CN106897733A - Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm - Google Patents
Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm Download PDFInfo
- Publication number
- CN106897733A CN106897733A CN201710032385.3A CN201710032385A CN106897733A CN 106897733 A CN106897733 A CN 106897733A CN 201710032385 A CN201710032385 A CN 201710032385A CN 106897733 A CN106897733 A CN 106897733A
- Authority
- CN
- China
- Prior art keywords
- video
- feature
- optimization algorithm
- swarm optimization
- particle
- 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
Classifications
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- 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]
Abstract
The invention discloses a kind of video stream characteristics selection based on particle swarm optimization algorithm and sorting technique, the feature selecting of the method is first by calculating difference of the feature in similar neighbour's sample and foreign peoples's neighbour's sample come the separating capacity of measures characteristic, removal part extraneous features, reach the purpose of quick dimensionality reduction, then using particle swarm optimization algorithm as searching algorithm, and optimal subset is selected in residue character subset as evaluation function using inconsistent rate as the initial population of population with the larger part excellent characteristic of feature weight.The method has lower computation complexity relative to existing particle swarm optimization algorithm, can effectively reduce the computation complexity in feature selection process.Meanwhile, the present invention using design three layers of SVM cascade sorter model to online SD video, online HD video, online seven kinds of business such as super clear video, online live video classify.Experiment shows that the inventive method obtains more preferable classification performance than existing method.
Description
Technical field
The invention belongs to pattern recognition and classification technical field, more particularly to a kind of video based on particle swarm optimization algorithm
Stream feature selecting and sorting technique.
Background technology
With developing rapidly for internet and stream media technology, the growth of the video traffic in network is particularly rapid.With this
Meanwhile, new application and the continuous quick appearance of agreement in network so that network environment is more complicated.Various types of networks should
With the rapidly increase with network traffics, greatly challenge is brought to Internet Service Provider.How network pipe is effectively carried out
The information security of reason, the service quality for ensureing different business and user has turned into problem in the urgent need to address.Taken for network
For business provider and network environment manager, rapidly and accurately identify that the different business stream in network is a kind of effective
Solution.
Traditional network traffics identification and sorting technique mainly have the method based on port, the side based on deep-packet detection
Method.Recognition methods based on port is that the port numbers that the suggestion of member management office is acted on behalf of according to Internet distinguish different nets
Network application, with the extensive use of dynamic end slogan so that the recognition efficiency of this method and the classification degree of accuracy be not high.Based on depth
The principle of the method for degree bag detection is the load by parsing packet, is compared with specific signature in known protocol, from
And distinguish different business.But the problems such as being protected with the popularization and privacy of user of network data encryption, causes based on depth
The net flow assorted method for wrapping detection is no longer applicable.Method based on statistical nature is by extracting the statistical nature pair of data flow
Data flow is classified.This method can both overcome the shortcoming of conventional method, and with Stability and veracity higher.Cause
This, network service flow identification field is widely applied to based on network statistical flow characteristic bonding machine device learning method.
Substantial amounts of statistical nature can be extracted from network service flow, how to select rational combinations of features is to improve classification essence
The key point of degree.Many studies have shown that, uncorrelated or redundancy the feature between feature can trigger over-fitting problem, Jin Eryan
Ghost image rings the accuracy of classification results.Meanwhile, the characteristic set of higher-dimension can also to grader bring substantial amounts of computing cost and when
Prolong.Therefore, the raising that selection is simple, combinations of features that is easily obtaining is to classifier performance plays the role of important.
The content of the invention
Statistical nature it is an object of the invention to be directed to network video service stream is selected and identification classification problem, is proposed
A kind of video stream characteristics selection and sorting technique based on particle swarm optimization algorithm, the method are (non-for online SD video
It is live), online HD video (non-live), online super clear video (non-live), online live video, HTTP download, IMU
Letter class video, seven kinds of business of P2P classes video are analyzed and study, and propose that a kind of video flowing based on particle swarm optimization algorithm is special
System of selection is levied, original video traffic is classified by three layers of SVM cascade grader.Test result indicate that, this hair
Bright method can obtain classification accuracy higher than existing congenic method.
To achieve the above object, technical scheme proposed by the present invention is that a kind of video flowing based on particle swarm optimization algorithm is special
Selection and sorting technique are levied, is comprised the following steps：
Step 1：Experimental data needed for being obtained using network package analysis software in open internet environment, then
Packet is filtered, finally these network video service streams is carried out with basic statistical nature and is calculated；
Step 2：Statistical nature to the above-mentioned video traffic for calculating is analyzed, and selecting can effectively distinguish industry
The combinations of features of business stream；
Step 3：Three layers of SVM cascade grader according to design carry out classification experiments to original video traffic, obtain
Final classification results.
Further, above-mentioned steps 1 are specifically included again：
Step 1-1：In open internet environment, the video traffic needed for being captured by network package analysis software
Then original data are simply pre-processed by data, are converted into what the five-tuple text formatting of standard, i.e. packet were reached
Time, source IP address, purpose IP address, agreement, packet packet size；
Step 1-2：Refer to filter the data lost interest in or will not classification results be produced with influence on Packet Filtering
Bag；
Step 1-3：Standard quintuple file to original video stream carries out basic statistical nature calculating, these feature bags
Include：Bag size, the average of bag size and variance, bag size information entropy, the average of inter-packet gap and variance, byte-rate, packet speed
The ratio between the ratio between rate, up-downgoing byte number, up-downgoing bag size.
Above-mentioned steps 2 are also specifically included：
Step 2-1：Statistical nature to all video traffics carries out discretization operations, in reduction feature selection process
Computing cost；
Step 2-2：The weight of each statistical nature is calculated using feature weight algorithm；
Step 2-3：According to the ranking of feature weight, remove part and the less feature of category associations, weight selection is maximum
N number of feature, reduce original feature space dimension, reduce subsequent operation computation complexity；
Step 2-3：In N number of character subset that previous step is chosen, the M feature conduct in the top of selected characteristic weight
Priori, instructs the initialization of population of particle swarm optimization algorithm, and the initial position of each particle is set into optimal location；Iteration
Number of times is set to 1；
Step 2-4：Using inconsistent rate as the fitness function of particle swarm optimization algorithm, grain is calculated using fitness function
The overall fitness of son, a pattern is referred to as by a combinations of features in sample instance, and all patterns of character subset are not
Consistent number, the total sample number for being equal to pattern appearance subtracts the sample number of the most a certain class label of occurrence number, inconsistent
Rate is equal to inconsistent number divided by total sample number；
Step 2-5：If the fitness of current particle is less than the fitness of particle itself optimal location, by particle itself most
Excellent location updating is current location；If the overall fitness of particle itself optimal location is less than the adaptation of the optimal location of population
Degree, particle itself optimal location is updated to by the optimal location of population；
Step 2-6：Position and velocity information according to current particle update position and the speed of population；
Step 2-7：If it is lasting constant in an iterative process to meet maximum iteration or inconsistent rate, export optimal
Solution；Otherwise, repeat step 2-5 to step 2-6.
Further, the N in above-mentioned steps 2-3 is preferably 10, M and is preferably 2.
Above-mentioned steps 3 can be specifically included：
Step 3-1：Original video Business Stream feature is selected using feature selection approach, and carries out ground floor SVM
Classification, obtains classification results C1, C2, C3, C4；Wherein, C1 is instant messaging class video, and C2 is P2P class videos, and C3 is under http
Carry, C4 is Online Video, comprising live and non-live two class；
Step 3-2：Reusing feature selection approach to the data flow characteristics of last layer classification results C4 carries out feature choosing
Select, and carry out second layer svm classifier, obtain classification results C41, C4；Wherein, C41 is online live video, and C42 is online non-
Live video；
Step 3-3：Reusing feature selection approach to the data flow characteristics of last layer classification results C42 carries out feature choosing
Select, and carry out third layer svm classifier, obtain classification results C421, C422, C433；Wherein, C421 is SD video, and C422 is
HD video, C423 is super clear video；
Step 3-4：Statistical classification output result.
Compared with prior art, beneficial outcomes of the invention：
1st, the video stream characteristics system of selection based on particle swarm optimization algorithm proposed by the present invention is based on compared to others
The feature selection approach of particle swarm optimization algorithm has lower computation complexity, in can effectively reducing feature selection process
Time and space expense, improve feature selecting efficiency.
2nd, the present invention devises a kind of three layers of SVM cascade grader to video traffic using the method for multistratum classification, coordinates
The combinations of features of feature selection approach selection proposed by the present invention, can obtain preferable classification results.
Brief description of the drawings
Fig. 1 is the FB(flow block) of video stream characteristics selection of the present invention based on particle swarm optimization algorithm and sorting technique.
Fig. 2 is effective proof diagram of the combinations of features selected by feature selection approach proposed by the present invention.
Specific embodiment
The invention is described in further detail below in conjunction with Figure of description.The present invention can be to video flowing industry
Simple, effective combinations of features is selected in business, and original video traffic is divided using three layers of SVM cascade grader
Class.Method flow is divided into following steps：
Step 1：Experimental data needed for being obtained using network package analysis software in open internet environment, then
Packet is filtered, finally these network video service streams is carried out with basic statistical nature and is calculated, concretely comprised the following steps：
Step 1-1：In open internet environment, the video traffic needed for being captured by network package analysis software
Then original data are simply pre-processed by data, are converted into what the five-tuple text formatting of standard, i.e. packet were reached
Time, source IP address, purpose IP address, agreement, packet packet size；
Step 1-2：Refer to filter the data lost interest in or will not classification results be produced with influence on Packet Filtering
Bag；
Step 1-3：Standard quintuple file to original video stream carries out basic statistical nature calculating, these feature bags
Include：Bag size, the average of bag size and variance, bag size information entropy, the average of inter-packet gap and variance, byte-rate, packet speed
The ratio between the ratio between rate, up-downgoing byte number, up-downgoing bag size.
Step 2：Statistical nature to video traffic is analyzed, select can effective differentiated service stream feature group
Close, concretely comprise the following steps：
Step 2-1：Statistical nature to all video traffics carries out discretization operations, in reduction feature selection process
Computing cost；
Step 2-2：The weight of each statistical nature is calculated using feature weight algorithm；
Step 2-3：According to the ranking of feature weight, remove part and the less feature of category associations, weight selection is maximum
N number of feature, reduce original feature space dimension, reduce subsequent operation computation complexity；
Step 2-4：In N number of character subset that previous step is chosen, the M feature conduct in the top of selected characteristic weight
Priori, instructs the initialization of population of particle swarm optimization algorithm, and the initial position of each particle is set into optimal location；Iteration
Number of times is set to 1；
Step 2-5：Using inconsistent rate as the fitness function of particle swarm optimization algorithm, grain is calculated using fitness function
The overall fitness of son, a pattern is referred to as by a combinations of features in sample instance, and all patterns of character subset are not
Consistent number, the total sample number for being equal to pattern appearance subtracts the sample number of the most a certain class label of occurrence number, inconsistent
Rate is equal to inconsistent number divided by total sample number；
Step 2-5：If the fitness of current particle is less than the fitness of particle itself optimal location, by particle itself most
Excellent location updating is current location；If the overall fitness of particle itself optimal location is less than the adaptation of the optimal location of population
Degree, particle itself optimal location is updated to by the optimal location of population；
Step 2-6：Position and velocity information according to current particle update position and the speed of population；
Step 2-7：If it is lasting constant in an iterative process to meet maximum iteration or inconsistent rate, export optimal
Solution；Otherwise, repeat step 2-5 to step 2-6.
Step 2-1：Statistical nature to all video traffics carries out discretization operations, in reduction feature selection process
Computing cost；
Step 2-2：The weight of each statistical nature is calculated using feature weight algorithm；
Step 2-3：According to the ranking of feature weight, remove part and the less feature of category associations, weight selection is maximum
10 features, reduce original feature space dimension, reduce subsequent operation computation complexity；
Step 2-4：In 10 character subsets that previous step is chosen, selected characteristic weight 2 features in the top are made
It is priori, instructs the initialization of population of particle swarm optimization algorithm, the initial position of each particle is set to optimal location；Repeatedly
Generation number is set to 1；
Step 2-5：Using inconsistent rate as the fitness function of particle swarm optimization algorithm, grain is calculated using fitness function
The overall fitness of son, a pattern is referred to as by a combinations of features in sample instance, and all patterns of character subset are not
Consistent number, the total sample number for being equal to pattern appearance subtracts the sample number of the most a certain class label of occurrence number, inconsistent
Rate is equal to inconsistent number divided by total sample number；
Step 2-5：If the fitness of current particle is less than the fitness of particle itself optimal location, by particle itself most
Excellent location updating is current location；If the overall fitness of particle itself optimal location is less than the adaptation of the optimal location of population
Degree, particle itself optimal location is updated to by the optimal location of population；
Step 2-6：Position and velocity information according to current particle update position and the speed of population；
Step 2-7：If it is lasting constant in an iterative process to meet maximum iteration or inconsistent rate, export optimal
Solution；Otherwise, repeat step 2-5 to step 2-6.
Step 3：Three layers of SVM cascade grader according to design carry out classification experiments to original video traffic, obtain
Final classification results, concretely comprise the following steps：
Step 3-1：Original video Business Stream feature is selected using feature selection approach, and carries out ground floor SVM
Classification, obtains classification results C1, C2, C3, C4；Wherein, C1 is instant messaging class video, and C2 is P2P class videos, and C3 is under http
Carry, C4 is Online Video, comprising live and non-live two class；
Step 3-2：Reusing feature selection approach to the data flow characteristics of last layer classification results C4 carries out feature choosing
Select, and carry out second layer svm classifier, obtain classification results C41, C4；Wherein, C41 is online live video, and C42 is online non-
Live video；
Step 3-3：Reusing feature selection approach to the data flow characteristics of last layer classification results C42 carries out feature choosing
Select, and carry out third layer svm classifier, obtain classification results C421, C422, C433；Wherein, C421 is SD video, and C422 is
HD video, C423 is super clear video；
Step 3-4：Statistical classification output result.
Above-mentioned steps are further described in detail in conjunction with drawings and Examples.
Step 1, acquisition and the statistical nature of network video service stream are calculated：Network is used in open internet environment
Package analysis software obtain needed for video traffic flow data, including online SD video (by taking youku SDs as an example),
Line HD video (by taking youku high definitions as an example), online super clear video (by youku it is super it is clear as a example by), online live video is (with Cbox
As a example by), HTTP downloads, instant messaging class video (by taking QQ as an example), seven kinds of video traffics of P2P videos (so that a sudden peal of thunder is looked at as an example).
Then the stream compression of acquisition is changed into time, source IP address, purpose that the five-tuple text formatting of standard, i.e. packet are reached
IP address, agreement, packet packet size.Finally the standard quintuple file to original video stream carries out basic statistical nature
Calculate.
Step 2, the video stream characteristics selection based on particle swarm optimization algorithm：Each is calculated first with feature weight algorithm
The feature weight of statistical nature, then the weight size according to feature, filters part extraneous features, so as to reach quick dimensionality reduction
Purpose.Then weight selection the best part feature instructs the initialization of population of particle swarm optimization algorithm as priori, choosing
Optimal character subset is selected in remaining character subset as fitness function with inconsistent rate.
We devise three layers of SVM cascade sorter model in an experiment, and this model can be in the grader of every one-level
The combinations of features selected using the inventive method identifies some certain types of applied business.The SVM classifier master of ground floor
Will be used for identify instant messaging class video (QQ), P2P classes video (Kankan), http download and other class data (network exists
Line video traffic and live broadcast service), optimal combinations of features is uplink packet size variance, downstream packets size information entropy.For side
Just observe, we have done operation of taking the logarithm to uplink packet size variance in Fig. 2 the first width figures.It can be seen that QQ business
There is obvious difference with other business in downstream packets size information entropy feature, http downloading services are also in uplink packet size side
There is obvious difference in difference feature with other business.Believed using uplink packet size variance and downstream packets size on two-dimensional space
A sudden peal of thunder can be looked at video traffic and Online Video business (including live and non-live) by breath entropy.
The SVM classifier of the second layer is mainly used to recognize live and non-live broadcast service.Optimal combinations of features：Overall bag is big
The ratio between small variance, ensemble average bag size and up-downgoing bag number.Observe for convenience, we are to overall in Fig. 2 the second width figures
Bag size variance and ensemble average bag size have done operation of taking the logarithm.From figure 2 it can be seen that using overall in three dimensions
The ratio between bag size variance, ensemble average bag size and up-downgoing bag number can effectively distinguish live and non-live broadcast service.
The SVM classifier of third layer is mainly used to recognize non-live SD, high definition and super clear.Optimal combinations of features：
Descending byte-rate, downlink data packet rate, descending average inter-packet gap, descending inter-packet gap variance.
Step 3, three layers of SVM cascade classification export statisticses, and its implementation is：Using above-mentioned three layers for designing
SVM cascade grader, to primitive network video traffic, implements multistratum classification.
Three layers of SVM cascade sorting technique of the invention includes：
Step 3-1：Original video Business Stream feature is selected using feature selection approach, and carries out ground floor SVM
Classification, obtains classification results C1, C2, C3, C4；Wherein, C1 is instant messaging class video, and C2 is P2P class videos, and C3 is under http
Carry, C4 is Online Video, comprising live and non-live two class；
Step 3-2：Reusing feature selection approach to the data flow characteristics of last layer classification results C4 carries out feature choosing
Select, and carry out second layer svm classifier, obtain classification results C41, C4；Wherein, C41 is online live video, and C42 is online non-
Live video；
Step 3-3：Reusing feature selection approach to the data flow characteristics of last layer classification results C42 carries out feature choosing
Select, and carry out third layer svm classifier, obtain classification results C421, C422, C433；Wherein, C421 is SD video, and C422 is
HD video, C423 is super clear video；
Step 3-4：Statistical classification output result.
Claims (5)
1. the video stream characteristics based on particle swarm optimization algorithm are selected and sorting technique, it is characterised in that comprised the following steps：
Step 1：Required experimental data is obtained using network package analysis software in open internet environment, then logarithm
Filtered according to bag, finally these network video service streams are carried out with basic statistical nature and is calculated；
Step 2：Statistical nature to the above-mentioned video traffic for calculating is analyzed, and selecting can effective differentiated service stream
Combinations of features；
Step 3：Three layers of SVM cascade grader according to design carry out classification experiments to original video traffic, obtain final
Classification results.
2. video stream characteristics selection and sorting technique based on particle swarm optimization algorithm according to claim 1, its feature
It is that 1 specifically includes the step of methods described：
Step 1-1：In open internet environment, the video traffic fluxion needed for being captured by network package analysis software
According to, then original data are simply pre-processed, be converted into the five-tuple text formatting of standard, i.e., packet reach when
Between, source IP address, purpose IP address, agreement, packet packet size；
Step 1-2：Refer to filter the packet lost interest in or will not classification results be produced with influence on Packet Filtering；
Step 1-3：Standard quintuple file to original video stream carries out basic statistical nature calculating, and these features include：
Bag size, the average of bag size and variance, bag size information entropy, the average of inter-packet gap and variance, byte-rate, packet rates,
The ratio between the ratio between up-downgoing byte number, up-downgoing bag size.
3. video stream characteristics selection and sorting technique based on particle swarm optimization algorithm according to claim 1, its feature
It is that 2 specifically include the step of methods described：
Step 2-1：Statistical nature to all video traffics carries out discretization operations, reduces the calculating in feature selection process
Expense；
Step 2-2：The weight of each statistical nature is calculated using feature weight algorithm；
Step 2-3：According to the ranking of feature weight, remove part and the less feature of category associations, maximum N number of of weight selection
Feature, reduces the dimension of original feature space, reduces the computation complexity of subsequent operation；
Step 2-3：In N number of character subset that previous step is chosen, selected characteristic weight M feature in the top is used as priori
Knowledge, instructs the initialization of population of particle swarm optimization algorithm, and the initial position of each particle is set into optimal location；Iterations
It is set to 1；
Step 2-4：Using inconsistent rate as the fitness function of particle swarm optimization algorithm, particle is calculated using fitness function
Overall fitness, a pattern is referred to as by a combinations of features in sample instance, all patterns of character subset it is inconsistent
Number, the total sample number for being equal to pattern appearance subtracts the sample number of the most a certain class label of occurrence number, and inconsistent rate is just
Equal to inconsistent number divided by total sample number；
Step 2-5：If the fitness of current particle is less than the fitness of particle itself optimal location, by the optimal position of particle itself
Put and be updated to current location；If the overall fitness of particle itself optimal location is less than the fitness of the optimal location of population,
The optimal location of population is updated to particle itself optimal location；
Step 2-6：Position and velocity information according to current particle update position and the speed of population；
Step 2-7：If it is lasting constant in an iterative process to meet maximum iteration or inconsistent rate, optimal solution is exported；It is no
Then, repeat step 2-5 to step 2-6.
4. video stream characteristics selection and sorting technique based on particle swarm optimization algorithm according to claim 3, its feature
It is the N in step 2-3 for 10, M is 2.
5. video stream characteristics selection and sorting technique based on particle swarm optimization algorithm according to claim 1, its feature
It is that 3 specifically include the step of methods described：
Step 3-1：Original video Business Stream feature is selected using feature selection approach, and carries out ground floor svm classifier,
Obtain classification results C1, C2, C3, C4；Wherein, C1 is instant messaging class video, and C2 is P2P class videos, and C3 is downloaded for http, C4
It is Online Video, comprising live and non-live two class；
Step 3-2：Reusing feature selection approach to the data flow characteristics of last layer classification results C4 carries out feature selecting, and
Second layer svm classifier is carried out, classification results C41, C4 is obtained；Wherein, C41 is online live video, and C42 non-live is regarded for online
Frequently；
Step 3-3：Reusing feature selection approach to the data flow characteristics of last layer classification results C42 carries out feature selecting,
And third layer svm classifier is carried out, obtain classification results C421, C422, C433；Wherein, C421 is SD video, and C422 is high definition
Video, C423 is super clear video；
Step 3-4：Statistical classification output result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710032385.3A CN106897733A (en) | 2017-01-16 | 2017-01-16 | Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710032385.3A CN106897733A (en) | 2017-01-16 | 2017-01-16 | Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106897733A true CN106897733A (en) | 2017-06-27 |
Family
ID=59198478
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710032385.3A Pending CN106897733A (en) | 2017-01-16 | 2017-01-16 | Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106897733A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107844801A (en) * | 2017-10-19 | 2018-03-27 | 苏翀 | A kind of sorting technique of spam |
| CN108494620A (en) * | 2018-02-28 | 2018-09-04 | 南京邮电大学 | Network service flow feature selecting and sorting technique based on multiple target Adaptive evolvement arithmetic |
| CN108874850A (en) * | 2018-02-14 | 2018-11-23 | 南京邮电大学 | Network video service feature selection approach based on PSOGSA-CI |
| CN110336637A (en) * | 2019-07-15 | 2019-10-15 | 北京航空航天大学 | A kind of unmanned plane interference signal feature selection approach |
| CN110991518A (en) * | 2019-11-28 | 2020-04-10 | 山东大学 | Two-stage feature selection method and system based on evolution multitask |
| CN112953851A (en) * | 2019-12-10 | 2021-06-11 | 华为数字技术（苏州）有限公司 | Traffic classification method and traffic management equipment |
| CN114513685A (en) * | 2022-01-28 | 2022-05-17 | 武汉绿色网络信息服务有限责任公司 | Method and device for identifying HTTPS encrypted video stream based on stream characteristics |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105809190A (en) * | 2016-03-03 | 2016-07-27 | 南京邮电大学 | Characteristic selection based SVM cascade classifier method |
-
2017
- 2017-01-16 CN CN201710032385.3A patent/CN106897733A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105809190A (en) * | 2016-03-03 | 2016-07-27 | 南京邮电大学 | Characteristic selection based SVM cascade classifier method |
Non-Patent Citations (3)
| Title |
|---|
| FATEME HAJIKARAMI等: "A Modular Two-layer System for Accurate and Fast Traffic Classification", 《2014 11TH INTERNATIONAL ISC CONFERENCE ON INFORMATION SECURITY AND CRYPTOLOGY》 * |
| 曾向阳: "《智能水中目标识别》", 31 March 2016, 国防工业出版社 * |
| 肖艳等: "基于Relief F和PSO混合特征选择的面向对象土地利用分类", 《农业工程学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107844801A (en) * | 2017-10-19 | 2018-03-27 | 苏翀 | A kind of sorting technique of spam |
| CN107844801B (en) * | 2017-10-19 | 2021-04-06 | 苏翀 | Classification method of junk mails |
| CN108874850A (en) * | 2018-02-14 | 2018-11-23 | 南京邮电大学 | Network video service feature selection approach based on PSOGSA-CI |
| CN108874850B (en) * | 2018-02-14 | 2022-02-22 | 南京邮电大学 | Network video service feature selection method based on PSOGSA-CI |
| CN108494620B (en) * | 2018-02-28 | 2021-07-27 | 南京邮电大学 | Network service flow characteristic selection and classification method |
| CN108494620A (en) * | 2018-02-28 | 2018-09-04 | 南京邮电大学 | Network service flow feature selecting and sorting technique based on multiple target Adaptive evolvement arithmetic |
| CN110336637A (en) * | 2019-07-15 | 2019-10-15 | 北京航空航天大学 | A kind of unmanned plane interference signal feature selection approach |
| CN110991518A (en) * | 2019-11-28 | 2020-04-10 | 山东大学 | Two-stage feature selection method and system based on evolution multitask |
| CN110991518B (en) * | 2019-11-28 | 2023-11-21 | 山东大学 | Two-stage feature selection method and system based on evolutionary multitasking |
| WO2021114844A1 (en) * | 2019-12-10 | 2021-06-17 | 华为技术有限公司 | Traffic classification method and traffic management device |
| CN112953851A (en) * | 2019-12-10 | 2021-06-11 | 华为数字技术（苏州）有限公司 | Traffic classification method and traffic management equipment |
| CN114513685A (en) * | 2022-01-28 | 2022-05-17 | 武汉绿色网络信息服务有限责任公司 | Method and device for identifying HTTPS encrypted video stream based on stream characteristics |
| CN114513685B (en) * | 2022-01-28 | 2022-10-11 | 武汉绿色网络信息服务有限责任公司 | Method and device for identifying HTTPS encrypted video stream based on stream characteristics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106897733A (en) | Video stream characteristics selection and sorting technique based on particle swarm optimization algorithm | |
| CN110290022B (en) | Unknown application layer protocol identification method based on adaptive clustering | |
| CN104052639B (en) | Real-time multi-application network flow identification method based on support vector machine | |
| EP3275124B1 (en) | Network traffic classification | |
| CN105809190B (en) | A kind of SVM cascade classifier methods based on Feature Selection | |
| CN102035698B (en) | HTTP tunnel detection method based on decision tree classification algorithm | |
| CN105871619B (en) | A kind of flow load type detection method based on n-gram multiple features | |
| CN107967311A (en) | A kind of method and apparatus classified to network data flow | |
| CN110417729B (en) | Service and application classification method and system for encrypted traffic | |
| CN108650194A (en) | Net flow assorted method based on K_means and KNN blending algorithms | |
| CN102394827A (en) | Hierarchical classification method for internet flow | |
| CN108696543B (en) | Distributed reflection denial of service attack detection and defense method based on deep forest | |
| CN112949739A (en) | Information transmission scheduling method and system based on intelligent traffic classification | |
| CN101902484A (en) | Method and system for classifying local area network http application services | |
| CN112800424A (en) | Botnet malicious traffic monitoring method based on random forest | |
| CN109286576A (en) | A kind of network agent encryption traffic characteristic extracting method of data packet frequency analysis | |
| CN111611280A (en) | Encrypted traffic identification method based on CNN and SAE | |
| CN111565156A (en) | Method for identifying and classifying network traffic | |
| CN112910853A (en) | Encryption flow classification method based on mixed characteristics | |
| Kohout et al. | Automatic discovery of web servers hosting similar applications | |
| CN112884121A (en) | Traffic identification method based on generation of confrontation deep convolutional network | |
| CN114553722B (en) | VPN and non-VPN network flow classification method based on multi-view one-dimensional convolutional neural network | |
| Dixit et al. | Internet traffic detection using naïve bayes and K-Nearest neighbors (KNN) algorithm | |
| CN108494620B (en) | Network service flow characteristic selection and classification method | |
| CN107832611B (en) | Zombie program detection and classification method combining dynamic and static characteristics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170627 |
|
| RJ01 | Rejection of invention patent application after publication |