CN112532429B - Multivariable QoS prediction method based on position information - Google Patents

Multivariable QoS prediction method based on position information Download PDF

Info

Publication number
CN112532429B
CN112532429B CN202011256189.2A CN202011256189A CN112532429B CN 112532429 B CN112532429 B CN 112532429B CN 202011256189 A CN202011256189 A CN 202011256189A CN 112532429 B CN112532429 B CN 112532429B
Authority
CN
China
Prior art keywords
qos
data
model
value
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011256189.2A
Other languages
Chinese (zh)
Other versions
CN112532429A (en
Inventor
刘博�
王名亮
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.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
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 Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN202011256189.2A priority Critical patent/CN112532429B/en
Publication of CN112532429A publication Critical patent/CN112532429A/en
Application granted granted Critical
Publication of CN112532429B publication Critical patent/CN112532429B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • 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

Abstract

A multivariable QoS prediction method based on position information belongs to the field of service calculation. The method acquires historical QoS historical data, adopts a collaborative filtering algorithm to process missing values of the historical data, and introduces position information of a user in the similarity calculation process to enable the filled missing values to be closer to actual values. And constructing the filled data into a moving sliding window data set, dividing the data set into training data and test data, training the multivariate LSTM model by using the training data, and finally testing the prediction result by using the test data. The present invention uses a collaborative filtering algorithm to fill missing values and uses multivariate LSTM to predict QoS. The position information of the user is introduced in the process of filling the missing value, so that the filled missing value is closer to the real QoS value. Meanwhile, compared with single QoS attribute prediction, the multivariate prediction model can effectively reduce error accumulation so as to improve prediction accuracy.

Description

Multivariable QoS prediction method based on position information
Technical Field
The invention belongs to the field of service calculation, and is mainly used for establishing a specific QoS attribute prediction model.
Background
Service Oriented Architecture (SOA) is a modern architectural paradigm that connects different functional units of an application by defining well-defined interfaces and contracts between services. Web services are the standard of implementation for SOAs. As Web services with similar functions are increasingly emerging on networks, quality of service (QoS) is gradually attracting attention and is applied in service selection, service composition and service scheduling processes.
QoS refers to non-functional attributes of a Web service such as throughput, response time, and call success rate. The most direct method of obtaining a QoS value of a Web service is by invoking a target service, but the uncertainty of the network environment and the difference of the user geographical location make it possible to observe completely different QoS values even if the same Web service is invoked. Moreover, such enumerated calls would incur significant costs to the user and the service. Therefore, the prediction of the QoS value becomes crucial in the service selection process. Prediction of QoS is important because if the QoS of the current or future is known, one can apply it to the service selection process. Currently, many scholars are studying QoS prediction of Web services. These studies can be roughly divided into two categories, one based on data-driven methods and the other based on temporal perception. Data-driven methods generally treat the problem as a filling problem of a missing value and are suitable for prediction in a static environment, and such methods mainly include a collaborative filtering method, a location-aware method, and a context-based method. The time-aware method generally models the prediction of QoS as a time-series prediction problem, and uses a time-series prediction model to predict the prediction problem, but this method rarely considers the influence of the relevant information of the deeper level user and service on the QoS value.
Disclosure of Invention
The invention combines the collaborative filtering method and the time perception method, and fully utilizes the advantages of the user position information and the multivariate time sequence to improve the prediction precision.
The prediction of the QoS has a cold start problem, the root of the cold start problem is the sparsity of QoS data, and most of calling records of the QoS are lost, so that the accuracy of the algorithm in time sequence prediction is low. So first we consider the padding problem of data, and common padding methods for multivariate timing problem include three categories: the first is direct deletion, but this approach may discard some important information of the data; the second type is a filling method based on statistics, such as mean filling, median filling and the like, which ignores the time sequence information of data; the third category is machine learning based methods such as KNN (K-Nearest Neighbor) and matrix decomposition. In consideration of the influence of the position information of the user and the service on the QoS, the position information is added to the collaborative filtering system to solve the filling problem of the data. After the data filling is completed, the invention adopts a multivariable time series model to predict the specific QoS attribute.
The prediction scheme adopted in the prediction stage of the invention is a multivariable time series prediction method based on LSTM, which comprises the following steps:
step 1, obtaining and preprocessing a QoS attribute to be predicted and a historical QoS record of related attributes thereof, and constructing the QoS attribute to be predicted into a time sequence containing a plurality of QoS attributes.
And 2, dividing the data into training data and testing data.
And 3, setting the hyper-parameters of the multivariate LSTM prediction model, including the size T of a sliding window, the time step number, the number of input layer nodes, the number of hidden layer layers, the number of hidden layer nodes and the number of output layer nodes.
Step 4, fitting a multivariate LSTM model on a training set, predicting in a test data set, and evaluating the error of the model:
and 5, finally, taking the preprocessed QoS data as input data of the model, learning the input data through the model, and outputting to obtain a specified QoS value.
1. A multivariable QoS prediction method based on position information is characterized in that: the method comprises the following steps:
step 1, acquiring historical QoS data, preprocessing historical QoS records, and filling missing values in the data:
1) Searching a similar user set S, and dividing z users which are in the same country, are coded by an ASN and are closest to a user to be predicted into similar users by using user information;
2) In the same moment, similar users call the average value of all QoS records of the service to be predicted as a missing value to be filled in the QoS records;
step 2, dividing the preprocessed data into training data and test data;
step 3, building an LSTM model on the multivariate input data, and setting parameters: learning rate η =0.001, sliding time window T =4, training time epoch =100, number of hidden neurons =30, number of output neurons =1, batch size =16, and parameter update optimizer Adam; (ii) a
Step 4, fitting a multivariable LSTM model on a training set, predicting in a test data set, and evaluating the error of the model;
and 5, predicting the target to be predicted.
2. Further, the implementation process of step 1 is as follows:
acquiring historical QoS data, preprocessing the historical QoS records, and filling missing values in the data, wherein the method specifically comprises the following steps:
given a user set U = { U } consisting of m users 1 ,u 2 ,u 3 ,...,u m-1 ,u m Service set S = { S } consisting of n services 1 ,s 2 ,s 3 ,...,s n-1 ,s n And f, qoS records in t time intervals are recorded, and the data to be filled in the kth time interval (k is more than or equal to 1 and less than or equal to t) is assumed as the user u i (1. Ltoreq. I. Ltoreq.m) calling service s j Generated record
Figure BDA0002773166780000032
J is more than or equal to 1 and less than or equal to n, and the filling process is as follows:
firstly, screening and user u from QoS information acquisition system i The users with the same country information and ASN code obtain the distance user u from the user by using the great circle distance i The nearest z users get a set of similar users Neig i The great circle distance calculation formula is as follows:
Figure BDA0002773166780000031
wherein r represents the earth radius, # i ,ψ j Respectively represent users u i And u j The latitude of (a), Δ ψ represents the absolute value of the difference in the latitudes of two users, and Δ λ is the absolute value of the difference in the longitudes thereof;
similar user set Neig in k time interval i All users in (1) invoke service s j The average of the recordings was taken as the missing recording
Figure BDA0002773166780000033
The value of (c).
3. Further, the sample obtained in step 2 is divided into training data and testing data, the training data is used for training the model, and the testing data is used for testing the effect of the model.
4. Further, in step 4, fitting a multivariate LSTM model on a training set, predicting in a test data set, and evaluating the error of the model;
assume that the input is the sequence x = (x) 1 ,x 2 ,...,x t ) User u i Each time interval response time from the first time interval to the t-th time interval represents QoS record under t different time moments, and the hidden layer vector sequence is h = (h =) 1 ,h 2 ,...,h t ) (ii) a The specific process of forward propagation of LSTM is:
the invention is characterized in that:
(1) The invention introduces the position information of the user in the process of processing the missing value, so that the filled missing data has higher interpretability and is closer to the actual data.
(2) The method adopts the multivariate time series model to predict the time series model, compared with the prior method, the method more fully releases the potential of data, predicts a certain characteristic by utilizing a plurality of characteristics, and can effectively improve the prediction precision.
Drawings
Fig. 1 is a flow chart of QoS record missing value padding.
FIG. 2 is a schematic view of the structure of a multi-element LSTM.
Fig. 3 is an internal structure view of the LSTM.
Fig. 4 is a flow chart of QoS prediction proposed by the present invention.
FIG. 5 shows the results of each comparative experiment (where IPCC, UPCC, UIPCC are conventional collaborative filtering methods, LSTM-1 uses univariate prediction, LSTM-2 uses univariate prediction, but geographical location information is introduced during the filling process, and LSTM-3 is the proposed method of the present invention).
Detailed Description
The following is a detailed description of the invention with reference to examples and drawings.
The invention uses a PC, the experimental environment is Python3.6 version, keras is adopted to define LSTM, and GPU with enough computing power is needed to accelerate training. The response time is predicted in this example using two variables, response time and throughput. As shown in fig. 1, the multivariate timing prediction method based on location information provided by the present invention comprises the following specific steps:
step 1, acquiring historical records of response time and throughput, preprocessing the historical records, and constructing the historical records into a data set containing 2 time sequences;
the acquired historical data is from a QoS data acquisition system, the data generally comprises response time and throughput generated by a user calling a specific service at different time intervals, and the calling success rate. Before the time series prediction model is used, it needs to be filled. U = { U } for a given 100 users 1 ,u 2 ,u 3 ,...,u 99 ,u 100 150 services S = { S = } 1 ,s 2 ,s 3 ,...,s 149 ,s 150 Record of QoS in past 64 time slices (15 minutes each), assuming that the record to be filled in the k-th time interval is user u i Invoking service generated records
Figure BDA0002773166780000041
The filling process comprises the following steps:
first screening and user u i The country information and ASN code of the user are the same, and then the user u is obtained from the user by using the great circle distance in a centralized way i The last 5 users get a similar user set Neig i The great circle distance calculation formula is as follows:
Figure BDA0002773166780000051
wherein r representsRadius of the earth, # i ,ψ j Respectively represent users u i And u j The latitude of (a) and (b) indicate the absolute value of the difference between the latitudes of two users, and (a) is the absolute value of the difference between the longitudes thereof.
Similar user set Neig in k time interval i All users in (1) invoke service s j The average of the recordings was taken as the missing recording
Figure BDA0002773166780000052
The value of (c).
The padded data is constructed to obtain a supervised data set containing 2 QoS attributes (response time, throughput). Building input data
Figure BDA0002773166780000053
Outputting the data
Figure BDA0002773166780000054
The output data is our target prediction data.
Step 2, dividing the processed QoS data set into training data and testing data:
and (2) dividing the sample processed in the step (1) into training data and testing data, wherein the training data is used for training the model, and the testing data is used for testing the effect of the model, and the training data accounts for 75%.
And 3, setting parameters of the multivariate LSTM prediction model. (different parameter settings will yield different results, in this example learning rate η =0.001, sliding time window T =4, training epoch =100, number of cryptic neurons =30, number of output layer neurons =1, batch _size =16, parameter update optimizer Adam).
And 4, fitting a multivariate LSTM model on the training set, predicting in the test data set and evaluating the error of the model, wherein the specific steps are as follows:
FIG. 2 is a schematic diagram of the architecture of a multivariate LSTM predictive model. As shown in fig. 2, the model is composed of an input layer, hidden layers and an output layer, wherein the hidden layer units are memory blocks specific to the LSTM prediction model rather than neuron nodes of a general neural network, and fig. 3 shows the internal structure of a memory block Cell.
Assume that the input is the sequence x = (x) 1 ,x 2 ,...,x t ) The QoS records at t different time points are represented, and the hidden vector sequence is h = (h) 1 ,h 2 ,...,h t ) (ii) a The specific process of forward propagation is:
f t ＝σ(W f [h t-1 ,x t ]+b f )
i t ＝σ(W i [h t-1 ,x t ]+b i )
Figure BDA0002773166780000061
Figure BDA0002773166780000062
o t ＝σ(W o [h t-1 ,x t ]+b o )
h t ＝o t *tanh(C t )
wherein i t 、f t 、o t Respectively representing the output of the input gate, the forgetting gate and the output gate updated at the time t. The input gate indicates whether the collected QoS record is allowed to be added to the current hidden layer node, if 1 (gate open), the input is allowed, if 0 (gate closed), the input is not allowed, and thus some non-available input information is filtered. And the forgetting gate represents whether to keep the historical QoS data stored by the current hidden layer node, if so, the current information is kept, and otherwise, the historical QoS record stored by the current node is emptied. The output gate represents whether the current node output is transmitted to the next layer (the next hidden layer or the output layer), if 1 (gate open), the output value of the current node is acted on the next layer, otherwise, the output value of the current node is not transmitted. W is a group of f ,W i ,W C ,W o Representing the weight matrices of the forgetting gate, the input gate, the state of the computing unit and the output gate respectively,[h t-1 ,x t ]representing the concatenation of two vectors into a longer vector, b f ,b i ,b C ,b o Then the bias terms representing the forgetting gate, the input gate, the state of the computational cell, and the output gate, respectively. C t The state of the cell is calculated by adding the information required to be memorized at the last moment and the information required to be memorized at the current moment.
Figure BDA0002773166780000063
As old cell state C t-1 Transition to C t Represents new information. In the above formula, tanh and σ are both activation functions, σ is a sigmoid function, and is applied to the input gate, the output gate, and the forgetting gate, and tanh function is a function when candidate memory is generated, and both functions are defined as:
Figure BDA0002773166780000064
Figure BDA0002773166780000065
the LSTM training algorithm is a back propagation algorithm and mainly comprises three steps:
1) Calculating the output value of each neuron in the forward direction, i.e. f for LSTM t ,i t ,c t ,h t ,o t 5 sets of vectors.
2) The error term for each neuron is calculated in reverse. Like the recurrent neural network, the back propagation of the LSTM error term also includes two directions: one is the backward propagation along the time, namely, the error term at each moment is calculated from the current moment t; one is to propagate the error term up one layer.
3) The gradient of each weight is calculated according to the corresponding error term.
There are 8 sets of parameters to be learned by LSTM, which are: w f ,W i ,W C ,W o And b f ,b i ,b C ,b o . The invention uses under gradientThe descent algorithm adjusts parameters in the model to minimize the loss function, and the gradient used for gradient descent can be calculated using a back propagation algorithm or an automatic differentiation tool.
The objective function in the training process is the square error between the predicted value and the true value, and is defined as:
Figure BDA0002773166780000071
wherein r is i,j Representing user u i Invoking a service s j The actual QoS value that is generated is,
Figure BDA0002773166780000072
indicating the predicted value of QoS. By minimizing the objective function, the various parameters of the model of the sphere can be minimized.
And measuring the prediction error of the trained model by using two indexes of MSE and RMSE, wherein the calculation formulas of MSE and RMSE are as follows:
Figure BDA0002773166780000073
Figure BDA0002773166780000074
where N represents the number of QoS predictors.
And 5, predicting the target to be predicted on the model. The method comprises the following specific steps:
and finally, taking the trained, verified and tested multivariable prediction model as a QoS prediction model. And processing the data obtained by the QoS acquisition system to be used as the input of the model, learning the input data through the model, and finally outputting to obtain the response time of the 65 th time slice.
As shown in fig. 5, the experimental results of this example show that compared with the conventional collaborative filtering algorithm or the univariate time series prediction method, the error of the present invention is significantly lower than that of other methods, and better performance is obtained, thus proving the effectiveness of the method.
The above embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and the scope of the present invention is defined by the claims. Various modifications and equivalents of the invention may be made by those skilled in the art within the spirit and scope of the invention, and such modifications and equivalents should also be considered as falling within the scope of the invention.

Claims (3)

1. A multivariable QoS prediction method based on position information is characterized in that: the method comprises the following steps:
step 1, acquiring historical QoS data, preprocessing historical QoS records, and filling missing values in the data:
1) Searching a similar user set S, and dividing z users which are in the same country, are coded by an ASN and are closest to a user to be predicted into similar users by using user information;
2) In the same moment, similar users call the average value of all QoS records of the service to be predicted as a missing value to be filled in the QoS records;
step 2, dividing the preprocessed data into training data and test data;
step 3, building an LSTM model on the multivariate input data, and setting parameters: learning rate η =0.001, sliding time window T =4, training time epoch =100, number of hidden neurons =30, number of output neurons =1, batch size =16, and parameter update optimizer Adam;
step 4, fitting a multivariable LSTM model on a training set, predicting in a test data set, and evaluating the error of the model;
step 5, predicting the target to be predicted;
the implementation process of the step 1 is as follows:
acquiring historical QoS data, preprocessing the historical QoS records, and filling missing values in the data, wherein the method specifically comprises the following steps:
given a user set U = { U } consisting of m users 1 ,u 2 ,u 3 ,...,u m-1 ,u m Service set S = { S } consisting of n services 1 ,s 2 ,s 3 ,...,s n-1 ,s n And f, qoS records in t time intervals are recorded, and the data to be filled in the kth time interval (k is more than or equal to 1 and less than or equal to t) is assumed as the user u i (1. Ltoreq. I. Ltoreq.m) calling service s j Generated record
Figure FDA0003964632540000011
The filling process comprises the following steps:
firstly, screening and user u from QoS information acquisition system i The users with the same country information and ASN code obtain the distance user u from the user by using the great circle distance i The nearest z users get a set of similar users Neig i The great circle distance calculation formula is as follows:
Figure FDA0003964632540000021
wherein r represents the earth radius, # i ,ψ j Respectively represent users u i And u j The latitude of (a), Δ ψ represents the absolute value of the difference in the latitudes of two users, and Δ λ is the absolute value of the difference in the longitudes thereof;
similar user set Neig in k time interval i All users in (1) invoke service s j The average of the recordings was taken as the missing recording
Figure FDA0003964632540000022
A value of (d);
step 4, fitting a multivariable LSTM model on a training set, predicting in a test data set, and evaluating a model error;
assume that the input is the sequence x = (x) 1 ,x 2 ,...,x t ) User u i Each time interval response time from the first time interval to the t-th time interval represents QoS record under t different time moments, and the hidden layer vector sequence is h = (h =) 1 ,h 2 ,...,h t ) (ii) a The specific process of forward propagation of LSTM is:
f t ＝σ(W f [h t-1 ,x t ]+b f )
i t ＝σ(W i [h t-1 ,x t ]+b i )
Figure FDA0003964632540000023
Figure FDA0003964632540000024
o t ＝σ(W o [h t-1 ,x t ]+b o )
h t ＝o t *tanh(C t )
wherein i t 、f t 、o t Respectively representing the output of the input gate, the forgetting gate and the output gate updated at the moment t; the input gate indicates whether the collected QoS records are allowed to be added into the current hidden layer node, if the collected QoS records are 1, input is allowed, and if the collected QoS records are 0, input is not allowed, so that some input information which is not available is filtered; the forgetting gate shows whether historical QoS data stored by the current hidden layer node is reserved or not, if so, current information is reserved, and otherwise, historical QoS records stored by the current node are emptied; the output gate indicates whether to transmit the output of the current node to the next layer, the next layer is a next hidden layer or an output layer, if the output value is 1, the output value of the current node is acted on the next layer, otherwise, the output value of the current node is not transmitted; w is a group of f ,W i ,W C ,W o Weight matrices representing the states of the forgetting gate, the input gate, the calculation unit and the output gate, respectively, [ h ] t-1 ,x t ]Is that two vectors are connected into one longer vector, b f ,b i ,b C ,b o Then respectively represent the bias items of a forgetting gate, an input gate, a computing unit state and an output gate; c t For calculating the state of the cell, the information needed to be memorized at the last moment is added with the information needed to be memorized at the current moment(ii) the information recalled; c ^ e t As old cell state C t-1 Transition to C t A candidate value vector of (a), representing new information; in the above formula, tanh and σ are both activation functions, σ is a sigmoid function, and is applied to the input gate, the output gate, and the forgetting gate, and tanh function is a function when candidate memory is generated, and both functions are defined as:
Figure FDA0003964632540000031
Figure FDA0003964632540000032
the training algorithm of the LSTM is a back propagation algorithm, and comprises three steps:
1) The output value of each neuron is calculated forward, i.e. f for LSTM t ,i t ,C t ,h t ,o t 5 sets of vectors;
2) Reversely calculating an error term of each neuron; like the recurrent neural network, the back propagation of the LSTM error term also includes two directions: one is the backward propagation along the time, namely, the error term of each moment is calculated from the current moment t; one is to propagate the error term up one layer;
3) Calculating a gradient for each weight based on the corresponding error term;
there are 8 sets of parameters to be learned by LSTM, which are: w f ,W i ,W C ,W o And b f ,b i ,b C ,b o (ii) a Parameters in the model are adjusted by using a gradient descent algorithm to minimize a loss function, and the gradient used by the gradient descent can be calculated by using a back propagation algorithm or an automatic differential tool;
the objective function in the training process is the square error between the predicted value and the true value, and is defined as:
Figure FDA0003964632540000041
wherein r is i,j Representing user u i Invoking a service s j The actual QoS value that is generated is,
Figure FDA0003964632540000042
a prediction value representing QoS; the parameters of the model are found by minimizing the objective function.
2. The method of claim 1, wherein the method comprises: and (3) dividing the sample obtained in the step (2) into training data and testing data, wherein the training data is used for training the model, and the testing data is used for testing the effect of the model.
3. The method of claim 1, wherein the method comprises: the implementation of step 5 is as follows,
finally, the multivariable prediction model after training, verification and test is used as a QoS prediction model; and taking the preprocessed QoS data as the input of the model, learning the input data through the model, and finally outputting to obtain the QoS record of the specified time interval.
CN202011256189.2A 2020-11-11 2020-11-11 Multivariable QoS prediction method based on position information Active CN112532429B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011256189.2A CN112532429B (en) 2020-11-11 2020-11-11 Multivariable QoS prediction method based on position information

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011256189.2A CN112532429B (en) 2020-11-11 2020-11-11 Multivariable QoS prediction method based on position information

Publications (2)

Publication Number Publication Date
CN112532429A CN112532429A (en) 2021-03-19
CN112532429B true CN112532429B (en) 2023-01-31

Family

ID=74981559

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011256189.2A Active CN112532429B (en) 2020-11-11 2020-11-11 Multivariable QoS prediction method based on position information

Country Status (1)

Country Link
CN (1) CN112532429B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116132347B (en) * 2023-04-06 2023-06-27 湖南工商大学 Bi-LSTM-based service QoS prediction method in computing network convergence environment

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
CN111131424A (en) * 2019-12-18 2020-05-08 武汉大学 Service quality prediction method based on combination of EMD and multivariate LSTM

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10855550B2 (en) * 2016-11-16 2020-12-01 Cisco Technology, Inc. Network traffic prediction using long short term memory neural networks
CN108734338A (en) * 2018-04-24 2018-11-02 阿里巴巴集团控股有限公司 Credit risk forecast method and device based on LSTM models

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
CN111131424A (en) * 2019-12-18 2020-05-08 武汉大学 Service quality prediction method based on combination of EMD and multivariate LSTM

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
S1-010399 "Missing LCS QoS, Priority, Request type, Assistance data, Client type, Stop reporting type parameters over Iu interface RANAP 25.413 (LOCATION REPORTING CONTROL and LOCATION REPORT messages).";TSG-SA WG2;《3GPP tsg_sa\WG1_Serv》;20010509;全文 *
一种基于矩阵填充的Web服务QoS预测算法;李刚等;《舰船电子工程》;20160120(第01期);全文 *
基于用户位置与反向预测的QoS预测方法;王珂等;《计算机工程与设计》;20170216(第02期);全文 *

Also Published As

Publication number Publication date
CN112532429A (en) 2021-03-19

Similar Documents

Publication Publication Date Title
CN110647900B (en) Intelligent safety situation prediction method, device and system based on deep neural network
CN113723007B (en) Equipment residual life prediction method based on DRSN and sparrow search optimization
CN110083125B (en) Machine tool thermal error modeling method based on deep learning
CN113905391A (en) Ensemble learning network traffic prediction method, system, device, terminal, and medium
CN112380773A (en) Fuel cell degradation prediction method, device and equipment based on transfer learning
CN112532429B (en) Multivariable QoS prediction method based on position information
CN113378939A (en) Structure digital twin modeling and parameter identification method based on physical driving neural network
CN112734002A (en) Service life prediction method based on data layer and model layer joint transfer learning
CN114694379B (en) Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN112990585A (en) Hen laying rate prediction method based on LSTM-Kalman model
CN113947133A (en) Task importance perception element learning method for small sample image recognition
CN112948743B (en) Coal mine gas concentration deficiency value filling method based on space-time fusion
CN112215421A (en) Deep learning water quality index prediction method based on generation countermeasure network
CN117494871A (en) Ship track prediction method considering ship interaction influence
CN111860975A (en) Rainfall approaching prediction method based on generation countermeasure network
CN116865251A (en) Short-term load probability prediction method and system
CN116192669A (en) Network flow prediction method based on dynamic space-time diagram convolution
CN111277710A (en) Telephone traffic prediction method, system, computer device and storage medium
CN116522747A (en) Two-stage optimized extrusion casting process parameter optimization design method
CN110009132A (en) A kind of short-term electric load fining prediction technique based on LSTM deep neural network
Yakushin et al. Neural network model for forecasting statistics of communities of social networks
CN114997464A (en) Popularity prediction method based on graph time sequence information learning
CN114726751A (en) Intelligent early warning method, system, equipment and storage medium for resource quality monitoring
CN113627594A (en) One-dimensional time sequence data amplification method based on WGAN
CN114398819A (en) Method and system for predicting effective wave height of unstructured grid based on deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant