CN113762737A - Method and system for predicting network service quality - Google Patents
Method and system for predicting network service quality Download PDFInfo
- Publication number
- CN113762737A CN113762737A CN202110956949.9A CN202110956949A CN113762737A CN 113762737 A CN113762737 A CN 113762737A CN 202110956949 A CN202110956949 A CN 202110956949A CN 113762737 A CN113762737 A CN 113762737A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- network service
- training
- trained
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 72
- 238000013528 artificial neural network Methods 0.000 claims description 32
- 238000011176 pooling Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 13
- 238000013140 knowledge distillation Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 13
- 238000012423 maintenance Methods 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 description 8
- 230000006835 compression Effects 0.000 description 8
- 238000007906 compression Methods 0.000 description 8
- 239000013598 vector Substances 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for predicting network service quality, which comprises the following steps: acquiring user data and service data, and inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; the soft label is prediction data output by the trained deep FM model after sample data is input. According to the method, the deep FM model is adopted, manual feature combination is not needed, a sparse data set can be processed, and prediction data output after sample data is input into the trained deep FM model is used as a soft label to train the network service quality prediction model, so that the scale of the network service quality prediction model is reduced, and the operation and maintenance load is reduced.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method and a system for predicting network service quality.
Background
In recent years, the explosive development of network services has led to the rapid expansion of the number of homogeneous network services provided by various network service providers, and it has become a great difficulty to select the network service most suitable for users from a large number of homogeneous network services. The network service quality prediction technology can effectively solve the problem by predicting the quality of a large number of homogeneous network services which have not been used.
In the predictive class of problems, it is important to acquire and learn knowledge information from a large amount of data. Common methods include Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Factorization Machines (FM), Wide & Deep, and the like. Wherein CNN is biased towards learning the combined features of neighboring features; RNN is applied to data with a time sequence relation; FM represents the combination feature by making inner product with hidden vector, thus solving the problem of low-order and high-order combination feature extraction theoretically, and being limited by computational complexity and very limited in practical application; wide & Deep can learn low-order and high-order combined features at the same time, but depends on artificial feature engineering, the model result is greatly influenced by the feature engineering, the requirements on manpower and computing resources are increased, a sparse data set and high-dimensional information are difficult to process, and the Wide & Deep is not suitable for network service data containing the high-dimensional information.
In summary, there is a need for a method for predicting network qos, which is used to solve the above-mentioned problems in the prior art.
Disclosure of Invention
Because the existing method has the problems, the invention provides a method and a system for predicting the network service quality.
In a first aspect, the present invention provides a method for predicting network service quality, including:
acquiring user data and service data;
inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result;
the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
Further, before the inputting the user data and the service data into the trained network service quality prediction model to obtain a network service quality prediction result, the method further includes:
acquiring the sample data and a data tag corresponding to the sample data;
constructing a deep FM model;
training the deep FM model according to the sample data and the data label to obtain the trained deep FM model and the soft label;
taking the trained DeepFM model as a teacher model under a knowledge distillation frame and constructing a student model;
and training the student model according to the sample data, the data label and the soft label to obtain a trained network service quality prediction model.
Further, the constructing the deep fm model includes:
and improving the pooling layer of the DeepFM model by adopting a self-adaptive pooling method to obtain an improved DeepFM model.
Further, the building of the student model comprises:
constructing a fully-connected neural network according to the trained DeepFM model; the hidden layer of the fully-connected neural network comprises an extension layer and two fully-connected layers;
and determining the student model according to the fully-connected neural network.
Further, the training the student model according to the sample data, the data label and the soft label to obtain a trained network service quality prediction model includes:
training the student model according to the sample data and the data labels to obtain a first student model;
training the first student model according to the sample data and the soft label to obtain a second student model;
and taking the second student model as a trained network service quality prediction model.
Further, the obtaining the sample data and the data tag corresponding to the sample data includes:
determining a first characteristic corresponding to the user training data according to the user training data; the first characteristic comprises a user number, an IP address, a country of the user, a number of an autonomous system, and a longitude or latitude of a user party;
determining a second characteristic corresponding to the service training data according to the service training data; the second characteristic comprises the number of a network service party, a network service uniform resource positioning system, a service provider, an IP address, a country of the network service party, the number of an autonomous system, and longitude or latitude.
In a second aspect, the present invention provides a system for predicting network service quality, including:
the acquisition module is used for acquiring user data and service data;
the processing module is used for inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
Further, the processing module is further configured to:
before the user data and the service data are input into a trained network service quality prediction model to obtain a network service quality prediction result, acquiring the sample data and a data label corresponding to the sample data;
constructing a deep FM model;
training the deep FM model according to the sample data and the data label to obtain the trained deep FM model and the soft label;
taking the trained DeepFM model as a teacher model under a knowledge distillation frame and constructing a student model;
and training the student model according to the sample data, the data label and the soft label to obtain a trained network service quality prediction model.
Further, the processing module is specifically configured to:
and improving the pooling layer of the DeepFM model by adopting a self-adaptive pooling method to obtain an improved DeepFM model.
Further, the processing module is specifically configured to:
constructing a fully-connected neural network according to the trained DeepFM model; the hidden layer of the fully-connected neural network comprises an extension layer and two fully-connected layers;
and determining the student model according to the fully-connected neural network.
Further, the processing module is specifically configured to:
training the student model according to the sample data and the data labels to obtain a first student model;
training the first student model according to the sample data and the soft label to obtain a second student model;
and taking the second student model as a trained network service quality prediction model.
Further, the processing module is specifically configured to:
determining a first characteristic corresponding to the user training data according to the user training data; the first characteristic comprises a user number, an IP address, a country of the user, a number of an autonomous system, and a longitude or latitude of a user party;
determining a second characteristic corresponding to the service training data according to the service training data; the second characteristic comprises the number of a network service party, a network service uniform resource positioning system, a service provider, an IP address, a country of the network service party, the number of an autonomous system, and longitude or latitude.
In a third aspect, the present invention also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the method for predicting network service quality according to the first aspect is implemented.
In a fourth aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method for network quality of service prediction as described in the first aspect.
According to the technical scheme, the method and the system for predicting the network service quality provided by the invention have the advantages that the deep FM model is adopted, the feature combination is not required manually, the sparse data set can be processed, the student model is constructed by adopting a model compression method of knowledge distillation, the student model is trained according to a knowledge distillation mechanism, the scale of the network service quality prediction model is reduced, and the operation and maintenance load is reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a system framework for a method of network quality of service prediction provided by the present invention;
FIG. 2 is a flow chart illustrating a method for predicting network QoS according to the present invention;
FIG. 3 is a flow chart illustrating a method for predicting network QoS according to the present invention;
FIG. 4 is a schematic diagram of an improved deep FM model provided by the present invention;
FIG. 5 is a schematic diagram of a knowledge distillation framework provided by the present invention;
FIG. 6 is a schematic diagram of a student model provided by the present invention;
FIG. 7 is a schematic diagram of a system for predicting network QoS according to the present invention;
fig. 8 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
The following further describes embodiments of the present invention with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
The method for predicting the network service quality provided by the embodiment of the invention can be applied to a system architecture as shown in fig. 1, wherein the system architecture comprises a deep fm model 100 and a network service quality prediction model 200.
Specifically, the deep fm model 100 is used to obtain user data, service data, and data labels, and obtain a trained deep fm model and soft labels after the user data, the service data, and the data labels are trained.
The network service quality prediction model 200 is used for obtaining a network service quality prediction result after user data and service data are input after training according to the user data, the service data and the soft label.
It should be noted that fig. 1 is only an example of a system architecture according to the embodiment of the present invention, and the present invention is not limited to this specifically.
Based on the above illustrated system architecture, fig. 2 is a flowchart corresponding to a method for predicting network service quality according to an embodiment of the present invention, as shown in fig. 2, the method includes:
It should be noted that the data characteristics corresponding to the service data include a number (ws _ id) of a network service party, a URL (ws _ URL) of a network service uniform resource locator system (wsjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj. The embodiment of the present invention is not particularly limited to this.
It should be noted that, the trained network service quality prediction model is obtained by training with different sample data and corresponding soft labels; the sample data comprises user training data and service training data; the soft label is prediction data output by the trained deep FM model after sample data is input.
According to the scheme, the deep FM model is adopted, manual feature combination is not needed, a sparse data set can be processed, prediction data output after sample data is input into the trained deep FM model is used as a soft label to train the network service quality prediction model, the scale of the network service quality prediction model is reduced, and operation and maintenance loads are reduced.
Before step 202, the embodiment of the present invention has a step flow as shown in fig. 3, which is specifically as follows:
Specifically, a first characteristic corresponding to user training data is determined according to the user training data; the first characteristic comprises a user number, an IP address, a country of the user, a number of an autonomous system, and longitude or latitude of a user party;
determining a second characteristic corresponding to the service training data according to the service training data; the second feature includes a number of a network service provider, a network service uniform resource location system, a service provider, an IP address, a country of the country, an autonomous system number, a longitude, or a latitude.
For example, the sample data and data tags are shown in table 1:
TABLE 1
ws_id | ws_ip | ws_as | ws_url | ws_latitude | ws_longitude | … | label |
0 | 1E+27 | 1 | 1.8E+101 | 38 | -97 | … | 5.394 |
1 | 1E+27 | 1 | 1.8E+125 | 38 | -97 | … | 5.55 |
2 | 1E+27 | 1 | 1.8E+123 | 38 | -97 | … | 7.067 |
3 | 1E+27 | 1 | 1.8E+103 | 38 | -97 | … | 6.415 |
4 | 1E+27 | 2 | 1.8E+99 | 38 | -97 | … | 5.828 |
5 | 1.1E+31 | 3 | 1.8E+123 | 40.7904 | -74.0246 | … | 5.875 |
6 | 0 | 4 | 1.8E+105 | -45.5881 | -69.07 | … | 5.782 |
7 | 0 | 4 | 1.8E+109 | -45.5881 | -69.07 | … | 5.776 |
8 | 1.1E+31 | 5 | 1.8E+115 | -34.5875 | -58.6725 | … | 5.565 |
For example, when the number (ws _ id) of the web service side is 0, the web service URL (ws _ URL) is 1.8E +101, the IP address (ws _ IP) is 1E +27, the AS number (ws _ AS) is 1, the longitude (ws _ longitude) is 38, and the latitude (ws _ longitude) is-97, the corresponding data tag (label) is 5.394.
According to the scheme, the data characteristics are extracted from the user data and the service data, and the prediction precision is improved.
And step 302, constructing a DeepFM model.
The DeepFM model mainly comprises a compression neural network FM and a deep neural network DNN, and the compression neural network is adopted to replace the traditional linear regression, so that the explicit feature learning of high-dimensional information can be effectively realized.
In one possible implementation, the improved deep fm model is obtained by improving the pooling layer of the deep fm model by using an adaptive pooling method.
In the embodiment of the invention, the specific steps of constructing the deep FM model comprise the following flows:
s1: user data and service data are converted into X and y vectors with the length of D and reconstructed into a matrix X0∈Rm×DWherein m is x + y.
Wherein X is the user data volume, y is the service data volume, XkIs a k-th hidden layer.
S2: for each hidden layer, XkThe calculation is made by the following formula:
wherein H is more than or equal to 1 and less than or equal to Hk,HkRepresents the k-ththThe number of feature vectors of a layer,is the h ththA parameter matrix of the layer feature vector,representing a Hadamard product (Hadamard product).
It should be noted that the calculation method of the hadamard product is shown as the following formula:
s3: an adaptive pooling method is employed. Each hidden layer X of the modelk,k∈[1,T]And connecting with an output neuron, wherein T is the iteration number of the adaptive pooling.
In the embodiment of the invention, the deep fm model includes a compression neural network, a deep neural network and the like, as shown in fig. 4. Wherein, the last layer of the pooling layer of the compressed neural network adopts a self-adaptive pooling method.
After compression HiOne vector U ═ of<u1,u2,…,un>For example, the pooling process includes multiple iterations, and the result of each iteration is used as an auxiliary parameter for the next iteration.
In one possible embodiment, the number of iterations T is set to 3.
Specifically, each round of iterative computation includes:
s31: representing the result of the previous iteration as a vectorWhere r represents the current iteration number, BrThe parameters are calculated for each iteration.
It should be noted that, since there is no previous iteration result in the first iteration, B will be usedrIs initialized to
S32: will be from HiVector U ═ of<u1,u2,…,un>Performing a neural network-like computation as an input with an output of srThe specific calculation formula is as follows:
the parameter C isiIn effect a softmax function. Since it is dynamically decided in the calculation process, and not learned, the calculation is called a neural network-like structure.
S33: adjustment srAnd limiting s in a preset range, wherein a specific calculation formula is as follows:
in the embodiment of the invention, if the current round is overlapped to be the last round, a is replacedrAnd outputting as a result y of the adaptive pooling, otherwise performing an operation of S34.
S34: by arCalculating vector Br+1And used for the next iteration of the calculation.
In particular, the method comprises the following steps of,the calculation formula of (a) is as follows:
s4: and constructing a parallel deep neural network.
The deep neural network is composed of three fully connected layers.
In the embodiment of the present invention, the output result of the deep fm model may be represented as:
wherein, yFMRepresenting the output of a compressed neural network, yDNNRepresenting the output of the deep neural network.
According to the scheme, the self-adaptive pooling method is adopted for carrying out information extraction self-adaptive optimization on the compressed neural network in the deep FM model aiming at the high-dimensional characteristics of the network service data, the high-dimensional information extraction capability is improved, high-dimensional and low-dimensional data information is considered, and artificial characteristic engineering is not needed, so that the prediction precision is improved.
The soft label is prediction data output by the trained deep fm model after sample data is input.
For example, as shown in table 2:
TABLE 2
ws_id | ws_ip | ws_as | ws_url | ws_latitude | ws_longitude | … | label |
0 | 1E+27 | 1 | 1.8E+101 | 38 | -97 | … | 5.261 |
1 | 1E+27 | 1 | 1.8E+125 | 38 | -97 | … | 5.294 |
2 | 1E+27 | 1 | 1.8E+123 | 38 | -97 | … | 7.184 |
3 | 1E+27 | 1 | 1.8E+103 | 38 | -97 | … | 6.451 |
4 | 1E+27 | 2 | 1.8E+99 | 38 | -97 | … | 5.814 |
5 | 1.1E+31 | 3 | 1.8E+123 | 40.7904 | -74.0246 | … | 5.991 |
6 | 0 | 4 | 1.8E+105 | -45.5881 | -69.07 | … | 5.561 |
7 | 0 | 4 | 1.8E+109 | -45.5881 | -69.07 | … | 5.981 |
8 | 1.1E+31 | 5 | 1.8E+115 | -34.5875 | -58.6725 | … | 5.419 |
AS can be seen from table 2, when the number (ws _ id) of the web service side is 0, the web service URL (ws _ URL) is 1.8E +101, the IP address (ws _ IP) is 1E +27, the AS number (ws _ AS) is 1, the longitude (ws _ longitude) is 38, and the latitude (ws _ longitude) is-97, the corresponding soft tag (label) is 5.261, which is different from the data tag 5.394.
And step 304, taking the trained deep FM model as a teacher model under a knowledge distillation frame and constructing a student model.
It should be noted that knowledge distillation adopts a teacher-student mode: the complex and large model is used as a teacher model, the structure of the student model is simple, the teacher model is used for assisting the training of the student model, the learning ability of the teacher model is strong, and the knowledge learned by the teacher model can be transferred to the student model with relatively weak learning ability, so that the generalization ability of the student model is enhanced.
And 305, training the student model according to the sample data, the data labels and the soft labels to obtain a trained network service quality prediction model.
Specifically, training a student model according to sample data and a data label to obtain a first student model;
training the first student model according to the sample data and the soft label to obtain a second student model;
and taking the second student model as a trained network service quality prediction model.
In the embodiment of the invention, as shown in fig. 5, the knowledge learned by the teacher model is used to guide the training of the student model, so that the student model has the performance equivalent to that of the teacher model, but the number of parameters is greatly reduced, thereby realizing the compression and acceleration of the model.
According to the scheme, the knowledge distillation method is adopted to compress and optimize the model, so that the requirements on computing resources and memory in the operation process are reduced, and the load pressure brought to the server by network service quality prediction is reduced.
In the embodiment of the present invention, the user data and the service data are input into the trained network service quality prediction model to obtain the network service quality prediction result, which is shown in table 3:
TABLE 3
ws_id | ws_ip | ws_as | ws_url | ws_latitude | ws_longitude | … | label |
0 | 1E+27 | 1 | 1.8E+101 | 38 | -97 | … | 5.518 |
1 | 1E+27 | 1 | 1.8E+125 | 38 | -97 | … | 5.291 |
2 | 1E+27 | 1 | 1.8E+123 | 38 | -97 | … | 7.01 |
3 | 1E+27 | 1 | 1.8E+103 | 38 | -97 | … | 6.507 |
4 | 1E+27 | 2 | 1.8E+99 | 38 | -97 | … | 5.794 |
5 | 1.1E+31 | 3 | 1.8E+123 | 40.7904 | -74.0246 | … | 5.812 |
6 | 0 | 4 | 1.8E+105 | -45.5881 | -69.07 | … | 5.791 |
7 | 0 | 4 | 1.8E+109 | -45.5881 | -69.07 | … | 5.681 |
8 | 1.1E+31 | 5 | 1.8E+115 | -34.5875 | -58.6725 | … | 5.597 |
According to the scheme, the deep FM model is adopted, manual feature combination is not needed, a sparse data set can be processed, a small full-connection neural network is constructed by adopting a knowledge distillation model compression method and is used as a student model, the student model is trained according to a knowledge distillation mechanism, the scale of a network service quality prediction model is reduced, and operation and maintenance loads are reduced.
Further, in step 304, the fully-connected neural network is constructed according to the trained deep fm model according to the embodiment of the present invention.
It should be noted that the hidden layer of the fully-connected neural network includes one extended layer and two fully-connected layers.
And determining the student model according to the fully-connected neural network.
In one possible implementation, a fully-connected neural network with the same prediction category is established according to the improved deep FM model.
It should be noted that the parameters of the fully-connected neural network are much smaller than the improved deep fm model, which is used as a student model in the teacher-student mode.
In the embodiment of the present invention, the hidden layer of the fully-connected neural network includes one extended layer and two fully-connected layers, as shown in fig. 6.
As can be seen from fig. 6, the second fully-connected layer is directly connected to the output layer, and the prediction result of the layer on the network service quality should be the same as the improved deep fm model.
It should be noted that the hidden layer of the fully-connected neural network may further include an extended layer and a fully-connected layer, which is not specifically limited in this embodiment of the present invention.
According to the scheme, the improved deep FM model is compressed by a knowledge distillation model compression method, a smaller student model is obtained, model deployment is facilitated, and computing resources are saved.
Based on the same inventive concept, fig. 7 exemplarily illustrates a system for network qos prediction according to an embodiment of the present invention, which may be a flow of a method for network qos prediction.
The system, comprising:
an obtaining module 701, configured to obtain user data and service data;
a processing module 702, configured to input the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
Further, the processing module 702 is further configured to:
before the user data and the service data are input into a trained network service quality prediction model to obtain a network service quality prediction result, acquiring the sample data and a data label corresponding to the sample data;
constructing a deep FM model;
training the deep FM model according to the sample data and the data label to obtain the trained deep FM model and the soft label;
taking the trained DeepFM model as a teacher model under a knowledge distillation frame and constructing a student model;
and training the student model according to the sample data, the data label and the soft label to obtain a trained network service quality prediction model.
Further, the processing module 702 is specifically configured to:
and improving the pooling layer of the DeepFM model by adopting a self-adaptive pooling method to obtain an improved DeepFM model.
Further, the processing module 702 is specifically configured to:
constructing a fully-connected neural network according to the trained DeepFM model; the hidden layer of the fully-connected neural network comprises an extension layer and two fully-connected layers;
and determining the student model according to the fully-connected neural network.
Further, the processing module 702 is specifically configured to:
training the student model according to the sample data and the data labels to obtain a first student model;
training the first student model according to the sample data and the soft label to obtain a second student model;
and taking the second student model as a trained network service quality prediction model.
Further, the processing module 702 is specifically configured to:
determining a first characteristic corresponding to the user training data according to the user training data; the first characteristic comprises a user number, an IP address, a country of the user, a number of an autonomous system, and a longitude or latitude of a user party;
determining a second characteristic corresponding to the service training data according to the service training data; the second characteristic comprises the number of a network service party, a network service uniform resource positioning system, a service provider, an IP address, a country of the network service party, the number of an autonomous system, and longitude or latitude.
Based on the same inventive concept, another embodiment of the present invention provides an electronic device, which specifically includes the following components, with reference to fig. 8: a processor 801, a memory 802, a communication interface 803, and a communication bus 804;
the processor 801, the memory 802 and the communication interface 803 complete mutual communication through the communication bus 804; the communication interface 803 is used for realizing information transmission between devices;
the processor 801 is configured to call a computer program in the memory 802, and the processor implements all the steps of the method for predicting the network service quality when executing the computer program, for example, the processor implements the following steps when executing the computer program: acquiring user data and service data; inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
Based on the same inventive concept, a further embodiment of the present invention provides a non-transitory computer-readable storage medium, having stored thereon a computer program, which when executed by a processor implements all the steps of the above-mentioned method for network quality of service prediction, for example, the processor implements the following steps when executing the computer program: acquiring user data and service data; inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
In addition, the logic instructions in the memory may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a system for predicting network service quality, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above-described system embodiments are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on such understanding, the above technical solutions may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for enabling a computer device (which may be a personal computer, a system for predicting network service quality, or a network device, etc.) to execute the method for predicting network service quality according to the embodiments or some parts of the embodiments.
In addition, in the present invention, terms such as "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Moreover, in the present invention, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Furthermore, in the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for network quality of service prediction, comprising:
acquiring user data and service data;
inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result;
the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
2. The method of claim 1, wherein before the inputting the user data and the service data into the trained network service quality prediction model to obtain the network service quality prediction result, the method further comprises:
acquiring the sample data and a data tag corresponding to the sample data;
constructing a deep FM model;
training the deep FM model according to the sample data and the data label to obtain the trained deep FM model and the soft label;
taking the trained DeepFM model as a teacher model under a knowledge distillation frame and constructing a student model;
and training the student model according to the sample data, the data label and the soft label to obtain a trained network service quality prediction model.
3. The method of network QoS prediction according to claim 2, wherein the constructing a DeepFM model comprises:
and improving the pooling layer of the DeepFM model by adopting a self-adaptive pooling method to obtain an improved DeepFM model.
4. The method for predicting the network service quality according to claim 2, wherein the constructing a student model comprises:
constructing a fully-connected neural network according to the trained DeepFM model; the hidden layer of the fully-connected neural network comprises an extension layer and two fully-connected layers;
and determining the student model according to the fully-connected neural network.
5. The method of claim 2, wherein the training the student model according to the sample data, the data label and the soft label to obtain a trained network QoS prediction model comprises:
training the student model according to the sample data and the data labels to obtain a first student model;
training the first student model according to the sample data and the soft label to obtain a second student model;
and taking the second student model as a trained network service quality prediction model.
6. The method according to claim 2, wherein the obtaining the sample data and the data tag corresponding to the sample data comprises:
determining a first characteristic corresponding to the user training data according to the user training data; the first characteristic comprises a user number, an IP address, a country of the user, a number of an autonomous system, and a longitude or latitude of a user party;
determining a second characteristic corresponding to the service training data according to the service training data; the second characteristic comprises the number of a network service party, a network service uniform resource positioning system, a service provider, an IP address, a country of the network service party, the number of an autonomous system, and longitude or latitude.
7. A system for network quality of service prediction, comprising:
the acquisition module is used for acquiring user data and service data;
the processing module is used for inputting the user data and the service data into a trained network service quality prediction model to obtain a network service quality prediction result; the trained network service quality prediction model is obtained by training different sample data and corresponding soft labels; the sample data comprises user training data and service training data; and the soft label is prediction data output by the trained deep FM model after the sample data is input.
8. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method according to any one of claims 1 to 6 when executed by a processor.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 6 are implemented when the processor executes the program.
10. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110956949.9A CN113762737A (en) | 2021-08-19 | 2021-08-19 | Method and system for predicting network service quality |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110956949.9A CN113762737A (en) | 2021-08-19 | 2021-08-19 | Method and system for predicting network service quality |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113762737A true CN113762737A (en) | 2021-12-07 |
Family
ID=78790588
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110956949.9A Pending CN113762737A (en) | 2021-08-19 | 2021-08-19 | Method and system for predicting network service quality |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113762737A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114760191A (en) * | 2022-05-24 | 2022-07-15 | 咪咕文化科技有限公司 | Data service quality early warning method, system, device and readable storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110046952A (en) * | 2019-01-30 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
CN110674880A (en) * | 2019-09-27 | 2020-01-10 | 北京迈格威科技有限公司 | Network training method, device, medium and electronic equipment for knowledge distillation |
CN112367273A (en) * | 2020-10-30 | 2021-02-12 | 上海瀚讯信息技术股份有限公司 | Knowledge distillation-based flow classification method and device for deep neural network model |
-
2021
- 2021-08-19 CN CN202110956949.9A patent/CN113762737A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110046952A (en) * | 2019-01-30 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
CN110674880A (en) * | 2019-09-27 | 2020-01-10 | 北京迈格威科技有限公司 | Network training method, device, medium and electronic equipment for knowledge distillation |
CN112367273A (en) * | 2020-10-30 | 2021-02-12 | 上海瀚讯信息技术股份有限公司 | Knowledge distillation-based flow classification method and device for deep neural network model |
Non-Patent Citations (2)
Title |
---|
MICROSTRONG: "深度学习中的知识蒸馏技术(下)-知识蒸馏与推荐系统", Retrieved from the Internet <URL:https://mp.weixin.qq.com/s/Noac4YLIimr1HM2fln2bjg> * |
杨杨等: "SRI-XDFM A Service Reliability Inference Method Based on Deep Neural Network", INTELLIGENT AUTOMATION & SOFT COMPUTING, vol. 1, no. 26, pages 1459 - 1475 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114760191A (en) * | 2022-05-24 | 2022-07-15 | 咪咕文化科技有限公司 | Data service quality early warning method, system, device and readable storage medium |
CN114760191B (en) * | 2022-05-24 | 2023-09-19 | 咪咕文化科技有限公司 | Data service quality early warning method, system, equipment and readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112035743B (en) | Data recommendation method and device, computer equipment and storage medium | |
CN111541570A (en) | Cloud service QoS prediction method based on multi-source feature learning | |
CN116664719B (en) | Image redrawing model training method, image redrawing method and device | |
CN109086463B (en) | Question-answering community label recommendation method based on regional convolutional neural network | |
CN113255798A (en) | Classification model training method, device, equipment and medium | |
CN109189922B (en) | Comment evaluation model training method and device | |
CN114580794B (en) | Data processing method, apparatus, program product, computer device and medium | |
CN113065013A (en) | Image annotation model training and image annotation method, system, device and medium | |
CN112287965A (en) | Image quality detection model training method and device and computer equipment | |
Zhu et al. | Semantic image segmentation with shared decomposition convolution and boundary reinforcement structure | |
CN111832637A (en) | Distributed deep learning classification method based on alternative direction multiplier method ADMM | |
CN112131261A (en) | Community query method and device based on community network and computer equipment | |
CN113762737A (en) | Method and system for predicting network service quality | |
CN114065033A (en) | Training method of graph neural network model for recommending Web service combination | |
CN117216281A (en) | Knowledge graph-based user interest diffusion recommendation method and system | |
Varshaneya et al. | Res‐SE‐Net: Boosting Performance of ResNets by Enhancing Bridge Connections | |
CN113792110A (en) | Equipment trust value evaluation method based on social networking services | |
CN111310996B (en) | User trust relation prediction method and system based on graph self-coding network | |
CN116975686A (en) | Method for training student model, behavior prediction method and device | |
Jin et al. | Blind image quality assessment for multiple distortion image | |
CN109299291A (en) | A kind of Ask-Answer Community label recommendation method based on convolutional neural networks | |
CN114580388A (en) | Data processing method, object prediction method, related device and storage medium | |
CN115203516A (en) | Information recommendation method, device, equipment and storage medium based on artificial intelligence | |
CN114493674A (en) | Advertisement click rate prediction model and method | |
CN114462571A (en) | Deep learning model training method, data processing method and device |
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 |