CN113259163B - Web service quality prediction method and system based on network topology perception - Google Patents
Web service quality prediction method and system based on network topology perception Download PDFInfo
- Publication number
- CN113259163B CN113259163B CN202110534218.5A CN202110534218A CN113259163B CN 113259163 B CN113259163 B CN 113259163B CN 202110534218 A CN202110534218 A CN 202110534218A CN 113259163 B CN113259163 B CN 113259163B
- Authority
- CN
- China
- Prior art keywords
- historical
- service quality
- target
- feature vector
- server
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000008447 perception Effects 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000013528 artificial neural network Methods 0.000 claims abstract description 16
- 239000013598 vector Substances 0.000 claims description 151
- 230000006870 function Effects 0.000 claims description 22
- 230000004044 response Effects 0.000 claims description 14
- 238000013507 mapping Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 11
- 230000003993 interaction Effects 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 9
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 230000002452 interceptive effect Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 5
- 230000004931 aggregating effect Effects 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 10
- 230000008569 process Effects 0.000 abstract description 9
- 238000004364 calculation method Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
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
- H04L41/147—Network analysis or design for predicting network behaviour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention relates to a Web service quality prediction method and a system based on network topology perception, belonging to the field of service calculation, wherein the Web service quality prediction method comprises the following steps: acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes; training the topology perception neural network according to the sample data to obtain a service quality prediction model; and based on the service quality prediction model, obtaining a target service quality prediction value according to the target user, the target server and a target shortest path corresponding to the target user and the target server. The influence of the communication path in the Web service calling process is considered when the service quality is predicted, and the prediction precision of the service quality is improved.
Description
Technical Field
The invention relates to the field of service calculation, in particular to a Web service quality prediction method and a Web service quality prediction system based on network topology perception.
Background
With the increasing popularization of World Wide Web (Web) service application, more and more service resources capable of being accessed publicly appear on the internet, and particularly under the technical drive of cloud computing, internet of things, service-oriented architecture and the like, a service ecosystem becomes mature. With the aggregation of service resources, how to effectively recommend appropriate service resources for application developers becomes an important research topic.
The traditional service recommendation mainly considers the matching degree of service functions and user requirements, and then recommends candidate service resources meeting specific function requirements, but as the number of homogeneous services increases, the quality of the service needs to be further distinguished from the aspect of non-functional depiction, and a refined recommendation result can be provided. Quality of service (QoS) generally characterizes non-functional characteristics of Web Services, such as performance, reliability, availability, and security, and is the key point for homogeneous service competition. Therefore, in the case of cloud application developers, selecting high quality services is critical to ensure their market competitiveness.
Due to the large scale of Web services, constrained by time, economic cost, and other factors, it is unlikely that a service provider will deploy a large number of software sensors in a network to monitor all quality of service information. Therefore, researchers have proposed a solution for service quality prediction, and by using the idea of collaborative filtering for reference, unknown QoS values are predicted by analyzing service invocation history and using intelligent collaborative filtering. Subject to data conditions, current research efforts focus primarily on easily quantifiable quality of service attributes such as response time, throughput, and reliability.
However, the existing service quality prediction technology only depends on users, services and the context environment where the users and the services are located, neglects the influence of communication paths, and reduces the prediction accuracy. As far as the service invocation procedure is concerned, it not only involves the client host and the server host, but also is closely related to the internet communication path. In the process, both user requests and service responses depend on complex interaction among a plurality of autonomous domain systems in the network topology, so that routing forwarding and communication of data are completed. It can be said that the difference in communication paths determines the quality of service to some extent. This also makes it difficult for the conventional prediction method neglecting the influence of the communication path to obtain a highly accurate prediction result. However, there is no effective solution for how to consider communication paths based on network topology to achieve accurate QoS prediction.
Disclosure of Invention
The invention aims to provide a Web service quality prediction method and a Web service quality prediction system based on network topology perception, which can improve the prediction precision of service quality.
In order to achieve the purpose, the invention provides the following scheme:
a world wide Web Web service quality prediction method based on network topology awareness comprises the following steps:
acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes; the historical quality of service predictor comprises response time and throughput;
training the topology perception neural network according to the sample data to obtain a service quality prediction model;
based on the service quality prediction model, obtaining a target service quality prediction value according to a target user, a target server and a target shortest path corresponding to the target user and the target server; the target shortest path comprises a target user, a target server and an intermediate autonomous domain node; the target quality of service predictor includes a response time and a throughput.
Optionally, the method for calculating the historical shortest path includes:
and aiming at each historical user and the corresponding historical server, determining the historical shortest path between the historical user and the corresponding historical server by adopting a shortest path algorithm based on a network topological graph according to the historical user and the corresponding historical server.
Optionally, the training the topology-aware neural network according to the sample data to obtain a service quality prediction model specifically includes:
mapping the historical user into a user embedded feature vector, mapping the historical server into a server embedded feature vector, and mapping each intermediate autonomous domain node into a corresponding embedded feature vector;
obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector;
determining a path characteristic vector according to the embedded characteristic vectors corresponding to the nodes of the intermediate autonomous domain;
determining a comprehensive characteristic vector according to the path characteristic vector and the cross characteristic vector;
and determining a service quality predicted value by adopting a multilayer full-connection network according to the comprehensive characteristic vector.
Optionally, the obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector specifically includes:
carrying out high-dimensional interaction on the user embedded characteristic vector and the server embedded characteristic vector through an outer product to obtain an interaction characteristic vector set;
and extracting features by adopting a convolutional neural network according to the interactive feature vector set to obtain cross feature vectors.
Optionally, the determining the path feature vector according to the embedded feature vector corresponding to each intermediate autonomous domain node specifically includes:
and based on the bidirectional long-short term memory network, aggregating the embedded characteristic vectors corresponding to the nodes of each intermediate autonomous domain to obtain corresponding path characteristic vectors.
Optionally, the obtaining a comprehensive feature vector according to the path feature vector and the cross feature vector specifically includes:
according to the path eigenvector and the cross eigenvector, adopting a gating mechanism to respectively obtain the weight coefficients of the path eigenvector and the cross eigenvector;
and obtaining a comprehensive characteristic vector according to the path characteristic vector and the corresponding weight coefficient, the cross characteristic vector and the corresponding weight coefficient.
Optionally, the synthetic feature vector is determined according to the following formula:
Fg＝(δ(WeFe+WpFp))⊙Fe+(1-δ(WeFe+WpFp))⊙Fp；
wherein, FeAs cross feature vectors, FgAs path feature vectors, WeFor linear transformation of the cross feature vectors, WpFor linear transformation of the path feature vector, δ is the activation function, δ (W)eFe+WpFp) As cross feature vectorsWeight coefficient of (1- δ) (W)eFe+WpFp) Weight coefficients of path eigenvectors, <' > as vector dot products, FgIs the synthetic feature vector.
Optionally, the determining a service quality prediction value by using a multi-layer fully-connected network according to the comprehensive feature vector specifically includes:
determining a quality of service prediction value according to the following formula:
wherein, W1Weight of the full connection layer, b1For bias of fully connected layers, WoFor a linear transformation matrix, FgIn order to synthesize the feature vectors,
is a predicted value of the service quality.
Optionally, the method for predicting Web service quality based on network topology awareness further includes:
determining an objective optimization function of the quality of service prediction model according to the following formula:
wherein loss is an objective optimization function of the service quality prediction model, m is the number of samples, yiAs the actual response time of the ith sample data,
predicting response time of ith sample data;
and performing iterative training on the service quality prediction model by adopting a random gradient descent algorithm and through a back propagation minimized objective function until the service quality prediction model corresponding to the objective optimization function with the minimum function value is obtained.
In order to achieve the above purpose, the invention also provides the following scheme:
a Web service quality prediction system based on network topology awareness comprises the following components:
the acquisition unit is used for acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes;
the training unit is connected with the acquisition unit and used for training the topology sensing neural network according to the sample data to obtain a service quality prediction model;
the prediction unit is connected with the training unit and used for obtaining a target service quality prediction value according to a target user, a target server and a target shortest path corresponding to the target user and the target server based on the service quality prediction model; the target shortest path comprises a target user, a target server and an intermediate autonomous domain node.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: acquiring sample data, wherein the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; training the topology perception neural network according to the sample data to obtain a service quality prediction model; based on a service quality prediction model, a target service quality prediction value is obtained according to a target user, a target server and a target shortest path corresponding to the target user and the target server, the influence of a communication path in a Web service calling process is considered in service quality prediction, and service quality prediction precision is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a Web service quality prediction method based on network topology awareness according to the present invention;
FIG. 2 is a flow chart of a method of establishing a quality of service prediction model;
FIG. 3 is a schematic diagram of obtaining an interactive feature vector set;
FIG. 4 is a schematic diagram of a convolutional neural network;
FIG. 5 is a schematic structural diagram of a topology aware neural network;
fig. 6 is a schematic structural diagram of the Web service quality prediction system based on network topology awareness according to the present invention.
Description of the symbols:
the system comprises an acquisition unit-1, a training unit-2 and a prediction unit-3.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a Web service quality prediction method and a Web service quality prediction system based on network topology perception.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
As shown in fig. 1, the method for predicting Web service quality based on network topology awareness of the present invention includes:
s1: acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values. The historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes. The historical quality of service predictor includes response time and throughput. The influence of the communication path in the Web service calling process is fully considered in the acquired sample data, and the precision of service quality prediction is further improved.
S2: and training the topology perception neural network according to the sample data to obtain a service quality prediction model.
S3: and based on the service quality prediction model, obtaining a target service quality prediction value according to the target user, the target server and a target shortest path corresponding to the target user and the target server. The target shortest path comprises a target user, a target server and an intermediate autonomous domain node. The target quality of service predictor includes a response time and a throughput.
In this embodiment, the target shortest path is expressed as
PATH '{ u', uas ', as 1', as1 ', ask', sas ', s'. Wherein u 'is a target user, s' is a target server, uas 'is the autonomous domain number to which the target user belongs, sas' is the autonomous domain number to which the target server belongs, and the intermediate node as1 '-ask' is the autonomous domain number on the route from uas 'to sas' for routing. The target shortest path comprises k intermediate autonomous domain nodes between an autonomous domain to which a target user belongs and an autonomous domain to which a target server belongs, and n is k +4 different nodes.
Wherein, S1: and in the sample data, the history user and the history server are all users and servers in the service calling history record, and a communication path sequence from the source user to the target server is obtained. Under the condition of a known network topological graph, all network nodes (including users, servers and autonomous domains) are numbered uniformly, and sample data comprises user numbers, server numbers and shortest paths acquired by service invocation historical data.
Specifically, the method for calculating the historical shortest path includes: and aiming at each historical user and the corresponding historical server, determining the historical shortest path between the historical user and the corresponding historical server by adopting a shortest path algorithm based on a network topological graph according to the historical user and the corresponding historical server. The history shortest path is directly used in the sample data, so that the complexity of subsequent pair topology-aware neural network training is simplified.
The calculation method of the target shortest path is the same as that of the historical shortest path.
As shown in fig. 2, S2: training the topology-aware neural network according to the sample data to obtain a service quality prediction model, which specifically comprises:
s21: and mapping the historical user into a user embedded feature vector, mapping the historical server into a server embedded feature vector, and mapping each intermediate autonomous domain node into a corresponding embedded feature vector. In this embodiment, the user embedded feature vector, the server embedded feature vector, and the dimension of the embedded feature vector corresponding to the mapping of each intermediate autonomous domain node are all fixed.
S22: and obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector. The cross feature vector is used to implicitly characterize the service invocation process.
S23: and determining the path characteristic vector according to the embedded characteristic vector corresponding to each intermediate autonomous domain node.
S24: and determining a comprehensive characteristic vector according to the path characteristic vector and the cross characteristic vector.
S25: and determining a service quality predicted value by adopting a multilayer full-connection network according to the comprehensive characteristic vector.
Specifically, the predicted value of the quality of service is determined according to the following formula:
wherein, W1Weight of the full connection layer, b1For bias of fully connected layers, WoIs a linear transformation matrix, WoLinear transformation matrix for output, linear transformation of the last layer output of the fully-connected layer, FgIn order to synthesize the feature vectors,
is a predicted value of the service quality.
In this embodiment, all history users, history servers, and intermediate autonomous domain nodes in the topology map are numbered uniformly. Setting a set PATH formed by the history user, the history server and the intermediate autonomous domain node as { u, uas, as1，...，askSa, s } is mapped uniformly to a fixed length h of a feature vector, Ei(ii) wherein EiAnd the characteristic vector corresponding to the node with the number i in the node set is represented, u is a historical user, s is a historical server, uas is the autonomous domain number to which the historical user belongs, sas is the autonomous domain number to which the historical server belongs, and the intermediate node as1-ask is the autonomous domain number on the route from the uas to the sas selected by the routing.
Wherein, S22: obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector, and specifically comprising:
s221: and carrying out high-dimensional interaction on the user embedded characteristic vector and the server embedded characteristic vector through an outer product to obtain an interactive characteristic vector set. In this embodiment, as shown in fig. 3, the process of performing high-dimensional interaction is as follows: multiplying the values of all dimensions of the user embedded characteristic vector and the server embedded characteristic vector pairwise to obtain h2Group cross feature vectors, thus peer-to-peer interactionsThe process is simulated.
S222: and extracting features by adopting a convolutional neural network according to the interactive feature vector set to obtain cross feature vectors. Because redundant data exists in the cross feature vector set, the redundant data can be removed by extracting through the convolutional neural network.
Preferably, S222: according to the interactive feature vector set, performing feature extraction by adopting a convolutional neural network to obtain a cross feature vector, which specifically comprises the following steps:
as shown in fig. 4, the convolutional neural network includes L convolutional layers and T convolutional kernels; extracting a feature from each convolution kernel; the input of each layer of convolution is represented as a matrix MlWherein l is the number of the convolutional layer.
The set of interaction feature vectors, denoted M, as the initial input matrix0(ii) a Taking the output matrix of the previous layer as the input matrix of the next layer, and so on, the features extracted by the ith convolution kernel of the ith layer are as follows:
wherein M isl-1Is the l-th layer input matrix, i.e. the l-1-th layer output matrix,
the ith convolution kernel for the ith layer,
for convolution, blFor biasing, Relu () is the activation function.
The output matrix (cross-feature) of layer i is:
further, S23: determining a path feature vector according to the embedded feature vectors corresponding to the nodes of the intermediate autonomous domain, specifically comprising:
and based on the bidirectional long-short term memory network, aggregating the embedded characteristic vectors corresponding to the nodes of each intermediate autonomous domain to obtain corresponding path characteristic vectors. The path feature vector is used for displaying a characterization service calling process.
As another embodiment, the determining a path feature vector according to an embedded feature vector corresponding to each intermediate autonomous domain node specifically includes: and inputting the embedded characteristic vectors corresponding to the nodes of each intermediate autonomous domain based on the recurrent neural network, and obtaining the path characteristic vectors by learning the dependency relationship among the nodes.
Further, S24: obtaining a comprehensive feature vector according to the path feature vector and the cross feature vector, and specifically comprising:
s241: and according to the path eigenvector and the cross eigenvector, respectively obtaining the weight coefficients of the path eigenvector and the cross eigenvector by adopting a gating mechanism.
S242: and obtaining a comprehensive characteristic vector according to the path characteristic vector and the corresponding weight coefficient, the cross characteristic vector and the corresponding weight coefficient.
Specifically, the synthetic feature vector is determined according to the following formula:
Fg＝(δ(WeFe+WpFp))⊙Fe+(1-δ(WeFe+WpFp))⊙Fp；
wherein, FeAs cross feature vectors, FpAs path feature vectors, WeFor linear transformation of the cross feature vectors, WpFor linear transformation of the path feature vector, δ is the activation function, δ (W)eFe+WpFp) Weight coefficients for the cross feature vector, 1-delta (W)eFe+WpFp) Weight coefficients of path eigenvectors, <' > as vector dot products, FgIs the synthetic feature vector. The activation function δ ensures the resulting weight distribution δ (W)eFe+WpFp) The value of (a) is between 0 and 1.
Optionally, as shown in fig. 5, the topology-aware neural network includes an input layer, a display path modeling layer, an implicit interaction modeling layer, a fusion layer, and a prediction layer, which are associated with each other.
The input layer is used for mapping the historical user into a user embedded characteristic vector, mapping the historical server into a server embedded characteristic vector and mapping each intermediate autonomous domain node into a corresponding embedded characteristic vector.
And the display path modeling layer is used for determining the path characteristic vectors according to the embedded characteristic vectors corresponding to the nodes of the middle autonomous domain.
And the implicit interactive modeling layer is used for obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector.
And the fusion layer is used for determining a comprehensive characteristic vector according to the path characteristic vector and the cross characteristic vector.
And the prediction layer is used for determining a service quality prediction value by adopting a multilayer full-connection network according to the comprehensive characteristic vector.
The invention constructs a specific neural network model to predict the service quality, and models the service calling process by fusing explicit path characteristics and implicit cross characteristics, thereby further improving the efficiency and accuracy of prediction.
In order to improve the prediction precision, the method for predicting the Web service quality based on the network topology perception further comprises the following steps:
parameters in the network topology aware neural network are initialized. Specifically, each feature vector, each weight and a bias vector are randomly initialized and subjected to normal distribution.
Determining an objective optimization function of the quality of service prediction model according to the following formula:
wherein loss is an objective optimization function of the service quality prediction model, m is the number of samples,yias the actual response time of the ith sample data,
the predicted response time of the ith sample data.
The prediction target may also be a throughput rate. When the prediction target is the throughput rate, yiFor the actual throughput rate of the ith sample data,
the predicted throughput rate for the ith sample data.
And performing iterative training on the service quality prediction model by adopting a random gradient descent algorithm and through a back propagation minimized objective function until the service quality prediction model corresponding to the objective optimization function with the minimum function value is obtained.
As shown in fig. 6, the Web service quality prediction system based on network topology awareness of the present invention includes: the device comprises an acquisition unit 1, a training unit 2 and a prediction unit 3.
The acquisition unit 1 is used for acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes.
The training unit 2 is connected with the acquisition unit 1, and the training unit 2 is used for training the topology-aware neural network according to the sample data to obtain a service quality prediction model.
The prediction unit 3 is connected with the training unit 2, and the prediction unit 3 is used for obtaining a target service quality prediction value according to a target user, a target server and a target shortest path corresponding to the target user and the target server based on the service quality prediction model; the target shortest path comprises a target user, a target server and an intermediate autonomous domain node.
Compared with the prior art, the Web service quality prediction system based on network topology perception has the same beneficial effects as the Web service quality prediction method based on network topology perception, and the details are not repeated herein.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A Web service quality prediction method based on network topology awareness is characterized by comprising the following steps:
acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes; the historical quality of service predictor comprises response time and throughput;
training the topology perception neural network according to the sample data to obtain a service quality prediction model;
based on the service quality prediction model, obtaining a target service quality prediction value according to a target user, a target server and a target shortest path corresponding to the target user and the target server; the target shortest path comprises a target user, a target server and an intermediate autonomous domain node; the target quality of service predictor includes a response time and a throughput.
2. The method for predicting Web service quality based on network topology awareness according to claim 1, wherein the method for calculating the historical shortest path includes:
and aiming at each historical user and the corresponding historical server, determining the historical shortest path between the historical user and the corresponding historical server by adopting a shortest path algorithm based on a network topological graph according to the historical user and the corresponding historical server.
3. The method according to claim 1, wherein the training of the topology-aware neural network according to the sample data to obtain a service quality prediction model specifically comprises:
mapping the historical user into a user embedded feature vector, mapping the historical server into a server embedded feature vector, and mapping each intermediate autonomous domain node into a corresponding embedded feature vector;
obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector;
determining a path characteristic vector according to the embedded characteristic vectors corresponding to the nodes of the intermediate autonomous domain;
determining a comprehensive characteristic vector according to the path characteristic vector and the cross characteristic vector;
and determining a service quality predicted value by adopting a multilayer full-connection network according to the comprehensive characteristic vector.
4. The method for predicting Web service quality based on network topology awareness according to claim 3, wherein the obtaining a cross feature vector according to the user embedded feature vector and the server embedded feature vector specifically includes:
carrying out high-dimensional interaction on the user embedded characteristic vector and the server embedded characteristic vector through an outer product to obtain an interaction characteristic vector set;
and extracting features by adopting a convolutional neural network according to the interactive feature vector set to obtain cross feature vectors.
5. The method for predicting Web service quality based on network topology awareness according to claim 3, wherein the determining a path feature vector according to the embedded feature vector corresponding to each intermediate autonomous domain node specifically includes:
and based on the bidirectional long-short term memory network, aggregating the embedded characteristic vectors corresponding to the nodes of each intermediate autonomous domain to obtain corresponding path characteristic vectors.
6. The method for predicting Web service quality based on network topology awareness according to claim 3, wherein the obtaining a comprehensive feature vector according to the path feature vector and the cross feature vector specifically includes:
according to the path eigenvector and the cross eigenvector, adopting a gating mechanism to respectively obtain the weight coefficients of the path eigenvector and the cross eigenvector;
and obtaining a comprehensive characteristic vector according to the path characteristic vector and the corresponding weight coefficient, the cross characteristic vector and the corresponding weight coefficient.
7. The method of claim 6, wherein the comprehensive feature vector is determined according to the following formula:
Fg＝(δ(WeFe+WpFp))⊙Fe+(1-δ(WeFe+WpFp))⊙Fp；
wherein, FeAs cross feature vectors, FpAs path feature vectors, WeFor linear transformation of the cross feature vectors, WpFor linear transformation of the path feature vector, δ is the activation function, δ (W)eFe+WpFp) Weight coefficients for the cross feature vector, 1-δ(WeFe+WpFp) Weight coefficients of path eigenvectors, <' > as vector dot products, FgIs the synthetic feature vector.
8. The method for predicting Web service quality based on network topology awareness according to claim 3, wherein the determining a predicted value of service quality by using a multi-layer fully-connected network according to the comprehensive feature vector specifically includes:
determining a quality of service prediction value according to the following formula:
wherein, W1Weight of the full connection layer, b1For bias of fully connected layers, WoFor a linear transformation matrix, FgIn order to synthesize the feature vectors,
is a predicted value of the service quality.
9. The method of claim 1, wherein the method further comprises:
determining an objective optimization function of the quality of service prediction model according to the following formula:
wherein loss is an objective optimization function of the service quality prediction model, m is the number of samples, yiAs the actual response time of the ith sample data,
for the i-th sample dataMeasuring the response time;
and performing iterative training on the service quality prediction model by adopting a random gradient descent algorithm and through a back propagation minimized objective function until the service quality prediction model corresponding to the objective optimization function with the minimum function value is obtained.
10. A Web service quality prediction system based on network topology awareness is characterized by comprising:
the acquisition unit is used for acquiring sample data; the sample data comprises a plurality of groups of historical users, historical servers, historical shortest paths corresponding to the historical users and the historical servers, and historical service quality predicted values; the historical shortest path comprises a historical user, a historical server and a plurality of intermediate autonomous domain nodes;
the training unit is connected with the acquisition unit and used for training the topology sensing neural network according to the sample data to obtain a service quality prediction model;
the prediction unit is connected with the training unit and used for obtaining a target service quality prediction value according to a target user, a target server and a target shortest path corresponding to the target user and the target server based on the service quality prediction model; the target shortest path comprises a target user, a target server and an intermediate autonomous domain node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110534218.5A CN113259163B (en) | 2021-05-17 | 2021-05-17 | Web service quality prediction method and system based on network topology perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110534218.5A CN113259163B (en) | 2021-05-17 | 2021-05-17 | Web service quality prediction method and system based on network topology perception |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113259163A CN113259163A (en) | 2021-08-13 |
| CN113259163B true CN113259163B (en) | 2022-02-08 |
Family
ID=77182171
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110534218.5A Expired - Fee Related CN113259163B (en) | 2021-05-17 | 2021-05-17 | Web service quality prediction method and system based on network topology perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113259163B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114065033B (en) * | 2021-11-12 | 2024-08-23 | 西安邮电大学 | Training method for recommending graph neural network model of Web service combination |
| CN115473838B (en) * | 2022-09-15 | 2024-10-11 | 中国电信股份有限公司 | Processing method and device of network request, computer readable medium and electronic equipment |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10728553B2 (en) * | 2017-07-11 | 2020-07-28 | Sony Corporation | Visual quality preserving quantization parameter prediction with deep neural network |
| CN108256695A (en) * | 2018-03-14 | 2018-07-06 | 北京邮电大学 | A kind of Web service QoS Forecasting Methodologies based on historical information |
| CN109756911B (en) * | 2019-01-31 | 2020-11-27 | 腾讯科技（深圳）有限公司 | Network quality prediction method, service adjustment method, related device and storage medium |
-
2021
- 2021-05-17 CN CN202110534218.5A patent/CN113259163B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN113259163A (en) | 2021-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109544306B (en) | Cross-domain recommendation method and device based on user behavior sequence characteristics | |
| CN113254792B (en) | Method for training recommendation probability prediction model, recommendation probability prediction method and device | |
| Li et al. | Topology-aware neural model for highly accurate QoS prediction | |
| CN113259163B (en) | Web service quality prediction method and system based on network topology perception | |
| CN112184391A (en) | Recommendation model training method, medium, electronic device and recommendation model | |
| CN113761388B (en) | Recommendation method and device, electronic equipment and storage medium | |
| CN114896899B (en) | Multi-agent distributed decision method and system based on information interaction | |
| CN115631008B (en) | Commodity recommendation method, device, equipment and medium | |
| CN112883170A (en) | User feedback guided self-adaptive conversation recommendation method and system | |
| CN113032583A (en) | User rights and interests matching method and device | |
| CN117633371B (en) | Recommendation method, device and readable storage medium based on multi-attention mechanism | |
| CN114925270A (en) | Session recommendation method and model | |
| Li et al. | Tell me where to go next: Improving poi recommendation via conversation | |
| CN115858911A (en) | Information recommendation method and device, electronic equipment and computer-readable storage medium | |
| CN117194771B (en) | Dynamic knowledge graph service recommendation method for graph model characterization learning | |
| CN116976402A (en) | Training method, device, equipment and storage medium of hypergraph convolutional neural network | |
| CN115048530A (en) | Graph convolution recommendation system fusing neighbor importance and feature learning | |
| CN115080856A (en) | Recommendation method and device and training method and device of recommendation model | |
| CN114969209A (en) | Training method and device, and method and device for predicting resource consumption | |
| Yao et al. | Application of Neural Network and Structural Model in AI Educational Performance Analysis. | |
| CN112052386A (en) | Information recommendation method and device and storage medium | |
| CN110880141A (en) | Intelligent deep double-tower model matching algorithm and device | |
| CN118395022B (en) | Time-space perception multi-component diagram attention time sequence QoS prediction method and system | |
| CN117786234B (en) | Multimode resource recommendation method based on two-stage comparison learning | |
| CN118469715B (en) | Risk management method and system for cross-border e-commerce transaction behaviors |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220208 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |