CN111679959A (en)
WO (1) WO2021212753A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111950928A (en) * 2020-08-24 2020-11-17 国网冀北电力有限公司 Loss reduction method and device for power distribution network, storage medium and computing equipment

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111679959A (en) * 2020-04-23 2020-09-18 平安科技(深圳)有限公司 Computer performance data determination method and device, computer equipment and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106570164A (en) * 2016-11-07 2017-04-19 中国农业大学 Integrated foodstuff safety text classification method based on deep learning
WO2017185347A1 (en) * 2016-04-29 2017-11-02 北京中科寒武纪科技有限公司 Apparatus and method for executing recurrent neural network and lstm computations
CN109347668A (en) * 2018-10-17 2019-02-15 网宿科技股份有限公司 A kind of training method and device of service quality assessment model
CN109471941A (en) * 2018-11-07 2019-03-15 中国电子科技集团公司第二十八研究所 A kind of charge classification method for coping with class imbalance
CN109873779A (en) * 2019-01-30 2019-06-11 浙江工业大学 A kind of grading type wireless identification of signal modulation method based on LSTM
CN111679959A (en) * 2020-04-23 2020-09-18 平安科技(深圳)有限公司 Computer performance data determination method and device, computer equipment and storage medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017185347A1 (en) * 2016-04-29 2017-11-02 北京中科寒武纪科技有限公司 Apparatus and method for executing recurrent neural network and lstm computations
CN106570164A (en) * 2016-11-07 2017-04-19 中国农业大学 Integrated foodstuff safety text classification method based on deep learning
CN109347668A (en) * 2018-10-17 2019-02-15 网宿科技股份有限公司 A kind of training method and device of service quality assessment model
CN109471941A (en) * 2018-11-07 2019-03-15 中国电子科技集团公司第二十八研究所 A kind of charge classification method for coping with class imbalance
CN109873779A (en) * 2019-01-30 2019-06-11 浙江工业大学 A kind of grading type wireless identification of signal modulation method based on LSTM
CN111679959A (en) * 2020-04-23 2020-09-18 平安科技(深圳)有限公司 Computer performance data determination method and device, computer equipment and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111950928A (en) * 2020-08-24 2020-11-17 国网冀北电力有限公司 Loss reduction method and device for power distribution network, storage medium and computing equipment
CN111950928B (en) * 2020-08-24 2024-02-06 国网冀北电力有限公司 Loss reduction method and device for power distribution network, storage medium and computing equipment

Also Published As

Publication number Publication date
CN111679959A (en) 2020-09-18

Similar Documents

Publication Publication Date Title
WO2021212752A1 (en) Device index data-based anomaly detection method and apparatus, device, and storage medium
US7406452B2 (en) Machine learning
US11531987B2 (en) User profiling based on transaction data associated with a user
CN110599335A (en) User financial risk assessment method and device based on multiple models
US10580272B1 (en) Techniques to provide and process video data of automatic teller machine video streams to perform suspicious activity detection
WO2022134581A1 (en) Test case sorting method and related device
CN110852881B (en) Risk account identification method and device, electronic equipment and medium
CN111612039A (en) Abnormal user identification method and device, storage medium and electronic equipment
WO2020197757A1 (en) Data-driven deep learning model generalization analysis and improvement
Lee et al. ReRe: A lightweight real-time ready-to-go anomaly detection approach for time series
WO2021212753A1 (en) Computer performance data determining method and apparatus, computer device, and storage medium
CN111125529A (en) Product matching method and device, computer equipment and storage medium
Gu et al. Online anomaly detection with sparse Gaussian processes
CN112883990A (en) Data classification method and device, computer storage medium and electronic equipment
US11551120B2 (en) Systems and methods for predicting performance
WO2021196474A1 (en) User interest profiling method and related device
WO2021147404A1 (en) Dependency relationship classification method and related device
CN112184059A (en) Scoring analysis method and device, electronic equipment and storage medium
CN115795466B (en) Malicious software organization identification method and device
CN113780675B (en) Consumption prediction method and device, storage medium and electronic equipment
CN116245630A (en) Anti-fraud detection method and device, electronic equipment and medium
US20230186214A1 (en) Systems and methods for generating predictive risk outcomes
US11487964B2 (en) Comprehensive data science solution for segmentation analysis
US10353928B2 (en) Real-time clustering using multiple representatives from a cluster
US20240202816A1 (en) Systems and methods for dynamically generating pre-approval data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20931963

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20931963

Country of ref document: EP

Kind code of ref document: A1