WO2020098016A1 - 一种网络流量预测方法、装置及电子设备 - Google Patents

一种网络流量预测方法、装置及电子设备 Download PDF

Info

Publication number
WO2020098016A1
WO2020098016A1 PCT/CN2018/119359 CN2018119359W WO2020098016A1 WO 2020098016 A1 WO2020098016 A1 WO 2020098016A1 CN 2018119359 W CN2018119359 W CN 2018119359W WO 2020098016 A1 WO2020098016 A1 WO 2020098016A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
business
network traffic
layer
traffic prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2018/119359
Other languages
English (en)
French (fr)
Inventor
柯婉婉
庄虔泰
林少惠
陈艺超
张雅虹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wangsu Science and Technology Co Ltd
Original Assignee
Wangsu Science and Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wangsu Science and Technology Co Ltd filed Critical Wangsu Science and Technology Co Ltd
Priority to EP18940464.3A priority Critical patent/EP3879754A4/en
Priority to US17/043,639 priority patent/US11252044B2/en
Publication of WO2020098016A1 publication Critical patent/WO2020098016A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour

Definitions

  • Embodiments of the present invention relate to the field of Internet technology, and in particular, to a network traffic prediction method, device, and electronic equipment.
  • the network traffic prediction model is the basis for network performance analysis and network design planning.
  • An accurate network traffic prediction model can accurately predict traffic, which is helpful to understand the trend of customer traffic changes, and can make necessary work preparations in advance for various changes in traffic to ensure service quality.
  • the traffic prediction method based on the time series is directly based on the time series method. No matter what kind of optimized time series model, the accuracy of the time series cannot be effectively guaranteed in the case of network abnormalities or strong human intervention, and it is easy Influenced by the nature of past historical cycles.
  • the flow prediction method based on linear programming has a simple model and fails to consider sufficient flow influencing factors and periodic fluctuations, which may easily lead to over-fitting and result in prediction failure.
  • the purpose of the embodiments of the present invention is to provide a network traffic prediction method, device and electronic equipment, so that the network traffic can be predicted by a multi-layer network traffic prediction model, and the influence of different service level variables on the bandwidth value can be explored to predict More accurate customer network traffic bandwidth value.
  • the embodiments of the present invention provide a network traffic prediction method, including the following steps:
  • the superior network traffic prediction model includes: a linear model
  • the calculation model of the lower-level network traffic prediction model is substituted into the upper-level network traffic prediction model to obtain a complete network traffic prediction model.
  • Embodiments of the present invention also provide a network traffic prediction device, including:
  • the upper-level traffic prediction module is used to establish a higher-level network traffic prediction model.
  • the upper-level network traffic prediction model includes: a linear model;
  • the lower-level traffic prediction module is configured to use the independent variables of the linear model in the upper-level network traffic prediction model established by the upper-level traffic prediction module as dependent variables, establish at least one layer of lower-level network traffic prediction models, and The independent variables in the network traffic prediction model are disassembled until the independent variables of each of the subordinate network traffic prediction models can be obtained naturally;
  • the superposition module is used to substitute the calculation model of the lower-level network traffic prediction model established by the lower-level traffic prediction module into the upper-level network traffic prediction model established by the upper-level traffic prediction module to obtain a complete network traffic prediction model.
  • Embodiments of the present invention also provide an electronic device, including:
  • At least one processor and,
  • a memory communicatively connected to the at least one processor; wherein,
  • the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the network traffic prediction method described above.
  • Embodiments of the present invention also provide an electronic device, including:
  • a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the network traffic prediction method described above is implemented.
  • the embodiments of the present invention establish a multi-level network traffic prediction model by subdividing customer service data to find the hierarchical business factors affected by traffic, and at this multi-level network traffic prediction model
  • it includes a higher-level network traffic prediction model and at least one lower-level network traffic prediction model.
  • the higher-level network traffic prediction model includes: linear model.
  • the independent variables of the linear model described in the upper-level network traffic prediction model are used as dependent variables, and the variables of the lower-level network traffic prediction model are used as independent variables.
  • This network traffic prediction method can explore the impact of different business-level variables on bandwidth values, thereby predicting more accurate customer network traffic bandwidth values. This method can also take into account the periodic changes of the business, as well as the development characteristics and actual situation of the business, which can effectively help to improve the prediction and management of changes in customer traffic.
  • the analysis of the business variables affecting network traffic and the establishment of a traffic layer model based on the business variables include: searching for business variables based on specific business characteristics; verifying whether the business variables pass parameter estimation; if passed, then Establish a flow layer linear model between the service variable and the flow value of the flow layer model; if not passed, set the flow layer of the flow layer model according to the magnitude and effect of the service variable on the flow value Error compensation model; sum the established linear model of the flow layer and the flow error compensation model to obtain a calculation model of the flow layer model.
  • the customer network traffic generation method can analyze the different influencing factors of the traffic business, analyze the changes of the customer's business variables from a horizontal perspective, and summarize all the business-level variables that affect the traffic to the traffic level, thereby establishing the traffic layer Linear model, and form an error compensation model based on historical abnormal fluctuations and effects.
  • analysis of the business characteristics of the business variables to determine the calculation model of the next business layer model specifically includes:
  • the changes of the customer's business variables can be analyzed from the longitudinal latitude, the calculation model of the business variable data can be determined according to the parameter analysis and evaluation results of the independent variables, and whether the current analysis level has been dismantled at the business level has been completed. Then continue to disassemble and analyze until the independent variables of each model level can be explained, and all independent variables can be obtained through natural methods.
  • FIG. 1 is a schematic flowchart of a network traffic prediction method according to a first embodiment of the present invention
  • FIG. 2 is a schematic flowchart of a network traffic prediction method according to a second embodiment of the present invention.
  • FIG. 3 is a schematic flowchart of an implementation manner of step S201 in FIG. 2;
  • FIG. 4 is a schematic flowchart of an implementation manner of step S202 in FIG. 2;
  • FIG. 5 is a structural block diagram of a network traffic prediction apparatus according to a third embodiment of the present invention.
  • FIG. 6 is a structural block diagram of a network traffic prediction apparatus according to a fourth embodiment of the present invention.
  • the first embodiment of the present invention relates to a network traffic prediction method.
  • the specific process is shown in FIG. 1 and includes:
  • Step S101 Establish a superior network traffic prediction model.
  • the superior network traffic prediction model includes: a linear model.
  • Step S102 using the independent variables of the linear model in the upper-level network traffic prediction model as dependent variables, establishing at least one layer of lower-level network traffic prediction models, and disassembling the independent variables in the lower-level network traffic prediction models,
  • the independent variables of each of the lower-level network traffic prediction models can be obtained naturally.
  • Step S103 Substituting the calculation model of the lower-level network traffic prediction model into the upper-level network traffic prediction model to obtain a complete network traffic prediction model.
  • the embodiment of the present invention finds a hierarchical business factor of traffic impact by subdividing customer business data, and establishes a multi-level network traffic prediction model.
  • the multi-level network traffic prediction model includes a superior network traffic prediction model And at least one layer of lower-level network traffic prediction models.
  • the higher-level network traffic prediction model includes: linear model.
  • the independent variables of the linear model described in the upper-level network traffic prediction model are used as dependent variables, and the variables of the lower-level network traffic prediction model are used as independent variables. Decompose and build a new lower-level traffic prediction model based on the characteristics of the independent variables until the independent variables of each lower-level network traffic prediction model can be obtained naturally.
  • This network traffic prediction method can explore the impact of different business-level variables on bandwidth values, thereby predicting more accurate customer network traffic bandwidth values. This method can also take into account the periodic changes of business, as well as the development characteristics and actual situation of the business, which can effectively help to improve the prediction and management of changes in customer traffic.
  • the second embodiment of the present invention relates to a network traffic prediction method.
  • the second embodiment is substantially the same as the first embodiment, and the main difference is that in the second embodiment of the present invention, the upper-level network traffic prediction model specifically includes a traffic layer model located at the upper layer, and the lower-level network traffic prediction model specifically includes at least In the first-level business layer model, the independent variables are specifically business variables.
  • the upper-level network traffic prediction model specifically includes a traffic layer model located at the upper layer
  • the lower-level network traffic prediction model specifically includes at least
  • the independent variables are specifically business variables.
  • the second embodiment of the present invention relates to a network traffic prediction method.
  • the specific process is shown in FIG. 2 and includes:
  • Step S201 Analyze the service variables affecting network traffic, and establish a traffic layer model according to the service variables.
  • the flow layer model includes: flow layer linear model and flow layer error compensation model.
  • flow layer linear model Before establishing the traffic layer model, first of all, according to the customer's business type, it is necessary to conduct a hierarchical analysis of the different influencing factors of the traffic business, first summarize all the business-level variables that affect the traffic to the traffic level, so as to establish a traffic layer linear model
  • the flow layer error compensation model is formed according to the historical abnormal fluctuation and influence.
  • Step S301 Find business variables based on specific business characteristics.
  • the factors affecting download traffic include but are not limited to: the size of download resources, the number of requests for download resources related websites / forums, and events corresponding to download resources Popularity, the number of users downloading resources, the average speed of downloading resources, etc.
  • step S302 is entered. If it is determined that there are no business variables, the process is exited.
  • step S302 it is verified whether the business variable passes parameter estimation.
  • step S303 is entered, and if the business variable fails the parameter evaluation, step S304 is entered.
  • Step S303 Establish a flow layer linear model between the service variable and the flow value of the flow layer model.
  • the flow layer linear model between the established business variables and the flow value of the flow layer model is:
  • Yx A + B 1 Y 1 + B 2 Y 2 + ... + B n Y n .
  • A, B1, B2 ... Bn are constants.
  • Step S304 Set a flow layer error compensation model of the flow layer model according to the magnitude and effect of the service variable on the flow value.
  • the flow layer error compensation model between the set business variables and the flow value of the flow layer model is:
  • Step S305 Sum the established linear model of the flow layer and the error compensation model of the flow layer to obtain a calculation model of the flow layer model.
  • a flow layer model is obtained, such as:
  • Y is the target curve value, that is, the network traffic value.
  • the independent variables of the linear model in the upper-level network traffic prediction model After establishing the upper-level network traffic prediction model, it is necessary to execute the independent variables of the linear model in the upper-level network traffic prediction model as dependent variables, establish at least one layer of lower-level network traffic prediction models, and perform The independent variables in are disassembled until each independent variable of the lower-level network traffic prediction model can be obtained naturally.
  • at least one business layer model needs to be established by using independent variables (ie, business variables) of the traffic layer linear model in the traffic layer model as dependent variables.
  • the independent variables (ie, business variables) of the linear model of the traffic layer are used as dependent variables
  • the variables of the first layer business layer model are used as independent variables
  • a new business layer model is established based on the characteristics of the independent variables.
  • step S202 the business characteristics of the business variables are analyzed to determine the calculation model of the next business layer model.
  • the calculation model of the business variable (as the dependent variable) and the independent variable in the business layer model when specifically executed, first determine whether the business variable has a periodic or linear relationship. If the business variable has a periodic or linear relationship, a business layer linear model between the business variable and the next independent variable of the business layer model is established; if the business variable does not have a periodic and linear relationship, Then, a qualitative analysis is performed on the business variables, and a qualitative analysis model is set when there is a qualitative relationship.
  • a time series model or a linear model can be established according to the characteristics of the business variable.
  • FIG. 4 which specifically includes:
  • Step S401 verify whether the service variable has periodicity.
  • Step S402 If the business variable has periodicity, then establish a time series model between the business variable and the time independent variable.
  • Step S403 If the service variable has no periodicity, it is further verified whether there is a linear relationship between the service variable and the independent variable.
  • steps S401-S403 specifically, Y 1 , Y 2 ,..., Y n and other business variables obtained by analyzing the traffic layer model are analyzed for periodicity, normal distribution, etc.
  • this includes testing the pure randomness and stationarity of the variable sequence.
  • the sequence can be divided into different types, and different analysis methods are used for different types of sequences, including:
  • step S403 For a pure random sequence, that is, a white noise sequence, there is no correlation between the items of the sequence, the sequence is performing random fluctuations in complete disorder, the periodic analysis of the sequence is terminated, and the linear relationship analysis of step S403 is entered.
  • ARMA model Auto-Regressive and Moving Average, autoregressive moving average model
  • step S402 For non-stationary series, convert it to a stationary series. If it is stationary after differential operation, call the series a differential stationary series. You can use the ARIMA model (Autoregressive Integrated Moving Average Model, differential autoregressive moving average model) Analysis and fitting, that is to say step S402, if there is no stationarity, the linear relationship analysis of step S403 is also used.
  • ARIMA model Autoregressive Integrated Moving Average Model, differential autoregressive moving average model
  • variable Y i is a time series object
  • p is the order of autoregression
  • q is the moving average term
  • step S403 For business variables that do not pass the period verification in step S402, linear analysis processing between influencing factors (independent variables) and predicted objects (dependent variables, ie, business variables) is performed to verify correlation coefficients and linearity (ie, step S403).
  • step S404 if it is verified in step S403 that there is a linear relationship, a linear model between the business variable and the independent variable in the currently analyzed business layer model is established, and finally a regression coefficient is obtained to establish a regression equation.
  • Y j a + b 1 z 1 + b 2 z 2 + ... + b j z j .
  • Y j is the business variable of the traffic layer model
  • the dependent variable of the currently analyzed business layer model a, b 1 , b 2 ... b j is a constant, z 1 , z 2 ... z j is the currently analyzed business
  • the independent variables of the layer model are the factors that affect Y j .
  • step S302-step S304 When establishing a time series model between the business variable and the time independent variable or establishing a linear model between the business variable and the independent variable in the currently analyzed business layer model, reference may be made to step S302-step S304. However, the disassembly of the independent variables of the traffic layer is performed in S302-step S304, and the disassembly of the independent variables of the business layer is performed in this embodiment.
  • step S405 if the judgment result in step S403 is negative, a qualitative analysis is performed on the business variable, and a qualitative analysis model is set when there is a qualitative relationship.
  • step S201 and step S202 are completed, the method further includes:
  • This step is mainly to determine whether the analysis and evaluation of all business variables of the currently analyzed business layer model have been completed from a horizontal perspective. If it is determined that the disassembly has been completed, then go to step S204, if not, go to the step S202.
  • the network traffic prediction method performs hierarchical analysis on different influencing factors of traffic services based on the generation method of customer network traffic, analyzes the changes of customer's business variables from a horizontal perspective, and summarizes all business-level variables that affect traffic To the flow level, to establish a linear model of the flow level, and form an error compensation model according to the historical abnormal fluctuations and effects.
  • the third embodiment of the present invention relates to a network traffic prediction device 500, as shown in FIG. 5, including:
  • the upper-level traffic prediction module 501 is used to establish a higher-level network traffic prediction model.
  • the upper-level network traffic prediction model includes: a linear model.
  • the lower-level traffic prediction module 502 is configured to use the independent variable of the linear model in the upper-level network traffic prediction model established by the upper-level traffic prediction module 501 as a dependent variable, establish at least one layer of lower-level network traffic prediction models, and The independent variables in the lower-level network traffic prediction model are disassembled until each independent variable of the lower-level network traffic prediction model can be obtained naturally.
  • the superposition module 503 is used to substitute the calculation model of the lower-level network traffic prediction model established by the lower-level traffic prediction module 502 into the upper-level network traffic prediction model established by the upper-level traffic prediction module 501 to obtain complete network traffic Forecasting model.
  • the core of the embodiments of the present invention is to establish a multi-level network traffic prediction model by subdividing customer business data to find the hierarchical business factors affected by traffic.
  • the multi-level network traffic prediction model includes superiors Network traffic prediction model and at least one layer of lower-level network traffic prediction model.
  • the higher-level network traffic prediction model includes: linear model.
  • the independent variables of the linear model in the upper-level network traffic prediction model are used as dependent variables, and the variables of the lower-level network traffic prediction model are used as independent variables. Decompose and build a new lower-level traffic prediction model based on the characteristics of the independent variables until the independent variables of each lower-level network traffic prediction model can be obtained naturally.
  • This network traffic prediction method can explore the impact of different business-level variables on bandwidth values, thereby predicting more accurate customer network traffic bandwidth values. This method can also take into account the periodic changes of business, as well as the development characteristics and actual situation of the business, which can effectively help to improve the prediction and management of changes in customer traffic.
  • this embodiment is a system example corresponding to the first embodiment, and this embodiment can be implemented in cooperation with the first embodiment.
  • the relevant technical details mentioned in the first embodiment are still valid in this embodiment, and in order to reduce repetition, they will not be repeated here.
  • the relevant technical details mentioned in this embodiment can also be applied in the first embodiment.
  • modules involved in this embodiment are all logical modules.
  • a logical unit may be a physical unit or a part of a physical unit, or multiple physical The combination of units is realized.
  • units that are not closely related to solving the technical problems proposed by the present invention are not introduced in this embodiment, but this does not mean that there are no other units in this embodiment.
  • the fourth embodiment of the present invention relates to a network traffic prediction device 500.
  • the fourth embodiment is substantially the same as the third embodiment, the main difference is that in the fourth embodiment of the present invention, the upper-level network traffic prediction model specifically includes a traffic layer model located at the upper layer, and the lower-level network traffic prediction model specifically includes at least In the first-level business layer model, the independent variables are specifically business variables.
  • the upper-level traffic prediction module is specifically the traffic layer prediction module 5011;
  • the traffic layer prediction module 5011 is used to: analyze service variables affecting network traffic and establish a traffic layer model according to the service variables; wherein, the traffic layer model includes: a traffic layer linear model and a traffic layer error compensation model. Before establishing the traffic layer model, the traffic layer prediction module 5011 first needs to carry out a hierarchical analysis of different influencing factors of the traffic business according to the customer's business type. Layer linear model, and form flow layer error compensation model according to historical abnormal fluctuations and effects.
  • the traffic layer prediction module 5011 looks for business variables based on specific business characteristics. And verify whether the business variables are estimated by parameters. Among them, need to search for business variables based on the characteristics of customer business processes. Different types of customers have different business models and traffic influencing factors. Taking download as an example, the factors affecting download traffic include but are not limited to: the size of download resources, the number of requests for download resources related websites / forums, and events corresponding to download resources Popularity, the number of users downloading resources, the average speed of downloading resources, etc.
  • searching for business variables it is first determined whether there are business variables. If it is determined that there are business variables, it is verified whether the business variables have passed parameter estimation. If it is determined that there are no business variables, the execution process is exited.
  • the purpose of verifying whether the business variables pass the parameter estimation is to establish a traffic prediction model based on the business variables.
  • commonly used parameter estimation methods are: iterative generalized least squares, restrictive generalized least squares estimation and Markov chain Monte Carlo method, by verifying whether there is a linear relationship between business variables and flow-level variables, and variables Whether to obey the normal distribution as a whole for analysis and judgment.
  • a traffic layer linear model between the business variable and the traffic value of the traffic layer model is established.
  • the established traffic layer linear model between the business variable and the traffic value of the traffic layer model is:
  • Yx A + B 1 Y 1 + B 2 Y 2 + ... + B n Y n .
  • A, B1, B2 ... Bn are constants.
  • the flow layer error compensation model of the flow layer model is set according to the magnitude and effect of the service variable on the flow value.
  • the established linear model of the flow layer and the flow error compensation model are summed to obtain a calculation model of the flow layer model.
  • the resulting traffic layer model is:
  • Y is the target curve value, that is, the network traffic value.
  • Lower-level traffic prediction module specifically business layer prediction module 5021;
  • the business layer prediction module 5021 is required to use the independent variable (that is, the business variable) of the traffic layer linear model in the traffic layer model as the dependent variable to establish at least one business layer model.
  • the business layer prediction module 5021 is used to: analyze the business characteristics of the business variables to determine the calculation model of the next business layer model; in determining the business layer model, business variables (as dependent variables) and calculation of independent variables in the business layer model
  • the operation mode may be, for example, to first determine whether the business variable has a periodic or linear relationship. If the business variable has a periodic or linear relationship, a business layer linear model between the business variable and the next independent variable of the business layer model is established; if the business variable does not have a periodic and linear relationship, Then, a qualitative analysis is performed on the business variables, and a qualitative analysis model is set when there is a qualitative relationship.
  • a time series model or a linear model can be established according to the characteristics of the business variable.
  • variable Y i is a time series object
  • p is the order of autoregression
  • q is the moving average term
  • the business variable is not periodic, it is further verified whether there is a linear relationship between the business variable and the independent variable. If it is verified that there is a linear relationship, a linear model between the business variable and the independent variable in the business layer model currently analyzed is established, and finally a regression coefficient is obtained to establish a regression equation. For example, assuming that the variable Y j is a linear relationship object, you can establish:
  • Y j a + b 1 z 1 + b 2 z 2 + ... + b j z j .
  • Y j is the business variable of the traffic layer model
  • the dependent variable of the currently analyzed business layer model a, b 1 , b 2 ... b j is a constant, z 1 , z 2 ... z j is the currently analyzed business
  • the independent variables of the layer model are the factors that affect Y j .
  • step S302-step S304 When establishing a time series model between the business variable and the time independent variable or establishing a linear model between the business variable and the independent variable in the currently analyzed business layer model, reference may be made to step S302-step S304. However, the disassembly of the independent variables of the traffic layer is performed in S302-step S304, and the disassembly of the independent variables of the business layer is performed in this embodiment.
  • the business layer traffic module 5021 After analyzing the business characteristics of the business variables to determine the calculation model of the next business layer model, the business layer traffic module 5021 also determines whether the currently analyzed business layer model has completed the disassembly of the business layer. If the disassembly is not completed, it will be The independent variable of the linear model in the currently analyzed business layer model is used as the dependent variable, and then returns to the operation of analyzing the business characteristics of the business variable to determine the calculation model of the next business layer model; The calculation model of the business layer model is substituted into the traffic layer prediction model to obtain a complete network traffic prediction model.
  • the network traffic prediction device performs hierarchical analysis on different influencing factors of the traffic service based on the generation method of the customer network traffic, analyzes the changes of the customer's business variables from a horizontal perspective, and summarizes all the business-level variables affecting the traffic To the flow level, to establish a linear model of the flow level, and form an error compensation model according to the historical abnormal fluctuations and effects.
  • the fifth embodiment of the present invention relates to an electronic device including at least one processor; and,
  • a memory communicatively connected to the at least one processor; wherein,
  • the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute any one of the first and second embodiments The network traffic prediction method.
  • the memory and the processor are connected in a bus manner.
  • the bus may include any number of interconnected buses and bridges, and the bus connects one or more processors and various circuits of the memory together.
  • the bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, etc., which are well known in the art, and therefore, they will not be described further herein.
  • the bus interface provides an interface between the bus and the transceiver.
  • the transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on the transmission medium.
  • the data processed by the processor is transmitted on the wireless medium through the antenna. Further, the antenna also receives the data and transmits the data to the processor.
  • the processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions.
  • the memory can be used to store data used by the processor when performing operations.
  • the sixth embodiment of the present invention relates to a computer-readable storage medium that stores a computer program.
  • the computer program is executed by the processor, the above method embodiments are implemented.
  • a program which is stored in a storage medium and includes several instructions to make a device ( It may be a single chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in the embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code .

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明实施例涉及互联网技术领域，公开了一种网络流量预测方法、装置及电子设备。本发明中，网络流量预测方法，包括：建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型；将上级网络流量预测模型中线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对下级网络流量预测模型中的自变量进行拆解，直至每一下级网络流量预测模型的自变量均能自然获取；将下级网络流量预测模型的计算模型代入上级网络流量预测模型中，得到完整的网络流量预测模型。使得能够通过多层网络流量预测模型对网络流量进行预测，可以探索不同业务层面变量对带宽值的影响，从而预测出更准确的客户网络流量。

Description

一种网络流量预测方法、装置及电子设备 技术领域
本发明实施例涉及互联网技术领域，特别涉及一种网络流量预测方法、装置及电子设备。
背景技术
在互联网技术发展中，针对网络流量进行建模和预测研究备受关注。网络流量预测模型是进行网络性能分析和网络设计规划的基础。一个准确的网络流量预测模型，能够准确预测流量，有利于了解客户流量变化趋势，对于流量的各种变化能提前做好必要的工作准备，保障服务质量。
在大数据背景下，网络业务类别的急剧增加导致网络流量性质多样化，传统的流量预测模型已不适用于当今乃至下一代互联网流量的分析与预测，因此对网络流量的预测势在必行。现有技术中，各种流量预测方法，存在但不限于以下几个问题：
1、基于时间序列的流量预测方法，直接基于时间序列方法，不论是何种优化时间序列模型，在出现网络异常或者较强的人为干预等情况时，时间序列的准确性无法有效保证，并且容易被过往的历史周期性质所影响。
2、基于线性规划的流量预测方法，模型简单，未能考虑足够的流量影响因素以及周期性的波动影响，容易造成过度拟合导致预测失败。
由此，有必要建立一种新的、能够适应新的业务发展需求的网络流量预测方法。
发明内容
本发明实施方式的目的在于提供一种网络流量预测方法、装置及电子设备，使得能够通过多层网络流量预测模型对网络流量进行预测，可以探索不同业务层面变量对带宽值的影响，从而预测出更准确的客户网络流量带宽值。
为解决上述技术问题，本发明的实施方式提供了一种网络流量预测方法，包括以下步骤：
建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型；
将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下 级网络流量预测模型的自变量均能自然获取；
将所述下级网络流量预测模型的计算模型代入所述上级网络流量预测模型中，得到完整的网络流量预测模型。
本发明的实施方式还提供了一种网络流量预测装置，包括：
上级流量预测模块，用于建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型；
下级流量预测模块，用于将所述上级流量预测模块建立的所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取；
叠加模块，用于将所述下级流量预测模块建立的所述下级网络流量预测模型的计算模型代入所述上级流量预测模块建立的所述上级网络流量预测模型中，得到完整的网络流量预测模型。
本发明的实施方式还提供了一种电子设备，包括：
至少一个处理器；以及，
与所述至少一个处理器通信连接的存储器；其中，
所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述所述的网络流量预测方法。
本发明的实施方式还提供了一种电子设备，包括：
计算机可读存储介质，存储有计算机程序，所述计算机程序被处理器执行时实现上述所述的网络流量预测方法。
本发明实施方式相对于现有技术而言，通过对客户业务数据的细分，找到流量影响的层次结构业务因素，建立一种多层级的网络流量预测模型，在该多层级的网络流量预测模型中，包括上级网络流量预测模型和至少一层下级网络流量预测模型。其中，上级网络流量预测模型包括：线性模型。同时，在下层的网络流量预测模型中，以上级网络流量预测模型中所述线性模型的自变量作为因变量，以下级网络流量预测模型的变量作为自变量，对每个变量进行因素的分析和分解，并基于自变量的特点建立新的下层流量预测模型，直至每一下级网络流量预测模型的自变量均能自然获取。最后，将下级网络流量预测模型的计算模型代入上级网络流量预测模型中，从而得出完整的流量预测模型。此网络流量预测方法可以探索不同业务层面变量对带宽值的影响，从而预测出更准确的客户网络流量带宽值。此方法还能兼顾业 务的周期性变化，也能兼顾业务的发展特点和实际情况，从而能有效帮助提高对于客户流量变化情况的预测和管理。
另外，所述分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型，具体包括：基于具体的业务特性，寻找业务变量；验证所述业务变量是否通过参数估计；若通过，则建立所述业务变量与所述流量层模型的流量值间的流量层线性模型；若未通过，则根据所述业务变量对流量值的影响大小和作用，设定所述流量层模型的流量层误差补偿模型；对建立的所述流量层线性模型和所述流量误差补偿模型求和，得到所述流量层模型的计算模型。这样可以基于客户网络流量产生方式，对流量业务的不同影响因素进行层次分析，从横向角度分析客户的业务变量的变化，将各种影响流量的业务层次变量全都汇总到流量层次，从而建立流量层线性模型，并根据历史的异常波动和影响形成误差补偿模型。
另外，所述分析所述业务变量的业务特性以确定下一业务层模型的计算模型，具体包括：
验证所述业务变量是否存在周期性和线性关系；若所述业务变量存在周期性或线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型；若所述业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。这样可以从纵向纬度分析客户的业务变量的变化，根据自变量的参数分析评估结果，决定业务变量数据的计算模型，并验证当前分析层级是否已经完成业务层级的拆解，若未拆解完成，则继续拆解和分析，直到各模型层级的自变量均可以得到解释，并且所有的自变量均可以通过自然的方式获取到。通过对客户网络流量的产生方式进行细分，找到能兼顾业务的周期性变化，也能兼顾业务的发展特点和实际情况的流量因素分析和预测方法，建立合理的流量评估数据模型。能够更好地解释自变量与因变量间变化的规律，与实际的业务变化情况更加贴切，也能更好的解释客户流量变化的情况。
附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定，附图中具有相同参考数字标号的元件表示为类似的元件，除非有特别申明，附图中的图不构成比例限制。
图1是根据本发明第一实施方式的网络流量预测方法的流程示意图；
图2是根据本发明第二实施方式的网络流量预测方法的流程示意图；
图3根据图2中步骤S201的实施方式的流程示意图；
图4根据图2中步骤S202的实施方式的流程示意图；
图5是根据本发明第三实施方式的网络流量预测装置的结构框图；
图6是根据本发明第四实施方式的网络流量预测装置的结构框图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合附图对本发明的各实施方式进行详细的阐述。然而，本领域的普通技术人员可以理解，在本发明各实施方式中，为了使读者更好地理解本申请而提出了许多技术细节。但是，即使没有这些技术细节和基于以下各实施方式的种种变化和修改，也可以实现本申请所要求保护的技术方案。
本发明的第一实施方式涉及一种网络流量预测方法，具体流程如图1所示，包括：
步骤S101,建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型。
步骤S102,将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取。
步骤S103,将所述下级网络流量预测模型的计算模型代入所述上级网络流量预测模型中，得到完整的网络流量预测模型。
本发明实施方式通过对客户业务数据的细分，找到流量影响的层次结构业务因素，建立一种多层级的网络流量预测模型，在该多层级的网络流量预测模型中，包括上级网络流量预测模型和至少一层下级网络流量预测模型。其中，上级网络流量预测模型包括：线性模型。同时，在下层的网络流量预测模型中，以上级网络流量预测模型中所述线性模型的自变量作为因变量，以下级网络流量预测模型的变量作为自变量，对每个变量进行因素的分析和分解，并基于自变量的特点建立新的下层流量预测模型，直至每一下级网络流量预测模型的自变量均能自然获取。最后，将下级网络流量预测模型的计算模型代入上级网络流量预测模型中，从而得出完整的流量预测模型。此网络流量预测方法可以探索不同业务层面变量对带宽值的影响，从而预测出更准确的客户网络流量带宽值。此方法还能兼顾业务的周期性变化，也能兼顾业务的发展特点和实际情况，从而能有效帮助提高对于客户流量变化情况的预测和管理。
本发明的第二实施方式涉及一种网络流量预测方法。第二实施方式与第一实施方式大致相同，主要区别之处在于：在本发明第二实施方式中，上级网络流量预测模型具体包括位于上层的流量层模型，而下级网络流量预测模型具体包括至少一级业务层模型，自变量具体为业务变量。下面，将通过图2对本发明实施方式进行详细说明。
本发明的第二实施方式涉及一种网络流量预测方法，具体流程如图2所示，包括：
步骤S201,分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型。
其中，流量层模型包括：流量层线性模型和流量层误差补偿模型。在建立流量层模型之前，首先需要根据客户的业务类型，对流量业务的不同影响因素进行层次分析，先将各种影响流量的业务层次变量全都汇总到流量层次，从而建立流量层线性模型，并根据历史的异常波动和影响形成流量层误差补偿模型。
本步骤具体操作时，可通过图3所示的流程来实现，具体包括：
步骤S301,基于具体的业务特性，寻找业务变量。
其中，需要基于客户业务流程的特性，进行业务变量的寻找。不同类型客户存在着不同的业务模式和流量的影响因素，以下载为例，影响下载流量的因素包括但不限于：下载资源的大小，下载资源相关网站/论坛的请求数，下载资源对应的事件的热度、下载资源的用户数、下载资源的平均速度等。在业务变量的寻找过程中，首先判断是否还有业务变量，若判断还有业务变量，则进入步骤S302，若判断已经没有业务变量，则退出流程。
步骤S302,验证所述业务变量是否通过参数估计。
验证业务变量是否通过参数估计的目的是：建立基于业务变量的流量预测模型。其中，常用的参数估计方法有：迭代广义最小二乘法、限制性的广义最小二乘估计和马尔科夫链蒙特卡罗法，通过验证业务变量与流量层级的变量间是否存在线性关系，以及业务变量总体上是否服从正态分布来进行分析和判断。若业务变量通过参数评估，则进入步骤S303，若业务变量未通过参数评估，则进入步骤S304。
步骤S303,建立业务变量与所述流量层模型的流量值间的流量层线性模型。
例如，建立的业务变量与流量层模型的流量值间的流量层线性模型为：
Yx＝A+B 1Y 1+B 2Y 2+...+B nY n。
其中，Yx为流量层线性模型的目标曲线值，即线性流量值，其中n>＝1,为在步骤S302分析评估出的可用的业务变量数目。A、B1、B2……Bn为常数。
步骤S304,根据所述业务变量对流量值的影响大小和作用，设定所述流量层模型的流量层误差补偿模型。
例如，设定的业务变量与流量层模型的流量值间的流量层误差补偿模型为：
Ym＝f(x 1)+f(x 2)...+f(x m).
其中，Ym为流量层误差补偿模型的目标曲线值，即误差流量值，其中m>＝0,为在步骤S302分析评估出的用于误差补偿的业务变量数目。
步骤S305，对建立的所述流量层线性模型和所述流量层误差补偿模型求和，得到所述流量层模型的计算模型。
根据步骤S304和步骤S305建立的流量层线性模型和流量层误差补偿模型求和，得到流量层模型，如:
Y＝Yx+Ym＝A+B 1Y 1+B 2Y 2+...+B nY n+f(x 1)+f(x 2)...+f(x m)。其中，Y为目标曲线值，即网络流量值。
在建立上级网络流量预测模型之后，需要执行将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取的步骤。具体到本实施方式来说，在建立上级的流量层模型之后，需要将流量层模型中流量层线性模型的自变量(即业务变量)作为因变量，建立至少一层业务层模型。
具体来说，即是把流量层线性模型的自变量(即业务变量)作为因变量，使用第一层业务层模型的变量作为自变量，并基于自变量的特点建立新的业务层模型。在建立业务层模型时，具体需要对于业务变量进行业务特性分析，根据业务层模型中自变量的参数分析评估结果，决定业务变量数据的计算模型，其中就包括时间序列、线性回归和误差补偿模型等，并验证当前分析层级是否已经完成业务层级的拆解，若未拆解完成，则继续进行分析和拆解，直至各业务层级的自变量均可以得到解释，并且所有的自变量均可以通过自然的方式获取到。即是说，在执行步骤S201之后，还可以包括：
步骤S202,分析业务变量的业务特性以确定下一业务层模型的计算模型。
本步骤中，即在确定业务层模型中，业务变量(作为因变量)与业务层模型中自变量的计算模型，具体执行时，首先判断所述业务变量是否存在周期性或线性关系。若所述业务变量存在周期性或线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型；若所述业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。而在建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型时，根据业务变量的特性，可以建立时间序列模型或线性模型。
下面，将通过图4所示的操作流程来实现，具体包括：
步骤S401，验证所述业务变量是否存在周期性。
步骤S402，若所述业务变量存在周期性，则建立所述业务变量与时间自变量之间的时间序列模型。
步骤S403，若所述业务变量无周期性，则进一步验证所述业务变量与所述自变量之间是否存在线性关系。
步骤S401-步骤S403中，具体是对在流量层模型分析得出的Y 1，Y 2，...，Y n等业务变量进行周期性、正态分布性等的分析。在对业务变量进行周期性验证时，其中包括对变量序列的纯随机性和平稳性进行检验，据检验结果可以将序列分为不同的类型，对不同类型的序列采用不同的分析方法，包括：
i.对于纯随机序列，即白噪声序列，序列的各项之间没有任何相关关系，序列在进行完全无序的随机波动，终止对该序列的周期分析，并进入步骤S403的线性关系分析。
ii.对于平稳非白噪声序列，它的均值和方差是常数，可以使用ARMA模型(Auto-Regressive and Moving Average Model，自回归滑动平均模型)作为平稳序列拟合模型，即进入步骤S402。
iii.对于非平稳序列，将其转变为平稳序列，若经差分运算后具有平稳性，称该序列为差分平稳序列，可以使用ARIMA模型(Autoregressive Integrated Moving Average Model,差分自回归移动平均模型)进行分析和拟合，即进入步骤S402，若未具有平稳性，则同样采用步骤S403的线性关系分析。
若经验证，业务变量存在周期性，则建立所述业务变量与时间自变量之间的时间序列模型。例如，假设其中变量Y i为时间序列对象，则可以建立
Figure PCTCN2018119359-appb-000001
其中，p为自回归的阶数，q为移动平均项。
对未通过步骤S402周期验证的业务变量，则进行影响因素(自变量)和预测对象(因变量，即业务变量)之间的线性分析处理，进行相关系数及线性的验证(即步骤S403)。
步骤S404，若步骤S403验证存在线性关系，则建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型，并最终求得回归系数建立回归方程。
例如，假设其中变量Y j为线性关系对象，则可以建立：
Y j＝a+b 1z 1+b 2z 2+…+b jz j。
其中，Y j为流量层模型的业务变量，当前分析的业务层模型的因变量，a、b 1、b 2……b j为常数，z 1、z 2……z j为当前分析的业务层模型的自变量，即影响Y j的因素。
在建立所述业务变量与时间自变量之间的时间序列模型或建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型时，可以参考步骤S302-步骤S304。只是，S302-步骤S304中进行的是流量层自变量的拆解，而本实施方式中，进行的是业务层自变量的拆解。
步骤S405，若步骤S403的判断结果为否，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。
本步骤中，对业务变量进行定性分析，假设其中业务变量Y z为定性变量，则设定如下定性分析模型：
Figure PCTCN2018119359-appb-000002
在完成步骤S201和步骤S202之后，还包括：
S203，判断当前分析的业务层模型是否完成业务层级的拆解。
本步骤，主要是从横向角度，判断是否已经完成当前分析的业务层模型所有业务变量的分析与评估，若判断已经完成拆解，则转入步骤S204,若未完成拆解，则转入步骤S202。
S204，若完成拆解，则将所述下级网络流量预测模型的计算模型代入所述上级网络流量预测模型中，得到完整的网络流量预测模型。
本步骤具体执行时，具体结合流量层模型和下层业务层模型，最终可得完整的流量预测模型：
Figure PCTCN2018119359-appb-000003
本发明实施方式的网络流量预测方法，基于客户网络流量产生方式，对流量业务的不同影响因素进行层次分析，从横向角度分析客户的业务变量的变化，将各种影响流量的业务层次变量全都汇总到流量层次，从而建立流量层线性模型，并根据历史的异常波动和影响形成误差补偿模型。同时，从纵向纬度分析客户的业务变量的变化，根据自变量的参数分析评估结果，决定业务变量数据的计算模型，并验证当前分析层级是否已经完成业务层级的拆解，若未拆解完成，则继续拆解和分析，直到各模型层级的自变量均可以得到解释，并且所有的自变量均可以通过自然的方式获取到。本发明实施方式通过建立多层线性模型，并结合误差补偿模型对波动影响因素进行补偿。通过对客户网络流量的产生方式进行细分，找到能兼顾业务的周期性变化，也能兼顾业务的发展特点和实际情况的流量因素分析和预测方法，建立合理的流量评估数据模型。本发明实施方式的技术方案，能够更好地解释自变量与因变量间变化的规律，与实际的业务变化情况更加贴切，也能更好的解释客户流量变化的情况。
上面各种方法的步骤划分，只是为了描述清楚，实现时可以合并为一个步骤或者对某些步骤进行拆分，分解为多个步骤，只要包括相同的逻辑关系，都在本专利的保护范围内；对算法中或者流程中添加无关紧要的修改或者引入无关紧要的设计，但不改变其算法和流程的核心设计都在该专利的保护范围内。
本发明第三实施方式涉及一种网络流量预测装置500，如图5所示，包括：
上级流量预测模块501，用于建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型。
下级流量预测模块502，用于将所述上级流量预测模块501建立的所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取。
叠加模块503，用于将所述下级流量预测模块502建立的所述下级网络流量预测模型的计算模型代入所述上级流量预测模块501建立的所述上级网络流量预测模型中，得到完整的网络流量预测模型。
本发明实施方式的核心在于，通过对客户业务数据的细分，找到流量影响的层次结构业务因素，建立一种多层级的网络流量预测模型，在该多层级的网络流量预测模型中，包括上级网络流量预测模型和至少一层下级网络流量预测模型。其中，上级网络流量预测模型包括：线性模型。同时，在下层的网络流量预测模型中，以上级网络流量预测模型中所述线性模型的自变量作为因变量，以下级网络流量预测模型的变量作为自变量，对每个变量进行因素的分析和分解，并基于自变量的特点建立新的下层流量预测模型，直至每一下级网络流量预测模型的自变量均能自然获取。最后，将下级网络流量预测模型的计算模型代入上级网络流量预测模型中，从而得出完整的流量预测模型。此网络流量预测方法可以探索不同业务层面变量对带宽值的影响，从而预测出更准确的客户网络流量带宽值。此方法还能兼顾业务的周期性变化，也能兼顾业务的发展特点和实际情况，从而能有效帮助提高对于客户流量变化情况的预测和管理。
不难发现，本实施方式为与第一实施方式相对应的系统实施例，本实施方式可与第一实施方式互相配合实施。第一实施方式中提到的相关技术细节在本实施方式中依然有效，为了减少重复，这里不再赘述。相应地，本实施方式中提到的相关技术细节也可应用在第一实施方式中。
值得一提的是，本实施方式中所涉及到的各模块均为逻辑模块，在实际应用中，一个逻辑单元可以是一个物理单元，也可以是一个物理单元的一部分，还可以以多个物理单元的组合实现。此外，为了突出本发明的创新部分，本实施方式中并没有将与解决本发明所提出的技术问题关系不太密切的单元引入，但这并不表明本实施方式中不存在其它的单元。
本发明第四实施方式涉及一种网络流量预测装置500。第四实施方式与第三实施方式大 致相同，主要区别之处在于：在本发明第四实施方式中，上级网络流量预测模型具体包括位于上层的流量层模型，而下级网络流量预测模型具体包括至少一级业务层模型，自变量具体为业务变量。下面，通过图6进行详细说明。其中，上级流量预测模块具体为流量层预测模块5011；
流量层预测模块5011用于：分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型；其中，流量层模型包括：流量层线性模型和流量层误差补偿模型。流量层预测模块5011在建立流量层模型之前，首先需要根据客户的业务类型，对流量业务的不同影响因素进行层次分析，先将各种影响流量的业务层次变量全都汇总到流量层次，从而建立流量层线性模型，并根据历史的异常波动和影响形成流量层误差补偿模型。
流量层预测模块5011基于具体的业务特性，寻找业务变量。并验证业务变量是否通过参数估计。其中，需要基于客户业务流程的特性，进行业务变量的寻找。不同类型客户存在着不同的业务模式和流量的影响因素，以下载为例，影响下载流量的因素包括但不限于：下载资源的大小，下载资源相关网站/论坛的请求数，下载资源对应的事件的热度、下载资源的用户数、下载资源的平均速度等。在业务变量的寻找过程中，首先判断是否还有业务变量，若判断还有业务变量，则验证所述业务变量是否通过参数估计，若判断已经没有业务变量，则退出执行进程。
其中，验证业务变量是否通过参数估计的目的是：建立基于业务变量的流量预测模型。其中，常用的参数估计方法有：迭代广义最小二乘法、限制性的广义最小二乘估计和马尔科夫链蒙特卡罗法，通过验证业务变量与流量层级的变量间是否存在线性关系，以及变量总体上是否服从正态分布来进行分析和判断。
在通过参数评估时，建立业务变量与所述流量层模型的流量值间的流量层线性模型，例如，建立的业务变量与流量层模型的流量值间的流量层线性模型为：
Yx＝A+B 1Y 1+B 2Y 2+...+B nY n。
其中，Yx为流量层线性模型的目标曲线值，即线性流量值，其中n>＝1,为分析评估出的可用变量数目。A、B1、B2……Bn为常数。
在未通过参数评估时，根据所述业务变量对流量值的影响大小和作用，设定所述流量层模型的流量层误差补偿模型。例如，设定的业务变量与流量层模型的流量值间的流量层误差补偿模型为：Ym＝f(x 1)+f(x 2)...+f(x m)。
其中，Ym为流程层误差补偿模型的目标曲线值，即误差流量值，其中m>＝0,为分析评估出的误差补偿变量数目。
最后，对建立的所述流量层线性模型和所述流量误差补偿模型求和，得到所述流量层模型的计算模型。例如得到的流量层模型为:
Y＝Yx+Ym＝A+B 1Y 1+B 2Y 2+...+B nY n+f(x 1)+f(x 2)...+f(x m)。其中，Y为目标曲线值，即网络流量值。
下级流量预测模块，具体为业务层预测模块5021；
在建立上级网络流量预测模型之后，需要执行将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取的步骤。具体到本实施方式来说，在建立上级的流量层模型之后，需要业务层预测模块5021将流量层模型中流量层线性模型的自变量(即业务变量)作为因变量，建立至少一层业务层模型。
业务层预测模块5021用于：分析所述业务变量的业务特性以确定下一业务层模型的计算模型；在确定业务层模型中，业务变量(作为因变量)与业务层模型中自变量的计算模型时，操作方式例如可以是：首先判断所述业务变量是否存在周期性或线性关系。若所述业务变量存在周期性或线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型；若所述业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。而在建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型时，根据业务变量的特性，可以建立时间序列模型或线性模型。
具体执行中，首先验证所述业务变量是否存在周期性，若所述业务变量存在周期性，则建立所述业务变量与时间自变量之间的时间序列模型，例如：
假设其中变量Y i为时间序列对象，则可以建立
Figure PCTCN2018119359-appb-000004
其中，p为自回归的阶数，q为移动平均项。
若所述业务变量无周期性，则进一步验证所述业务变量与所述自变量之间是否存在线性关系。若验证存在线性关系，则建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型，并最终求得回归系数建立回归方程。例如，假设其中变量Y j为线性关系对象，则可以建立：
Y j＝a+b 1z 1+b 2z 2+…+b jz j。
其中，Y j为流量层模型的业务变量，当前分析的业务层模型的因变量，a、b 1、b 2……b j为常数，z 1、z 2……z j为当前分析的业务层模型的自变量，即影响Y j的因素。
在建立所述业务变量与时间自变量之间的时间序列模型或建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型时，可以参考步骤S302-步骤S304。只是，S302-步骤S304中进行的是流量层自变量的拆解，而本实施方式中，进行的是业务层自变量的拆解。
若判断业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。对业务变量进行定性分析，假设其中业务变量Y z为定性变量，则设定如下定性分析模型：
Figure PCTCN2018119359-appb-000005
在分析所述业务变量的业务特性以确定下一业务层模型的计算模型之后，业务层流量模块5021还判断当前分析的业务层模型是否完成业务层级的拆解，若未完成拆解，则将当前分析的业务层模型中线性模型的自变量作为因变量，然后返回执行分析所述业务变量的业务特性以确定下一业务层模型的计算模型的操作；若拆解完成，则通过叠加模块503将业务层模型的计算模型代入流量层预测模型中，得到完整的网络流量预测模型。
本发明实施方式的网络流量预测装置，基于客户网络流量产生方式，对流量业务的不同影响因素进行层次分析，从横向角度分析客户的业务变量的变化，将各种影响流量的业务层次变量全都汇总到流量层次，从而建立流量层线性模型，并根据历史的异常波动和影响形成误差补偿模型。同时，从纵向纬度分析客户的业务变量的变化，根据自变量的参数分析评估结果，决定业务变量数据的计算模型，并验证当前分析层级是否已经完成业务层级的拆解，若未拆解完成，则继续拆解和分析，直到各模型层级的自变量均可以得到解释，并且所有的自变量均可以通过自然的方式获取到。本发明实施方式通过建立多层线性模型，并结合误差补偿模型对波动影响因素进行补偿。通过对客户网络流量的产生方式进行细分，找到能兼顾业务的周期性变化，也能兼顾业务的发展特点和实际情况的流量因素分析和预测方法，建立合理的流量评估数据模型。本发明实施方式的技术方案，能够更好地解释自变量与因变量间变化的规律，与实际的业务变化情况更加贴切，也能更好的解释客户流量变化的情况。
由于第二实施方式与本实施方式相互对应，因此本实施方式可与第二实施方式互相配合实施。第二实施方式中提到的相关技术细节在本实施方式中依然有效，在第二实施方式中所能达到的技术效果在本实施方式中也同样可以实现，为了减少重复，这里不再赘述。相应地，本实施方式中提到的相关技术细节也可应用在第二实施方式中。
本发明第五实施方式涉及一种电子设备，包括至少一个处理器；以及，
与所述至少一个处理器通信连接的存储器；其中，
所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行上述第一、第二实施方式中任一所述的网络流量预测方法。
其中，存储器和处理器采用总线方式连接，总线可以包括任意数量的互联的总线和桥，总线将一个或多个处理器和存储器的各种电路连接在一起。总线还可以将诸如外围设备、稳压器和功率管理电路等之类的各种其他电路连接在一起，这些都是本领域所公知的，因此，本文不再对其进行进一步描述。总线接口在总线和收发机之间提供接口。收发机可以是一个元件，也可以是多个元件，比如多个接收器和发送器，提供用于在传输介质上与各种其他装置通信的单元。经处理器处理的数据通过天线在无线介质上进行传输，进一步，天线还接收数据并将数据传送给处理器。
处理器负责管理总线和通常的处理，还可以提供各种功能，包括定时，外围接口，电压调节、电源管理以及其他控制功能。而存储器可以被用于存储处理器在执行操作时所使用的数据。
本发明第六实施方式涉及一种计算机可读存储介质，存储有计算机程序。计算机程序被处理器执行时实现上述方法实施例。
即，本领域技术人员可以理解，实现上述实施例方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序存储在一个存储介质中，包括若干指令用以使得一个设备(可以是单片机，芯片等)或处理器(processor)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
本领域的普通技术人员可以理解，上述各实施方式是实现本发明的具体实施例，而在实际应用中，可以在形式上和细节上对其作各种改变，而不偏离本发明的精神和范围。

Claims (15)

  1. 一种网络流量预测方法，其特征在于，包括：
    建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型；
    将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取；
    将所述下级网络流量预测模型的计算模型代入所述上级网络流量预测模型中，得到完整的网络流量预测模型。
  2. 根据权利要求1所述的网络流量预测方法，其特征在于，所述上级网络流量预测模型包括：流量层模型；所述建立上级网络流量预测模型，具体包括：
    分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型；所述流量层模型包括：流量层线性模型和流量层误差补偿模型。
  3. 根据权利要求2所述的网络流量预测方法，其特征在于，所述分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型，具体包括：
    基于具体的业务特性，寻找业务变量；
    验证所述业务变量是否通过参数估计；
    若通过，则建立所述业务变量与所述流量层模型的流量值间的流量层线性模型；
    若未通过，则根据所述业务变量对流量值的影响大小和作用，设定所述流量层模型的流量层误差补偿模型；
    对建立的所述流量层线性模型和所述流量层误差补偿模型求和，得到所述流量层模型的计算模型。
  4. 根据权利要求2或3所述的网络流量预测方法，其特征在于，所述下级网络流量预测模型包括：业务层模型；所述自变量具体为所述业务变量；
    所述将所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取，具体包括：
    分析所述业务变量的业务特性以确定下一业务层模型的计算模型；
    判断当前分析的业务层模型是否完成业务层级的拆解；
    若未完成拆解，则返回执行分析所述业务变量的业务特性以确定下一业务层模型的计算 模型的步骤；
    若拆解完成，则退出循环流程。
  5. 根据权利要求4所述的网络流量预测方法，其特征在于，所述分析所述业务变量的业务特性以确定下一业务层模型的计算模型，具体包括：
    验证所述业务变量是否存在周期性和线性关系；
    若所述业务变量存在周期性或线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型；
    若所述业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。
  6. 根据权利要求5所述的网络流量预测方法，其特征在于，所述验证所述业务变量是否存在周期性和线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型，包括：
    验证所述业务变量是否存在周期性；
    若所述业务变量存在周期性，则建立所述业务变量与时间自变量之间的时间序列模型；
    若所述业务变量无周期性，则进一步验证所述业务变量与所述自变量之间是否存在线性关系；
    若存在线性关系，则建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型。
  7. 根据权利要求6所述的网络流量预测方法，其特征在于，所述验证所述业务变量是否存在周期性，包括：
    对所述业务变量进行纯随机性检验和平稳性检验。
  8. 一种网络流量预测装置，其特征在于，包括：
    上级流量预测模块，用于建立上级网络流量预测模型，所述上级网络流量预测模型包括：线性模型；
    下级流量预测模块，用于将所述上级流量预测模块建立的所述上级网络流量预测模型中所述线性模型的自变量作为因变量，建立至少一层下级网络流量预测模型，并对所述下级网络流量预测模型中的自变量进行拆解，直至每一所述下级网络流量预测模型的自变量均能自然获取；
    叠加模块，用于将所述下级流量预测模块建立的所述下级网络流量预测模型的计算模型代入所述上级流量预测模块建立的所述上级网络流量预测模型中，得到完整的网络流量预测 模型。
  9. 根据权利要求8所述的网络流量预测装置，其特征在于，所述上级网络流量预测模型包括：流量层模型；所述上级流量预测模块，具体为流量层预测模块；
    所述流量层预测模块用于：分析影响网络流量的业务变量，并根据所述业务变量建立流量层模型；所述流量层模型包括：流量层线性模型和流量层误差补偿模型。
  10. 根据权利要求9所述的网络流量预测装置，其特征在于，所述流量层预测模块具体用于：基于具体的业务特性，寻找业务变量；验证所述业务变量是否通过参数估计；若通过，则建立所述业务变量与所述流量层模型的流量值间的流量层线性模型；若未通过，则根据所述业务变量对流量值的影响大小和作用，设定所述流量层模型的流量层误差补偿模型。
  11. 根据权利要求9或10所述的网络流量预测装置，其特征在于，所述下级网络流量预测模型包括：业务层模型；所述下级流量预测模块，具体为业务层预测模块；所述自变量具体为所述业务变量；
    所述业务层预测模块用于：分析所述业务变量的业务特性以确定下一业务层模型的计算模型；判断当前分析的业务层模型是否完成业务层级自变量的拆解；若未完成拆解，则将当前分析的业务层模型中线性模型的自变量作为因变量，然后返回执行分析所述业务变量的业务特性以确定下一业务层模型的计算模型的操作；若拆解完成，则退出循环操作。
  12. 根据权利要求11所述的网络流量预测装置，所述业务层预测模块具体用于：
    验证所述业务变量是否存在周期性和线性关系；
    若所述业务变量存在周期性或线性关系，则建立所述业务变量与下一所述业务层模型自变量间的业务层线性模型；
    若所述业务变量不存在周期性且不存在线性关系，则对所述业务变量进行定性分析，并在存在定性关系时设定定性分析模型。
  13. 根据权利要求11所述的网络流量预测装置，所述业务层预测模块具体用于：
    验证所述业务变量是否存在周期性；
    若所述业务变量存在周期性，则建立所述业务变量与时间自变量之间的时间序列模型；
    若所述业务变量无周期性，则进一步验证所述业务变量与所述自变量之间是否存在线性关系；
    若存在线性关系，则建立所述业务变量与当前分析的所述业务层模型中自变量之间的线性模型。
  14. 一种电子设备，其特征在于，包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行如权利要求1至7中任一所述的网络流量预测方法。
  15. 一种计算机可读存储介质，存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至7中任一项所述的网络流量预测方法。
PCT/CN2018/119359 2018-11-14 2018-12-05 一种网络流量预测方法、装置及电子设备 Ceased WO2020098016A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP18940464.3A EP3879754A4 (en) 2018-11-14 2018-12-05 NETWORK TRAFFIC PREDICTION PROCESS, DEVICE AND ELECTRONIC DEVICE
US17/043,639 US11252044B2 (en) 2018-11-14 2018-12-05 Network traffic prediction method, apparatus, and electronic device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811354343.2A CN109586954B (zh) 2018-11-14 2018-11-14 一种网络流量预测方法、装置及电子设备
CN201811354343.2 2018-11-14

Publications (1)

Publication Number Publication Date
WO2020098016A1 true WO2020098016A1 (zh) 2020-05-22

Family

ID=65922515

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/119359 Ceased WO2020098016A1 (zh) 2018-11-14 2018-12-05 一种网络流量预测方法、装置及电子设备

Country Status (4)

Country Link
US (1) US11252044B2 (zh)
EP (1) EP3879754A4 (zh)
CN (1) CN109586954B (zh)
WO (1) WO2020098016A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113242156A (zh) * 2021-04-30 2021-08-10 新华三半导体技术有限公司 一种芯片验证方法、装置及系统
CN114629797A (zh) * 2022-03-11 2022-06-14 阿里巴巴（中国）有限公司 带宽预测方法、模型生成方法及设备
CN115037642A (zh) * 2022-03-30 2022-09-09 武汉烽火技术服务有限公司 一种识别流量瓶颈的方法和装置

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113572627B (zh) 2020-04-28 2022-09-02 华为技术有限公司 一种数据处理方法和数据处理装置
CN112819608A (zh) * 2021-02-05 2021-05-18 建信金融科技有限责任公司 一种基于多元回归与时间序列的区域信贷预测方法及系统
CN112994978B (zh) * 2021-02-25 2023-01-24 网宿科技股份有限公司 一种网络流量监测方法及装置
CN115348179B (zh) * 2022-08-10 2024-02-02 中国电信股份有限公司 网络流量边界预测方法、装置、设备及介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105376097A (zh) * 2015-11-30 2016-03-02 沈阳工业大学 网络流量的一种混合预测方法
CN105634787A (zh) * 2014-11-26 2016-06-01 华为技术有限公司 网络关键指标的评估方法、预测方法及装置和系统
US20160352584A1 (en) * 2013-09-26 2016-12-01 Amazon Technologies, Inc. Mixture model approach for network forecasting
CN107171848A (zh) * 2017-05-27 2017-09-15 华为技术有限公司 一种流量预测方法和装置

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7774440B1 (en) * 2001-07-25 2010-08-10 Scalable Network Technologies, Inc. Method and system for enhancing performance of a physical network under real-time control using simulation of a reference model
WO2011160002A2 (en) * 2010-06-17 2011-12-22 Velocent Systems Incorporated Determining an average effective data throughput as corresponds to a network-served end user
CN102916851B (zh) * 2012-09-24 2016-03-23 浪潮(北京)电子信息产业有限公司 一种网络流量预测方法及装置
US9204319B2 (en) * 2014-04-08 2015-12-01 Cellco Partnership Estimating long term evolution network capacity and performance
US10318882B2 (en) * 2014-09-11 2019-06-11 Amazon Technologies, Inc. Optimized training of linear machine learning models
CN105787136A (zh) * 2016-05-24 2016-07-20 北京京东尚科信息技术有限公司 数据挖掘方法和装置
CN106357456B (zh) * 2016-10-11 2019-05-21 广东工业大学 一种网络流量的预测方法及装置
CN108737455B (zh) * 2017-04-13 2020-12-25 国家电网有限公司 网络业务识别装置及方法
CN107026763B (zh) * 2017-06-02 2019-11-26 广东电网有限责任公司中山供电局 一种基于流量分解的数据通信网流量预测方法
CN108733970A (zh) * 2018-05-16 2018-11-02 常州大学 一种基于QSAR/QEcoSAR方法联用预测有机磷阻燃剂对斑马鱼急性毒性的方法
CN108768877B (zh) * 2018-07-20 2021-06-22 网宿科技股份有限公司 一种突发流量的分配方法、装置及代理服务器
US11134016B2 (en) * 2018-10-26 2021-09-28 Hughes Network Systems, Llc Monitoring a communication network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160352584A1 (en) * 2013-09-26 2016-12-01 Amazon Technologies, Inc. Mixture model approach for network forecasting
CN105634787A (zh) * 2014-11-26 2016-06-01 华为技术有限公司 网络关键指标的评估方法、预测方法及装置和系统
CN105376097A (zh) * 2015-11-30 2016-03-02 沈阳工业大学 网络流量的一种混合预测方法
CN107171848A (zh) * 2017-05-27 2017-09-15 华为技术有限公司 一种流量预测方法和装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3879754A4 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113242156A (zh) * 2021-04-30 2021-08-10 新华三半导体技术有限公司 一种芯片验证方法、装置及系统
CN113242156B (zh) * 2021-04-30 2022-05-27 新华三半导体技术有限公司 一种芯片验证方法、装置及系统
CN114629797A (zh) * 2022-03-11 2022-06-14 阿里巴巴（中国）有限公司 带宽预测方法、模型生成方法及设备
CN114629797B (zh) * 2022-03-11 2024-03-08 阿里巴巴（中国）有限公司 带宽预测方法、模型生成方法及设备
CN115037642A (zh) * 2022-03-30 2022-09-09 武汉烽火技术服务有限公司 一种识别流量瓶颈的方法和装置
CN115037642B (zh) * 2022-03-30 2023-11-21 武汉烽火技术服务有限公司 一种识别流量瓶颈的方法和装置

Also Published As

Publication number Publication date
US20210021482A1 (en) 2021-01-21
EP3879754A4 (en) 2021-12-15
CN109586954A (zh) 2019-04-05
CN109586954B (zh) 2020-09-18
EP3879754A1 (en) 2021-09-15
US11252044B2 (en) 2022-02-15

Similar Documents

Publication Publication Date Title
WO2020098016A1 (zh) 一种网络流量预测方法、装置及电子设备
CN1987922B (zh) 实际产品it系统的采用推测的在线性能建模方法和系统
US7881920B2 (en) Systemic enterprise management method and apparatus
US7783468B2 (en) Automated system and method for service and cost architecture modeling of enterprise systems
CN114430826A (zh) 用于预测计算工作负载的时间序列分析
US20180253728A1 (en) Optimizing fraud analytics selection
US11810022B2 (en) Contact center call volume prediction
CN105786681A (zh) 数据中心的服务器性能评估及服务器更新方法
CN117934135A (zh) 一种网点运营管理方法、装置、电子设备及存储介质
US20180315502A1 (en) Method and apparatus for optimization and simulation of patient flow
Morichetta et al. Demystifying deep learning in predictive monitoring for cloud-native SLOs
Tuli et al. SimTune: Bridging the simulator reality gap for resource management in edge-cloud computing
CN119576584A (zh) 基于多目标优化的边缘计算资源分配方法及系统
Baharvand et al. Optimizing service function chaining and scheduling with deep Q-learning for enhanced quality of experience in edge networks
US12229219B2 (en) Dynamic evaluation of model acceptability
CN119721341A (zh) 基于元学习与注意力机制的光伏功率预测方法及设备
CN114844889A (zh) 视频处理模型的更新方法、装置、电子设备及存储介质
CN114237856A (zh) 运算类型识别方法、装置、电子设备及存储介质
US20150100391A1 (en) Systems And Methods For Discovering An Optimal Operational Strategy For A Desired Service Delivery Outcome
Xiao et al. Optimal software testing-resource allocation with operational profile: computational aspects
He et al. Predicting effect and cost of microservice system evolution using graph neural network
Huo et al. Joint Prediction and Matching for Computing Resource Exchange Platforms
Rista et al. Simulation-Based Parameter Optimization for Self-adaptive HPL on Parallel Systems
Singh et al. Analytical and Computational Study of Bi-level Vacation Policy for Transient Markovian Queue with State-Dependent Rates, Waiting Server, and Discouragement
CN115238147B (zh) 软件服务推荐方法、装置、设备及介质

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: 18940464

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2018940464

Country of ref document: EP

Effective date: 20210609