CN109120961A - The prediction technique of the QoE of IPTV unbalanced dataset based on PNN-PSO algorithm - Google Patents
The prediction technique of the QoE of IPTV unbalanced dataset based on PNN-PSO algorithm Download PDFInfo
- Publication number
- CN109120961A CN109120961A CN201810803152.3A CN201810803152A CN109120961A CN 109120961 A CN109120961 A CN 109120961A CN 201810803152 A CN201810803152 A CN 201810803152A CN 109120961 A CN109120961 A CN 109120961A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- pnn
- iptv
- qoe
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000013528 artificial neural network Methods 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims abstract description 3
- 210000002569 neuron Anatomy 0.000 claims description 39
- 239000002245 particle Substances 0.000 claims description 23
- 238000012549 training Methods 0.000 claims description 18
- 239000000284 extract Substances 0.000 claims description 8
- 230000002596 correlated effect Effects 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000000875 corresponding effect Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000002860 competitive effect Effects 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000012552 review Methods 0.000 claims description 3
- 210000005036 nerve Anatomy 0.000 claims 1
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000000605 extraction Methods 0.000 abstract 1
- 230000006872 improvement Effects 0.000 description 9
- 238000004891 communication Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 208000018910 keratinopathic ichthyosis Diseases 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
- H04N21/25866—Management of end-user data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5003—Managing SLA; Interaction between SLA and QoS
- H04L41/5009—Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N17/00—Diagnosis, testing or measuring for television systems or their details
- H04N17/004—Diagnosis, testing or measuring for television systems or their details for digital television systems
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
- H04N21/25866—Management of end-user data
- H04N21/25891—Management of end-user data being end-user preferences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4662—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms
- H04N21/4666—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms using neural networks, e.g. processing the feedback provided by the user
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
Abstract
The invention proposes a kind of prediction techniques of the user experience quality (QoE) of IPTV unbalanced dataset based on PNN-PSO algorithm.This method comprises the following steps: extracting the related influence factor of Quality of experience with user first from the set-top box recording of IPTV, obtains the feature input of PNN-PSO model, and carry out the definition of related notion and the extraction of data and pretreatment；Then being associated with for each feature and user experience quality (QoE) is analyzed；It is then based on PNN-PSO neural network user experience quality (QoE) prediction model, the prediction of user experience quality (QoE) is realized using IPTV data set.Using this method, overcomes local minimum problem present in PNN algorithm and the unbalanced brought problem of data, study convergence process faster, are predicted more accurate, efficient；The data set of IPTV set top box is taken full advantage of, IPTV operator can be helped to adjust measure in time, to improve user experience quality.
Description
Technical field
The present invention relates to user experience quality (QoE) forecast analysis technical fields in Interactive Internet TV (IPTV), especially
It is related to a kind of QoE prediction technique of IPTV unbalanced dataset based on PNN-PSO algorithm.
Background technique
Media stream media services have become the main reason for global mobile data flow newest growth behind.According to system
Meter, until 2018, the summation of the video (TV, video on demand (VoD), internet and P2P) of form of ownership will be more than complete
The 80% of ball customer flow.Video flowing on internet is becoming increasingly popular.To the end of the year 2018, mobile video all over the world
For the traffic by the one third for the mobile data flow for being more than, wherein most mobile video flow is wireless network (WiFi),
Rather than mobile communication (3G, 4G or 5G network).Meanwhile network protocol television (IPTV) is quickly grown in radio honeycomb communication
Stage comes into being.IPTV is integrated with the multiple technologies such as internet, multimedia, communication.It provides the interaction based on communication network
Formula Video service provides multiple choices for user, to guarantee the satisfaction of user.
Due to measuring and predicting that QoE is a challenging task under dynamic network condition, International Telecommunication Association is proposed
No. 1011 suggestions provide reference guide for QoE appraisal procedure, this service provider is adjusted in time service have it is very big
Benefit.Since performance indicator (KPIs) can be used to verify whether network can satisfy the quantisation metric standard of target user, institute
KPIs can be mapped as to QoE value.But only consider that performance indicator can not adequately embody the experience matter of user
Amount may also need the Characteristic of Interest in view of user itself.And with the optimization of network, more and more users are satisfied
Their QoE may only have only a few just to will appear unsatisfied situation in special circumstances, this will lead to the unbalanced of data
Property.The situation is totally unfavorable for the prediction of QoE.
In consideration of it, it is necessory to propose that a kind of QoE of improved IPTV unbalanced dataset based on PNN-PSO algorithm is pre-
Survey method, to solve the above problems.
Summary of the invention
The purpose of the present invention is to provide a kind of IPTV based on PNN-PSO algorithm for improving user QoE prediction accuracy
The prediction technique of the QoE of unbalanced dataset.
To achieve the goals above, the present invention adopts the following technical scheme: the IPTV imbalance number based on PNN-PSO algorithm
According to the prediction technique of the QoE of collection, comprising the following steps:
S1 acquires user data from IPTV set top box, extracts from user data relevant with the Quality of experience of user
Influence factor data；
S2 analyzes the correlation degree between each influence factor and the Quality of experience of user；
S3 establishes the prediction model of the Quality of experience of user based on PNN-PSO neural network algorithm；
S4 is trained the prediction with the Quality of experience of user in the prediction model.
Technical solution as a further improvement of that present invention, the step S3 include the following steps:
What S31, definition step S1 were extracted is input vector X with the related influence factor of Quality of experience of user, according to
The dimension of the input vector X determines the quantity of neuron in the input layer of PNN neural network；
S32, the input vector X are transmitted to the input layer, extract several datas at random in the user data
As sample data, default number of branches are extracted according to as training sample, the input layer receives the training sample and passes data
It is delivered to hidden layer；The hidden layer is radial base, and the neuron node of each hidden layer has a center, the center pair
Answer a sample data；
S33 determines the neuron number of the hidden layer according to the number of the training sample；The hidden layer calculates institute
After input vector X is stated at a distance from the corresponding center, a scalar value is exported；The jth of i-th quasi-mode in the hidden layer
The output scalar value Φ of neuronij(X) meet following relationship with the input vector X:
Wherein, the σ is the expansion rate of radial basis function；I=1 ..., M, M are total class number in training sample；D is
The dimension of sample space data, XijIt is j-th of center of the i-th quasi-mode；
S34 is optimized the expansion rate σ of the radial basis function using particle swarm algorithm, searches for most suitable propagation automatically；
S35, data are transferred to summation layer；The neuron number of the summation layer is determined according to total classification number M；It is described
Summation layer exports v after being weighted and averaged the output scalar value for belonging to of a sort neuron in the hidden layeri, export vi
Meet following relational expression:
Wherein, viIndicate the output of the i-th class classification；N indicates the number of the i-th class neuron；
S36, data are transferred to output layer；The output layer is made of competition neurons, and neuron number is asked with described
It is identical with the neuron number of layer；The output layer receives the output of the summation layer and does threshold value discrimination, and output y meets such as
Lower relational expression:
Y=compet (vi)
Wherein, competitive function compet (x) expression finds one in all output layer neurons has maximum a posteriori
The neuron of probability density, output are 1, and the output of remaining neuron is 0.
Technical solution as a further improvement of that present invention, specific step is as follows by the step S32:
S321 randomly chooses 10000,15000,20000 records respectively from the user data, is respectively formed data
Collection 1, data set 2 and data set 3；
S322 determines the input vector of the data set 1, data set 2 and data set 3；Using ten folding cross validations by institute
It states data set 1, data set 2 and data set 3 and is respectively split as ten packets, randomly select nine packets in each data set as instruction
Practice sample, a remaining packet is used to predict the Quality of experience of user；
S323, the input layer receive the training sample and pass it to hidden layer.
Technical solution as a further improvement of that present invention, in step S31, according to relevant with the Quality of experience of user
The number of influence factor determine the dimension of the input vector X.
Technical solution as a further improvement of that present invention, the step S34 include the following steps:
S341 determines the fitness function in particle swarm algorithm, and determines adaptive value to judge current position according to it
Quality；The fitness function is for judging the comprehensive measurement index G-mean function of imbalanced data sets prediction accuracy；It is described
G-mean function meets following relational expression:
Wherein, TP is positive the quantity that class is correctly classified；FP is that the prediction class that is positive really is negative the quantity of class；TN is being negative class just
The quantity really classified；FN is that the prediction class that is negative really is positive the quantity of class；
S342, by the speed and position of continuous iteration and adjustment population, finding makes G-mean function be optimal value
σ；In the Optimization Steps of each step, the speed of each particle and position are updated according to following relational expression:
vk(t+1)=vk(t)+c1(pbestk-xk)+c2(gbest-xk)
xk(t+1)=xk(t)+vk(t+1)
Wherein, t represents t suboptimization；K represents k-th of particle；K=1,2 ..., L, L indicate the quantity of particle；vk、xkPoint
The speed and position of k-th of particle are not represented；c1、c2Representative is distributed in the uniform random variable on [0,2]；pbestkRepresent grain
The optimum position that sub- k is passed through；Gbest represents any one particle optimum position experienced in all populations.
Technical solution as a further improvement of that present invention sets optimization number as 30 in the step S342；Particle
Number is 10；c1、c2It is 2, initial velocity range is [- 0.5,0.5].
Technical solution as a further improvement of that present invention, the user data include satisfied data and unsatisfied number
According to total class number M is 2；The neuron number of the output layer and the summation layer is 2.
Technical solution as a further improvement of that present invention passes through correlation coefficient ρ in step s 2X,YMeasure the body of user
The correlation degree of the amount of checking the quality and each influence factor, the correlation coefficient ρX,YMeet following relational expression:
Wherein, Y is the Quality of experience of user；X is any one in the influence factor；σX、σYThe respectively mark of X, Y
It is quasi- poor；Cov (X, Y) is covariance function.
Technical solution as a further improvement of that present invention, the influence factor include it is relevant to IPTV service it is objective because
Subjective factor plain and related to user；The objective factor include abnormal end number, packet loss, Media Loss Ratio, shake,
Delay and mean bit rate；The subjective factor includes viewing type and channel popularity；The viewing type include live streaming,
It reviews and program request；The channel popularity is judged by accumulated view time.
Technical solution as a further improvement of that present invention, the body of the mean bit rate and the viewing type and user
The amount of checking the quality is positively correlated；The abnormal end number, packet loss, Media Loss Ratio, shake, delay and channel popularity and user
Quality of experience it is negatively correlated.
The beneficial effects of the present invention are: the present invention is using predicting unbalanced data based on PNN-PSO neural network model
The prediction technique of the user experience quality (QoE) of concentration, uses using G-mean as the PSO algorithm of objective function and adjusts in PNN
Unique variable σ, overcome the tune ginseng process for needing Heuristics and iteration to attempt in PNN algorithm.The model can be effective
The accuracy of the user QoE prediction in imbalanced data sets is promoted, and without local minimum problem present in PNN algorithm, study
Fast convergence rate.This method comprehensively considers the subjective factor of user and the objective factor of network, and analyzes user experience accordingly
Correlation between quality (QoE) and each influent factor.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the QoE prediction technique of the IPTV unbalanced dataset of PNN-PSO algorithm.
Fig. 2 is that the present invention is based on a kind of embodiment party of the QoE prediction technique of the IPTV unbalanced dataset of PNN-PSO algorithm
The schematic diagram of formula.
Fig. 3 is that the present invention is based on the convergence property of the QoE prediction technique of the IPTV unbalanced dataset of PNN-PSO algorithm songs
Line.
Fig. 4 be the present invention is based on the PNN-PSO of the QoE prediction technique of the IPTV unbalanced dataset of PNN-PSO algorithm with
The prediction result of BPNN, PNN and GRNN compare.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and detailed description
The present invention will be described in detail.
It please refers to shown in Fig. 1 to Fig. 2, a kind of prediction side of the QoE of the IPTV unbalanced dataset based on PNN-PSO algorithm
Method, comprising the following steps:
S1 acquires user data from IPTV set top box, extracts from user data relevant with the Quality of experience of user
Influence factor data；
S2 analyzes the correlation degree between each influence factor and the Quality of experience of user；
S3 establishes the prediction model of the Quality of experience of user based on PNN-PSO neural network algorithm；
S4 is trained the prediction with the Quality of experience of user in the prediction model.
In step sl, it includes 100000 records that user data is acquired from IPTV set top box, wherein most of
User is dissatisfied all in satisfactory state, only small part user, belongs to imbalanced data sets；Define total class number of user data
M is 2；It needs to remove the data for wherein having missing values, and therefrom extracts the related influence factor of Quality of experience with user
Data.
The influence factor includes objective factor relevant to IPTV service and the (interest and habit of user related with user
It is used) subjective factor.The objective factor includes abnormal end number (Abnum), packet loss (Lpr), Media Loss Ratio
(Mlr), (Df), delay (Jitter) and mean bit rate (Avg_bit_rate) are shaken.
The subjective factor includes viewing type (type) and channel popularity (cha_pop).The viewing type includes
It is broadcast live, reviews and program request；It indicated to be broadcast live, reviewed and program request respectively with 0,1,2.The channel popularity is seen by accumulative
See that the time judges, that is, the pouplarity of each channel is indicated with accumulated view time, n, rear n before being marked respectively with 1,3,2
And other records.In S2, pass through correlation coefficient ρX,YMeasure the Quality of experience of user and being associated with for each influence factor
Degree, the correlation coefficient ρX,YMeet following relational expression:
Wherein, Y is the Quality of experience of user；X is any one in the influence factor；σX、σYThe respectively mark of X, Y
It is quasi- poor；Cov (X, Y) is covariance function；That is, the correlation coefficient ρX,YEqual to variable covariance cov (X, Y) divided by respective
The product σ of standard deviationXσY.The ρX,YValue between -1 and 1, be used to describe two groups of linear data and change shifting together
Dynamic trend；When the enhancing of the linear relationship of two variables, related coefficient tends to 1 or -1；When a variable increases, another becomes
When amount also increases, show to be positively related between them, related coefficient is greater than 0；If a variable increases, another variable subtracts
It is small, show to be negatively correlated between them, related coefficient is less than 0；If related coefficient is equal to 0, show it is linear between them
Unrelated.
According to the calculating correlation coefficient ρX,YIt is found that the experience of the mean bit rate and the viewing type and user
Quality is positively correlated；The abnormal end number, packet loss, Media Loss Ratio, shake, delay and channel popularity and user's
Quality of experience is negatively correlated.
The step S3 includes the following steps:
S31, definition step S1 extract with the related influence factor of Quality of experience of user (8 influence factors, i.e., 8
A input feature vector) it is input vector X, neuron in the input layer of PNN neural network is determined according to the dimension of the input vector X
Quantity.That is, the dimension of the input vector X is determined according to the number with the related influence factor of the Quality of experience of user,
That is, the dimension of the input vector is 8.
S32, the input vector X are transmitted to the input layer, extract several datas at random in the user data
As sample data, default number of branches are extracted according to as training sample；The input layer, which is undergone training, sample and to be passed data to
Hidden layer；The hidden layer is radial base, and the neuron node of each hidden layer has a center, the center corresponding one
A sample data；
Specifically, specific step is as follows by the step S32:
S321 randomly chooses 10000,15000,20000 records respectively from the user data, is respectively formed data
Collect 1 (dataset1), data set 2 (dataset2) and data set 3 (dataset3)；
S322 determines the input vector of the data set 1, data set 2 and data set 3；Using ten folding cross validations by institute
It states data set 1, data set 2 and data set 3 and is respectively split as ten packets, randomly select nine packets in each data set as instruction
Practice sample, a remaining packet is used to predict the Quality of experience of user；
S323, the input layer receive the training sample and pass it to hidden layer.
S33 determines the neuron number of the hidden layer according to the number of the training sample value；The hidden layer calculates
After the input vector X is at a distance from the corresponding center, a scalar value is exported；I-th quasi-mode in the hidden layer
The output scalar value Φ of jth neuronij(X) meet following relationship with the input vector X:
Wherein, the σ is the expansion rate of radial basis function；I=1,2 ..., M, M are total class number in training sample；d
It is the dimension of sample space data, XijIt is j-th of center of the i-th quasi-mode；
S34 is optimized the expansion rate σ of the radial basis function using particle swarm algorithm, searches for most suitable propagation automatically；
The step S34 includes the following steps:
S341 determines the fitness function in particle swarm algorithm (PSO), and determines adaptive value to judge current position according to it
The quality set；The accuracy that user experience quality is predicted in imbalanced data sets is measured using G-mean function；The adaptation
Degree function is for judging the comprehensive measurement index G-mean function of imbalanced data sets prediction accuracy；The G-mean function meets
Following relational expression:
Wherein, TP is positive the quantity that class is correctly classified；FP is that the prediction class that is positive really is negative the quantity of class；TN is being negative class just
The quantity really classified；FN is that the prediction class that is negative really is positive the quantity of class；
S342, by the speed and position of continuous iteration and adjustment population, finding reaches objective function G-mean function
To the expansion rate σ of optimal value；In the Optimization Steps of each step, the parameter (speed and position) of each particle is according to such as ShiShimonoseki
It is that formula updates:
vk(t+1)=vk(t)+c1(pbestk-xk)+c2(gbest-xk)
xk(t+1)=xk(t)+vk(t+1)
Wherein, t represents t suboptimization；K represents k-th of particle；K=1,2 ..., L, L indicate the quantity of particle；vk、xkPoint
The speed and position of k-th of particle are not represented；c1、c2Representative is distributed in the uniform random variable on [0,2]；pbestkRepresent grain
The optimum position that sub- k is passed through；Gbest represents any one particle optimum position experienced in all populations.That is, grain
Son combine them before optimum position and the optimum position of neighbours can have higher G- to maximize them to some
The mobile probability of the area of space of mean.Specifically, in this embodiment, optimization number is set as 30；Population is 10；
c1、c2Studying factors are 2, and initial velocity range is [- 0.5,0.5].
The data of S35, hidden layer output are transferred to summation layer；The mind of the summation layer is determined according to total classification number M
Through first number, i.e. the neuron number of summation layer is 2；The summation layer will belong to of a sort neuron in the hidden layer
Output scalar value exports v after being weighted and averagedi, export viMeet following relational expression:
Wherein, viIndicate the output of the i-th class classification；N indicates the number of the i-th class neuron；
S36, data are transferred to output layer；The output layer is made of competition neurons, and neuron number is asked with described
It is identical with the neuron number of layer, that is, the neuron number of the output layer is 2；The output layer receives the summation layer
Threshold value discrimination is exported and does, output y meets following relational expression:
Y=compet (vi)
Wherein, competitive function compet (x) expression finds one in all output layer neurons has maximum a posteriori
The neuron of probability density, output are 1, and the output of remaining neuron is 0.
In step s 4, a remaining packet is used as prediction data packet in step S322, and prediction data bag data is transferred to
The PNN-PSO prediction model of foundation carries out QoE prediction, exports prediction result.
It please refers to shown in Fig. 3, fitness function (fitness-function) increases with the increase of the number of iterations, most
It tends towards stability eventually.Therefore, the number of iterations is set as 30 in step s 4, completes training and prediction.
Please refer to shown in Fig. 4, compared PNN-PSO model, BPNN, PNN, GRNN (BPNN, PNN, PNN-PSO and
GRNN from left to right successively arranges) G-mean value.It can be seen from the figure that the comparison result of three data sets is shown: this
The prediction technique of the QoE of IPTV unbalanced dataset of the invention based on PNN-PSO algorithm can be very well by using PNN-PSO model
Study imbalanced data sets in QoE characteristic, effectively increase prediction accuracy.
In conclusion the present invention is based on the prediction techniques of the QoE of the IPTV unbalanced dataset of PNN-PSO algorithm, first
Data prediction is carried out, removal there are the data of missing values, and the objective factor of the subjective factor and network that comprehensively consider user is selected
Influence factor related with user experience quality is taken, the input feature vector of model is obtained；Training is then completed using PNN-PSO model
And prediction.The model can effectively promote the accuracy of the prediction of the user QoE in imbalanced data sets, and without in PNN algorithm
Existing local minimum problem learns fast convergence rate.This method comprehensively consider user subjective factor and network it is objective because
Element, and the correlation between user experience quality (QoE) and each influent factor is analyzed accordingly.
Above embodiments are merely to illustrate the present invention and not limit the technical scheme described by the invention, to this specification
Understanding should based on person of ordinary skill in the field, although this specification referring to the above embodiments to the present invention
Detailed description is had been carried out, still, those skilled in the art should understand that, person of ordinary skill in the field is still
Can so modify or equivalently replace the present invention, and all do not depart from the spirit and scope of the present invention technical solution and
It is improved, and should all be covered in scope of the presently claimed invention.
Claims (10)
1. a kind of prediction technique of the QoE of the IPTV unbalanced dataset based on PNN-PSO algorithm, comprising the following steps:
S1 acquires user data from IPTV set top box, and the related shadow of Quality of experience with user is extracted from user data
The data of the factor of sound；
S2 analyzes the correlation degree between each influence factor and the Quality of experience of user；
S3 establishes the prediction model of the Quality of experience of user based on PNN-PSO neural network algorithm；
S4 is trained the prediction with the Quality of experience of user in the prediction model.
2. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 1 based on PNN-PSO algorithm,
It is characterized in that, the step S3 includes the following steps:
What S31, definition step S1 were extracted is input vector X with the related influence factor of Quality of experience of user, according to described
The dimension of input vector X determines the quantity of neuron in the input layer of PNN neural network；
S32, the input vector X are transmitted to the input layer, extract several data conducts at random in the user data
Sample data extracts default number of branches according to as training sample, and the input layer receives the training sample and passes data to
Hidden layer；The hidden layer is radial base, and the neuron node of each hidden layer has a center, the center corresponding one
A sample data；
S33 determines the neuron number of the hidden layer according to the number of the training sample；The hidden layer calculates described defeated
After incoming vector X is at a distance from the corresponding center, a scalar value is exported；The jth nerve of i-th quasi-mode in the hidden layer
The output scalar value Φ of memberij(X) meet following relationship with the input vector X:
Wherein, the σ is the expansion rate of radial basis function；I=1 ..., M, M are total class number in training sample；D is sample
The dimension of spatial data, XijIt is j-th of center of the i-th quasi-mode；
S34 is optimized the expansion rate σ of the radial basis function using particle swarm algorithm, searches for most suitable propagation automatically；
S35, data are transferred to summation layer；The neuron number of the summation layer is determined according to total classification number M；The summation
Layer exports v after being weighted and averaged the output scalar value for belonging to of a sort neuron in the hidden layeri, export viMeet
Following relational expression:
Wherein, viIndicate the output of the i-th class classification；N indicates the number of the i-th class neuron；
S36, data are transferred to output layer；The output layer is made of competition neurons, neuron number and the summation layer
Neuron number it is identical；The output layer receives the output of the summation layer and does threshold value discrimination, and output y meets such as ShiShimonoseki
It is formula:
Y=compet (vi)
Wherein, competitive function compet (x) expression finds one in all output layer neurons has maximum a posteriori probability
The neuron of density, output are 1, and the output of remaining neuron is 0.
3. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 2 based on PNN-PSO algorithm,
It is characterized in that, specific step is as follows by the step S32:
S321 randomly chooses 10000,15000,20000 records respectively from the user data, be respectively formed data set 1,
Data set 2 and data set 3；
S322 determines the input vector of the data set 1, data set 2 and data set 3；Using ten folding cross validations by the number
It is respectively split as ten packets according to collection 1, data set 2 and data set 3, randomly selects nine packets in each data set as training sample
This, a remaining packet is used to predict the Quality of experience of user；
S323, the input layer receive the training sample and pass it to hidden layer.
4. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 2 based on PNN-PSO algorithm,
Be characterized in that: in step S31, according to the number with the related influence factor of the Quality of experience of user determine it is described input to
Measure the dimension of X.
5. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 2 based on PNN-PSO algorithm,
It is characterized in that, the step S34 includes the following steps:
S341 determines the fitness function in particle swarm algorithm, and determines adaptive value according to it to judge the quality of current position；
The fitness function is for judging the comprehensive measurement index G-mean function of imbalanced data sets prediction accuracy；The G-mean
Function meets following relational expression:
Wherein, TP is positive the quantity that class is correctly classified；FP is that the prediction class that is positive really is negative the quantity of class；The TN class that is negative correctly is divided
The quantity of class；FN is that the prediction class that is negative really is positive the quantity of class；
S342 finds the σ for making G-mean function be optimal value by the speed and position of continuous iteration and adjustment population；
In the Optimization Steps of each step, the speed of each particle and position are updated according to following relational expression:
vk(t+1)=vk(t)+c1(pbestk-xk)+c2(gbest-xk)
xk(t+1)=xk(t)+vk(t+1)
Wherein, t represents t suboptimization；K represents k-th of particle；K=1,2 ..., L, L indicate the quantity of particle；vk、xkGeneration respectively
The speed of k-th of particle of table and position；c1、c2Representative is distributed in the uniform random variable on [0,2]；pbestkRepresent particle k institute
The optimum position of process；Gbest represents any one particle optimum position experienced in all populations.
6. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 5 based on PNN-PSO algorithm,
It is characterized in that: in the step S342, setting optimization number as 30；Population is 10；c1、c2It is 2, initial velocity range
For [- 0.5,0.5].
7. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 3 based on PNN-PSO algorithm,
Be characterized in that: the user data includes satisfied data and unsatisfied data, and total class number M is 2；The output layer and institute
The neuron number for stating summation layer is 2.
8. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 1 based on PNN-PSO algorithm,
It is characterized in that, in step s 2, passes through correlation coefficient ρX,YMeasure the Quality of experience of user and being associated with for each influence factor
Degree, the correlation coefficient ρX,YMeet following relational expression:
Wherein, Y is the Quality of experience of user；X is any one in the influence factor；σX、σYThe respectively standard deviation of X, Y；
Cov (X, Y) is covariance function.
9. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 8 based on PNN-PSO algorithm,
Be characterized in that: the influence factor includes objective factor relevant to IPTV service and subjective factor related to user；It is described
Objective factor includes abnormal end number, packet loss, Media Loss Ratio, shake, delay and mean bit rate；It is described it is subjective because
Element includes viewing type and channel popularity；The viewing type includes live streaming, reviews and program request；The channel popularity is logical
Cross accumulated view time judge.
10. the prediction technique of the QoE of the IPTV unbalanced dataset according to claim 9 based on PNN-PSO algorithm,
Be characterized in that: the Quality of experience of the mean bit rate and the viewing type and user are positively correlated；The abnormal end number,
Packet loss, Media Loss Ratio, shake, delay and channel popularity and the Quality of experience of user are negatively correlated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810803152.3A CN109120961B (en) | 2018-07-20 | 2018-07-20 | QoE prediction method of IPTV unbalanced data set based on PNN-PSO algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810803152.3A CN109120961B (en) | 2018-07-20 | 2018-07-20 | QoE prediction method of IPTV unbalanced data set based on PNN-PSO algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109120961A true CN109120961A (en) | 2019-01-01 |
| CN109120961B CN109120961B (en) | 2020-11-03 |
Family
ID=64863212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810803152.3A Active CN109120961B (en) | 2018-07-20 | 2018-07-20 | QoE prediction method of IPTV unbalanced data set based on PNN-PSO algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109120961B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111797937A (en) * | 2020-07-15 | 2020-10-20 | 东北大学 | Greenhouse environment assessment method based on PNN network |
| WO2021026944A1 (en) * | 2019-08-09 | 2021-02-18 | 东北大学 | Adaptive transmission method for industrial wireless streaming media employing particle swarm and neural network |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106869990A (en) * | 2017-03-02 | 2017-06-20 | 新疆大学 | Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms |
-
2018
- 2018-07-20 CN CN201810803152.3A patent/CN109120961B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106869990A (en) * | 2017-03-02 | 2017-06-20 | 新疆大学 | Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms |
Non-Patent Citations (1)
| Title |
|---|
| 贺媛: ""无线业务QOE建模及分析"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021026944A1 (en) * | 2019-08-09 | 2021-02-18 | 东北大学 | Adaptive transmission method for industrial wireless streaming media employing particle swarm and neural network |
| CN111797937A (en) * | 2020-07-15 | 2020-10-20 | 东北大学 | Greenhouse environment assessment method based on PNN network |
| CN111797937B (en) * | 2020-07-15 | 2023-06-13 | 东北大学 | Greenhouse environment assessment method based on PNN network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109120961B (en) | 2020-11-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106228386B (en) | A kind of information-pushing method and device | |
| CN108419274B (en) | Heterogeneous wireless network selection method based on utility function | |
| CN105635762B (en) | A kind of video temperature prediction technique and its system based on deepness belief network | |
| CN103546773A (en) | Television program recommendation method and system | |
| US7696427B2 (en) | Method and system for recommending music | |
| CN111211994A (en) | Network traffic classification method based on SOM and K-means fusion algorithm | |
| CN111563275A (en) | Data desensitization method based on generation countermeasure network | |
| CN106791964B (en) | Broadcast TV program recommender system and method | |
| WO2021129464A1 (en) | Channel identification method and apparatus, transmission method, transmission device, base station, and medium | |
| WO2020135642A1 (en) | Model training method and apparatus employing generative adversarial network | |
| CN101675431A (en) | Method of organising content items | |
| CN109151953B (en) | Network access selection calculation method based on user and network bilateral benefits | |
| CN109120961A (en) | The prediction technique of the QoE of IPTV unbalanced dataset based on PNN-PSO algorithm | |
| CN108132964A (en) | A kind of collaborative filtering method to be scored based on user item class | |
| CN109831705A (en) | A kind of subjective QoE appraisal procedure for HTTP video stream traffic | |
| CN109451352A (en) | A kind of video playing method for predicting and device | |
| CN109862431A (en) | A kind of TV programme mixed recommendation method based on MCL-HCF algorithm | |
| CN105872792A (en) | Voice-based service recommending method and device | |
| CN110933473A (en) | Video playing heat determining method and device | |
| CN108650532A (en) | Catv on demand program commending method and system | |
| CN109508407A (en) | The tv product recommended method of time of fusion and Interest Similarity | |
| KR100868764B1 (en) | Method and system of recommending a music using user model, and update method of a conditional user model | |
| CN115277354B (en) | Fault detection method for command control network management system | |
| CN106604068B (en) | A kind of method and its system of more new media program | |
| CN106056160B (en) | User fault reporting prediction method under unbalanced IPTV data set |
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 | ||
| CB02 | Change of applicant information |
Address after: Room 201, building 2, phase II, No.1 Kechuang Road, Yaohua street, Qixia District, Nanjing City, Jiangsu Province Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS Address before: 210003, No. 66, new exemplary Road, Nanjing, Jiangsu Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20190101 Assignee: NANJING NANYOU INSTITUTE OF INFORMATION TECHNOVATION Co.,Ltd. Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2021980014141 Denomination of invention: QoE prediction method of IPTV unbalanced data set based on pnn-pso algorithm Granted publication date: 20201103 License type: Common License Record date: 20211206 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EC01 | Cancellation of recordation of patent licensing contract |
Assignee: NANJING NANYOU INSTITUTE OF INFORMATION TECHNOVATION Co.,Ltd. Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2021980014141 Date of cancellation: 20231107 |
|
| EC01 | Cancellation of recordation of patent licensing contract |