CN106612511A - Wireless network throughput evaluation method and device based on support vector machine - Google Patents
Wireless network throughput evaluation method and device based on support vector machine Download PDFInfo
- Publication number
- CN106612511A CN106612511A CN201510686017.1A CN201510686017A CN106612511A CN 106612511 A CN106612511 A CN 106612511A CN 201510686017 A CN201510686017 A CN 201510686017A CN 106612511 A CN106612511 A CN 106612511A
- Authority
- CN
- China
- Prior art keywords
- base station
- throughput
- handling capacity
- base
- sequence
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/08—Testing, supervising or monitoring using real traffic
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a wireless network throughput evaluation method and device based on a support vector machine, which relates to the technical field of wireless communication and data mining. The method comprises the steps that the historical data of the throughput of N base stations are collected; according to the collected historical data of the throughput of N base stations, the base station relational network of N base stations is constructed; according to the base station relational network and the historical data of the throughput, M base stations which play an important role in evaluating the network throughput of the base station relationship are found, and are used as important base stations; and the throughput historical data of M important base stations are used to evaluate the throughput of the remaining N to M base stations. According to the invention, the throughput of selected small number of base stations can reflect the characteristics of other large number of base stations, which reduces the complexity of data analysis.
Description
Technical field
The present invention relates to radio communication and data mining technology field, including social network analysis, support vector machine
Etc. method.The appraisal procedure and device of more particularly to a kind of throughput of wireless networks based on support vector machine.
Background technology
With the fast development of wireless network, the species and flow of mobile Internet data business have very big
Improve, flow explosive increase, type of service are extremely enriched, to network traffic behavior analysis also with regard to more sophisticated.
In order to effectively realize Network Programe Design, Internet resources distribution, lean operation management etc., it is necessary to divide exactly
Analysis network throughput.Due to the multiformity of data service, randomness and it is sudden the features such as, traditional data point
Analysis method can not meet current network throughput behavior analysiss.In recent years, with the development of big data,
Related data mining algorithm is also more and more ripe, and these algorithms are also provided for throughput of wireless networks behavior analysiss
Strong instrument.The corresponding data mining algorithm of reasonable employment analyzes network throughput behavior, can be to net
Network planning, optimization, dilatation etc. provide certain guidance, so as to improve network holding to the data service that becomes increasingly abundant
Loading capability.
The content of the invention
It is an object of the invention to provide a kind of appraisal procedure of the throughput of wireless networks based on support vector machine
And device, solve in prior art due to the multiformity of data service, randomness and sudden, it is impossible to enough full
The problem of the current network throughput behavior analysiss of foot.
According to an aspect of the invention, there is provided a kind of throughput of wireless networks based on support vector machine is commented
Estimate method, comprise the following steps：
Gather the historical data of the handling capacity of N number of base station；
According to the historical data of the handling capacity of the N number of base station for being gathered, the base station relationship of N number of base station is built
Network；
Using the historical data according to the base station relationship network and the handling capacity, find to base station relationship net
The M base station that network throughput evaluation effect plays an important role, and using the M base station as significant base stations；
Using the throughput history data of the M significant base stations, remaining N-M base station throughput is carried out
Assessment；
Wherein, N and M are positive integer, and N is more than M.
Preferably, the historical data of the handling capacity of the N number of base station of the collection includes：
The handling capacity sequence of each base station is obtained, and calculates the meansigma methodss of handling capacity in the handling capacity sequence；
Described handling up is replaced with by front 3% big handling capacity in the handling capacity sequence by each acquired base station
The meansigma methodss of amount, obtain the new handling capacity sequence of each base station；
It is normalized by the time serieses of the new handling capacity to each base station, obtains returning for each base station
One changes handling capacity sequence.
Preferably, the historical data of the handling capacity according to the N number of base station for being gathered, builds N number of base
The base station relationship network stood includes：
According to the Normalized throughput sequence of resulting each base station, base two-by-two is calculated in N number of base station respectively
Correlation coefficient between standing；
When the correlation coefficient be more than correlation coefficient threshold when, then it is described two-by-two between base station generate one it is undirected
Side；
By the nonoriented edge for being generated between base station in N number of base station two-by-two, the base station of N number of base station is built
Relational network.
Preferably, the historical data using according to the base station relationship network and the handling capacity, finds
The M base station that base station relationship network throughput Evaluated effect plays an important role is included：
By the undirected edge strip number for counting each base station in the base station relationship network, the degree of each base station is obtained；
The big base station of front m degree in N number of base station is chosen successively, assesses N-m base according to support vector machine
The handling capacity stood, obtains the throughput evaluation effect of m kind base station relationship networks；
In the throughput evaluation effect of resulting m kind base station relationship networks, best throughput evaluation is chosen
Effect, and using m selected best throughput evaluation effect corresponding base station as to base station relationship network
The M base station that throughput evaluation effect plays an important role；
Wherein, m, M, N be positive integer, M<=m, M<N, m<N.
Preferably, the undirected edge strip number of each base station is directly proportional to the size of base station degree.
Preferably, the throughput history data using the M significant base stations, to remaining N-M base
Handling capacity of standing be estimated including：
The throughput concerns mould of remaining N-M base station and M significant base stations is constructed by algorithm of support vector machine
Type；
Using the throughput concerns model and the throughput history data of the M significant base stations, residue is obtained
N-M base station assessment handling capacity.
According to a further aspect in the invention, there is provided a kind of throughput of wireless networks based on support vector machine is commented
Estimate device, including：
Acquisition module, for gathering the historical data of the handling capacity of N number of base station；
Module is built, for the historical data of the handling capacity according to the N number of base station for being gathered, the N is built
The base station relationship network of individual base station；
Searching modul, for utilizing according to the base station relationship network and the historical data of the handling capacity, looks for
To the M base station played an important role to base station relationship network throughput Evaluated effect, and using the M base station as
Significant base stations；
Evaluation module, for the throughput history data using the M significant base stations, to remaining N-M
Individual base station throughput is estimated；
Wherein, N and M are positive integer, and N is more than M.
Preferably, the acquisition module includes：
Handling capacity averaging unit is calculated, for obtaining the handling capacity sequence of each base station, and described gulping down is calculated
The meansigma methodss of handling capacity in the amount of telling sequence；
Acquiring unit, for by front 3% big handling capacity in the handling capacity sequence by each acquired base station
The meansigma methodss of the handling capacity are replaced with, the new handling capacity sequence of each base station is obtained, and by each base
The time serieses of the new handling capacity stood are normalized, and obtain the Normalized throughput sequence of each base station.
Preferably, the structure module includes：
Correlation coefficient unit is calculated, for the Normalized throughput sequence according to resulting each base station, is counted respectively
Calculate correlation coefficient two-by-two between base station in N number of base station；
Generate nonoriented edge unit, for when the correlation coefficient be more than correlation coefficient threshold when, then it is described two-by-two
A nonoriented edge is generated between base station；
Construction unit, for the nonoriented edge by being generated between base station in N number of base station two-by-two, builds described
The base station relationship network of N number of base station.
Preferably, the searching modul includes：
Acquiring unit, for the undirected edge strip number by counting each base station in the base station relationship network, obtains
The degree of each base station, and the big base station of front m degree in N number of base station is chosen successively, according to supporting vector
Machine assesses the handling capacity of N-m base station, obtains the throughput evaluation effect of m kind base station relationship networks；
Searching unit, for, in the throughput evaluation effect of resulting m kind base station relationship networks, choosing most
Good throughput evaluation effect, and using m selected best throughput evaluation effect corresponding base station as
The M base station played an important role by base station relationship network throughput Evaluated effect；
Wherein, m, M, N be positive integer, M<=m, M<N, m<N.
Compared with prior art, the beneficial effects of the present invention is：
The present invention enables a small amount of base station throughput for selecting to embody the characteristic of other a large amount of base stations, is several
Reduce complexity according to analysis；Simultaneously, enabling in the case of known spatial portion of base stations handling capacity, assess
Go out the handling capacity of other a large amount of unknown base stations in spatial dimension, so as to the optimization to wireless network resource provides ginseng
Examine.
Description of the drawings
Fig. 1 is a kind of assessment of throughput of wireless networks based on support vector machine provided in an embodiment of the present invention
Method flow diagram；
Fig. 2 is a kind of assessment of throughput of wireless networks based on support vector machine provided in an embodiment of the present invention
Schematic device；
Fig. 3 is the throughput of wireless networks appraisal procedure based on support vector machine provided in an embodiment of the present invention
Flow chart；
Fig. 4 is the algorithm flow chart of support vector machine provided in an embodiment of the present invention；
Fig. 5 is the structure base station relationship network that first embodiment of the invention is provided；
Fig. 6 is SMAPE (the Symmetric mean absolute that first embodiment of the invention is provided
Percentage error, Symmetric mean relative error) meansigma methodss with m situation of change figure；
Fig. 7 is the assessment result figure of two base stations that first embodiment of the invention is provided；
Fig. 8 is the structure base station relationship network that second embodiment of the invention is provided；
Fig. 9 is the situation of change figure of the SMAPE meansigma methodss with m of second embodiment of the invention offer；
Figure 10 is the assessment result figure of two base stations that second embodiment of the invention is provided.
Specific embodiment
Below in conjunction with accompanying drawing to a preferred embodiment of the present invention will be described in detail, it will be appreciated that described below
Preferred embodiment be merely to illustrate and explain the present invention, be not intended to limit the present invention.
Fig. 1 is a kind of assessment of throughput of wireless networks based on support vector machine provided in an embodiment of the present invention
Method flow diagram, as shown in figure 1, comprising the following steps：
Step S101：Gather the historical data of the handling capacity of N number of base station；
Step S102：According to the historical data of the handling capacity of the N number of base station for being gathered, N number of base is built
The base station relationship network stood；
Step S103：Using the historical data according to the base station relationship network and the handling capacity, find
The M base station played an important role by base station relationship network throughput Evaluated effect, and using the M base station as weight
Want base station；
Step S104：Using the throughput history data of the M significant base stations, to remaining N-M base
Handling capacity of standing is estimated；
Wherein, N and M are positive integer, and N is more than M.
Wherein, the historical data of the handling capacity of the N number of base station of the collection includes：Obtain the handling capacity of each base station
Sequence, and calculate the meansigma methodss of handling capacity in the handling capacity sequence；By by each acquired base station
In handling capacity sequence, front 3% big handling capacity replaces with the meansigma methodss of the handling capacity, obtains the new of each base station
Handling capacity sequence；It is normalized by the time serieses of the new handling capacity to each base station, obtains each
The Normalized throughput sequence of base station.
Wherein, the historical data of the handling capacity according to the N number of base station for being gathered, builds N number of base station
Base station relationship network include：According to the Normalized throughput sequence of resulting each base station, calculate respectively described
Correlation coefficient in N number of base station two-by-two between base station；When the correlation coefficient is more than correlation coefficient threshold, then
A nonoriented edge is generated two-by-two between base station described；By what is generated between base station in N number of base station two-by-two
Nonoriented edge, builds the base station relationship network of N number of base station.
Wherein, the historical data using according to the base station relationship network and the handling capacity, it is right to find
The M base station that base station relationship network throughput Evaluated effect plays an important role includes：Closed by counting the base station
It is the undirected edge strip number of each base station in network, obtains the degree of each base station；Chosen in N number of base station successively
First m is spent big base station, assesses the handling capacity of N-m base station according to support vector machine, obtains m kinds base station pass
It is the throughput evaluation effect of network；In the throughput evaluation effect of resulting m kind base station relationship networks,
Choose best throughput evaluation effect, and by m selected best throughput evaluation effect corresponding base
Stand as the M base station played an important role to base station relationship network throughput Evaluated effect；Wherein, m, M, N
For positive integer, M<=m, M<N, m<N.Wherein, the undirected edge strip number of each base station is big with base station degree
It is little to be directly proportional.
Wherein, the throughput history data using the M significant base stations, to remaining N-M base station
Handling capacity be estimated including：Remaining N-M base station is constructed with M important base by algorithm of support vector machine
The throughput concerns model stood；Gone through using the handling capacity of the throughput concerns model and the M significant base stations
History data, obtain the assessment handling capacity of remaining N-M base station.
Fig. 2 is a kind of assessment of throughput of wireless networks based on support vector machine provided in an embodiment of the present invention
Schematic device, as shown in Fig. 2 including：Acquisition module 201, structure module 202, searching modul 203
And evaluation module 204.The acquisition module 201, for gathering the historical data of the handling capacity of N number of base station；
The structure module 202, for the historical data of the handling capacity according to the N number of base station for being gathered, builds described
The base station relationship network of N number of base station；The searching modul 203, for utilizing according to the base station relationship network
And the historical data of the handling capacity, find what base station relationship network throughput Evaluated effect was played an important role
M base station, and using the M base station as significant base stations；The evaluation module 204, for using the M
The throughput history data of individual significant base stations, are estimated to remaining N-M base station throughput；Wherein, N
Positive integer is with M, and N is more than M.
Wherein, the acquisition module 201 includes：Handling capacity averaging unit is calculated, for obtaining each base station
Handling capacity sequence, and calculate the meansigma methodss of handling capacity in the handling capacity sequence；Acquiring unit, for leading to
Cross front 3% big handling capacity in the handling capacity sequence by each acquired base station and replace with the flat of the handling capacity
Average, obtains the new handling capacity sequence of each base station, and by the time sequence of the new handling capacity to each base station
Row are normalized, and obtain the Normalized throughput sequence of each base station.The structure module 202 includes：
Correlation coefficient unit is calculated, for the Normalized throughput sequence according to resulting each base station, institute is calculated respectively
State correlation coefficient two-by-two between base station in N number of base station；Nonoriented edge unit is generated, for when the correlation coefficient
During more than correlation coefficient threshold, then a nonoriented edge is generated two-by-two between base station described；Construction unit, is used for
By the nonoriented edge for being generated between base station in N number of base station two-by-two, the base station relationship of N number of base station is built
Network.The searching modul 203 includes：Acquiring unit, for by counting every in the base station relationship network
The undirected edge strip number of individual base station, obtains the degree of each base station, and chooses front m in N number of base station successively
The big base station of individual degree, assesses the handling capacity of N-m base station according to support vector machine, obtains m kind base station relationship nets
The throughput evaluation effect of network；Searching unit, for commenting in the handling capacity of resulting m kind base station relationship networks
Estimate in effect, choose best throughput evaluation effect, and will be selected best throughput evaluation effect relative
The m base station answered is used as the M base station played an important role to base station relationship network throughput Evaluated effect；Wherein,
M, M, N be positive integer, M<=m, M<N, m<N.
It is of the invention mainly to include following four module：Data preprocessing module, base station relationship network struction module,
Significant base stations choose module, space throughput evaluation module.The data preprocessing module, waits to grind for choosing
The N number of base station studied carefully, rejects exceptional data point therein；The base station relationship network struction module, for basis
The N number of base station historical data for having gathered builds the relational network between base station；The significant base stations choose module,
Evaluated effect for basis in historical data, selects M significant base stations from N number of base station；The space
Throughput evaluation module, for for the time to be assessed, going out which according to the throughput evaluation of known M base station
The handling capacity of his N-M base station.
Specifically, the data preprocessing module includes：
A1. choose the N number of base station on locus in the same area；
A2. reject the abnormity point in each base station throughput；
A3. data are carried out with a normalization.
The base station relationship network struction module includes：
B1. calculate N number of base station correlation coefficient between any two；
B2. according to correlation coefficient, build the base station relationship network of a given threshold value.
The significant base stations choose module to be included：
C1. in accounting base-station relational network each base station degree size；
C2. base station m (m=1,2 ... N) big before selection degree successively is used as significant base stations, according to support vector machine
Assess the handling capacity of other N-m base station；
M base station when being C3. chosen at Evaluated effect in historical data preferably is used as significant base stations.
The space throughput evaluation module includes：
D1. according to the M significant base stations selected, assess the handling capacity of other N-M base station.
Fig. 3 is the throughput of wireless networks appraisal procedure based on support vector machine provided in an embodiment of the present invention
Flow chart, as shown in figure 3, including：
Step 1, data prediction；
In order to future is according to the handling capacity of other a large amount of base stations of throughput evaluation of portion of base stations, need first to collect
Then data are carried out pretreatment by the history service data of all base stations.Data prediction is mainly comprising following several
Individual step：
N number of base station on a, according to demand selection locus in the same area；
Front 3% big handling capacity in each sequence is replaced with the sequence by b, the N number of base station throughput sequence of arrangement
Meansigma methodss, obtain handling capacity sequence p of i-th base stationi(i=1,2 ... N)；
C, the time serieses to each base station throughput are normalized, and the normalization for obtaining i-th base station is gulped down
The amount of telling sequence Si。
Wherein,For the Normalized throughput of i-th base station t, max (pi), min (pi) difference table
Show the maxima and minima of raw throughput sequence, L is sequence total length.
Step 2, structure base station relationship network；
To N number of base station to be studied, L is the total duration of institute's gathered data, and in taking L, front T is (generalIt is left
It is right) individual time data calculates the i-th (i=1,2,3 ... N) individual base station and jth (j=1,2,3 ... N) individual base
Correlation coefficient ρ between standingij, computing formula is
SiFor the handling capacity sequence of i-th base station,The average throughput for being i-th base station in total duration,The handling capacity size (t=1,2,3 ... T) for being i-th base station in moment t；SjFor j-th base station
Handling capacity sequence,The average throughput for being j-th base station in total duration,Exist for j-th base station
Handling capacity size (t=1,2,3 ... T) during moment t.Correlation coefficient threshold c given for one, if ρij
More than c, then it is assumed that base station i and base station j has obvious dependency relation, addition one is undirected between them
Side, can thus construct the relational network of N number of base station.
Step 3, selection significant base stations；
In the present invention, using SVM (Support Vector Machine, support vector machine) assessing base
Stand handling capacity.The algorithm flow of SVM is as shown in figure 4, comprise the following steps：
1st, according to assessment Sample Establishing training sample set and test sample collection；
2nd, object function is set up according to training sample set；
3rd, object function is solved, obtains optimized parameter；
4th, optimized parameter is substituted into into object function, obtains decision-making regression equation；
5th, decision-making regression equation is verified using test data；
6th, whether it is less than assigned error e；
When judging less than assigned error e, into step 7, when judging not less than assigned error e, adjustment
Parameter, and return to step 3.
7th, by other base station throughputs of assessment sample input decision-making regression equation calculation.
In the present invention, final throughput evaluation effect quality is weighed using SMAPE, and SMAPE is reflected and commented
The size of relative error between valuation and actual value, while solving may bring as actual value is too small relative
The too big problem of error, its concrete formula is：
Wherein Ft is assessed value, and At is actual value.
In order to filter out portion of base stations as significant base stations, base m (m=1,2 ... N) big before selection degree successively
Stand as significant base stations, the handling capacity of other N-m base station is assessed according to support vector machine.Calculating is assessed out
Each base station SMAPE, choose average SMAPE it is minimum when M base station as significant base stations.
Step 4, other base station throughputs are assessed using significant base stations.
In the present invention, according to the M significant base stations selected, using SVM algorithm, the training of usage history data
Go out the throughput concerns model of other N-M base station and M significant base stations.By the M in the time period to be assessed
The handling capacity of individual base station is input in relational model, you can the handling capacity of the corresponding N-M base station of output.
In order that the above objects, features and advantages of the present invention can become apparent from it is understandable, below in conjunction with the accompanying drawings 5
To accompanying drawing, 10 couples of present invention are further described in detail.
Embodiment one
The data of all base station statistics in Data Source certain large size city in existing network, its time granularity in this example
For 60 minutes, total length of time was continuous 21 days.Wireless network space handling capacity in the embodiment of the present invention is commented
The method of estimating is comprised the steps of：
Step one：Data prediction；
A. 95 base stations on locus in the same area are chosen according to demand；
B. the exceptional data point in 95 base stations is rejected, the handling capacity sequence of each base station is obtained；
C. the time serieses of each base station throughput are normalized, the normalization for obtaining i-th base station is gulped down
The amount of telling sequence Si。
Step 2：To 95 base stations to be studied, base station relationship network is built；
A. before taking this 95 base stations, the data of 18 days calculate the i-th (i=1,2,3 ... 95) individual base station and jth
Correlation coefficient ρ between (j=1,2,3 ... 95) individual base stationij, computing formula is
Wherein T=432, SiFor the handling capacity sequence of i-th base station,It is i-th base station in total duration
Average throughput,The handling capacity size (t=1,2,3 ... 432) for being i-th base station in moment t；Sj
For the handling capacity sequence of j-th base station,The average throughput for being j-th base station in total duration,For
Handling capacity size (t=1,2,3 ... 432) of j-th base station in moment t.
B. in the present invention, correlation coefficient threshold c=0.6 is given (it is generally acknowledged that correlation coefficient more than 0.6 is
For strong correlation), if ρijMore than 0.6, then add a nonoriented edge between base station i and base station j, thus may be used
To construct the relational network of 95 base stations.As shown in figure 5, its midpoint represents base station, nonoriented edge is embodied
Dependency between base station, puts the bigger degree for representing the base station bigger.
Step 4：Choose significant base stations；
A., from the historical data of 18 days, front 15 day data is chosen as training sample set, afterwards 3 day data
As test sample collection；Using 15 day datas before all base stations as algorithm of support vector machine (SVM) input,
Other 95-m base station that output training is obtained and the m base station throughput relational model chosen；
B. rear 3 day data of m maximum base station of degree is exported into which as the input of throughput concerns model
The estimated value of 3 days behind his 95-m base station；
C. the respective SMAPE in 95-m base station of calculating, makes situation of change of the SMAPE meansigma methodss with m, such as
Shown in Fig. 6, meansigma methodss of the stain for 95-m base station SMAPE, it can be seen that as m=8,
The average SMAPE of other base stations is minimum, that is, prediction effect is optimal, therefore we choose in the present embodiment
Significant base stations number be M=8.
Step 5：Space handling capacity is assessed using algorithm of support vector machine.
In the present invention, we are according to 8 significant base stations selected, using algorithm of support vector machine (SVM)
Usage history data train the throughput concerns model of other 87 base stations and 8 significant base stations.Will be original
The handling capacity of 8 base stations of last 3 days in 21 day datas is input in relational model, you can output correspondence
87 base stations handling capacity.
As shown in fig. 7, illustrate the assessment result of portion of base stations in 87 base stations, wherein 1 is assessed value, 2
For actual value.The assessment errors of 87 base stations are calculated, average SMAPE=30.3% is obtained, it is seen that the method has
There is higher accuracy
Embodiment two
The statistical data of Data Source representative region in certain large size city in existing network, its time granule in this example
Spend for 60 minutes, total length of time is continuous 18 days.Wireless network space handling capacity in the embodiment of the present invention
Appraisal procedure is comprised the steps of：
Step one：Data prediction；
A. 117 base stations on locus in the same area are chosen according to demand；
B. the exceptional data point in 117 base stations is rejected, the handling capacity sequence of each base station is obtained；
C. the time serieses of each base station throughput are normalized, obtain the normalization of i-th base station
Handling capacity sequence Si。
Step 2：To 117 base stations to be studied, base station relationship network is built；
A. the data of 15 days before this 117 base stations are taken and calculates the i-th (i=1,2,3 ... 117) individual base station and the
Correlation coefficient ρ between j (j=1,2,3 ... 117) individual base stationij, computing formula is
Wherein T=360, SiFor the handling capacity sequence of i-th base station,It is i-th base station in total duration
Average throughput,The handling capacity size (t=1,2,3 ... 360) for being i-th base station in moment t；Sj
For the handling capacity sequence of j-th base station,The average throughput for being j-th base station in total duration,For
Handling capacity size (t=1,2,3 ... 360) of j-th base station in moment t.
B. in the present invention, we give correlation coefficient threshold c=0.6 (it is generally acknowledged that correlation coefficient is more than 0.6
As strong correlation), if ρijMore than 0.6, then add a nonoriented edge between base station i and base station j, thus
The relational network of 117 base stations can be constructed.As shown in figure 8, its midpoint represents base station, while embodying base
Dependency between standing, puts the bigger degree for representing the base station bigger.
Step 4：Choose significant base stations；
A., from the historical data of 15 days, front 12 day data is chosen as training set, afterwards three day data conduct
Test set；Using 12 day datas before all base stations as the input of algorithm of support vector machine (SVM), training is exported
Other 117-m base station for obtaining and the m base station throughput relational model chosen；
B. rear 3 day data of m maximum base station of degree is exported into which as the input of throughput concerns model
The estimated value of three days behind his 117-m base station；
C. the respective SMAPE in base station of 117-m is calculated, situations of change of the average SMAPE with m is made, such as
Shown in Fig. 9, meansigma methodss of the stain for 117-m base station SMAPE, it can be seen that as m=11,
The average SMAPE of other base stations is minimum, that is, prediction effect is optimal, therefore we choose in the present embodiment
Significant base stations number be M=11.
Step 5：Space handling capacity is assessed using SVM algorithm.
In the present invention, we are according to 11 significant base stations selected, using algorithm of support vector machine (SVM)
Usage history data train the throughput concerns model of other 106 base stations and 11 significant base stations.To treat
The handling capacity of 11 base stations in the assessment time period is input in relational model, you can output is corresponding 106
The handling capacity of base station.
As shown in Figure 10, as assessment result example, wherein 3 is assessed value, 4 is actual value.Calculate 106
The assessment errors of individual base station, obtain average SMAPE=36.4%, and assessment result has higher accuracy.
In sum, the present invention has following technique effect：
The present invention obtains handling capacity variation relation between base station according to base station historical data, and builds base station relationship net
Network, selects minority significant base stations from the network, so as to evaluate the handling capacity of other a large amount of base stations.Have
Very high practical value, such as in base station data collection, the data for having many base stations have disappearance, using this
Invention, it can be estimated that go out missing data, so as to make further analysis of network.Meanwhile, can according to demand,
It is flexible to choose different regions or the historical data of time period to assess, with the universal suitability and preferably
Prediction accuracy.
Although being described in detail to the present invention above, the invention is not restricted to this, the art technology
Personnel can carry out various modifications with principle of the invention.Therefore, all modifications made according to the principle of the invention,
All should be understood to fall into protection scope of the present invention.
Claims (10)
1. a kind of appraisal procedure of the throughput of wireless networks based on support vector machine, it is characterised in that include
Following steps：
Gather the historical data of the handling capacity of N number of base station；
According to the historical data of the handling capacity of the N number of base station for being gathered, the base station relationship of N number of base station is built
Network；
Using the historical data according to the base station relationship network and the handling capacity, find to base station relationship net
The M base station that network throughput evaluation effect plays an important role, and using the M base station as significant base stations；
Using the throughput history data of the M significant base stations, remaining N-M base station throughput is carried out
Assessment；
Wherein, N and M are positive integer, and N is more than M.
2. method according to claim 1, it is characterised in that the handling capacity of the N number of base station of the collection
Historical data include：
The handling capacity sequence of each base station is obtained, and calculates the meansigma methodss of handling capacity in the handling capacity sequence；
Described handling up is replaced with by front 3% big handling capacity in the handling capacity sequence by each acquired base station
The meansigma methodss of amount, obtain the new handling capacity sequence of each base station；
It is normalized by the time serieses of the new handling capacity to each base station, obtains returning for each base station
One changes handling capacity sequence.
3. method according to claim 2, it is characterised in that described according to the N number of base station for being gathered
Handling capacity historical data, the base station relationship network for building N number of base station includes：
According to the Normalized throughput sequence of resulting each base station, base two-by-two is calculated in N number of base station respectively
Correlation coefficient between standing；
When the correlation coefficient be more than correlation coefficient threshold when, then it is described two-by-two between base station generate one it is undirected
Side；
By the nonoriented edge for being generated between base station in N number of base station two-by-two, the base station of N number of base station is built
Relational network.
4. method according to claim 3, it is characterised in that described using according to the base station relationship
The historical data of network and the handling capacity, finds and plays important work to base station relationship network throughput Evaluated effect
M base station includes：
By the undirected edge strip number for counting each base station in the base station relationship network, the degree of each base station is obtained；
The big base station of front m degree in N number of base station is chosen successively, assesses N-m base according to support vector machine
The handling capacity stood, obtains the throughput evaluation effect of m kind base station relationship networks；
In the throughput evaluation effect of resulting m kind base station relationship networks, best throughput evaluation is chosen
Effect, and using m selected best throughput evaluation effect corresponding base station as to base station relationship network
The M base station that throughput evaluation effect plays an important role；
Wherein, m, M, N be positive integer, M<=m, M<N, m<N.
5. method according to claim 4, it is characterised in that the undirected edge strip number of each base station
It is directly proportional to the size of base station degree.
6. method according to claim 5, it is characterised in that described using the M significant base stations
Throughput history data, remaining N-M base station throughput is estimated including：
The throughput concerns mould of remaining N-M base station and M significant base stations is constructed by algorithm of support vector machine
Type；
Using the throughput concerns model and the throughput history data of the M significant base stations, residue is obtained
N-M base station assessment handling capacity.
7. a kind of apparatus for evaluating of the throughput of wireless networks based on support vector machine, it is characterised in that include：
Acquisition module, for gathering the historical data of the handling capacity of N number of base station；
Module is built, for the historical data of the handling capacity according to the N number of base station for being gathered, the N is built
The base station relationship network of individual base station；
Searching modul, for utilizing according to the base station relationship network and the historical data of the handling capacity, looks for
To the M base station played an important role to base station relationship network throughput Evaluated effect, and using the M base station as
Significant base stations；
Evaluation module, for the throughput history data using the M significant base stations, to remaining N-M
Individual base station throughput is estimated；
Wherein, N and M are positive integer, and N is more than M.
8. device according to claim 7, it is characterised in that the acquisition module includes：
Handling capacity averaging unit is calculated, for obtaining the handling capacity sequence of each base station, and described gulping down is calculated
The meansigma methodss of handling capacity in the amount of telling sequence；
Acquiring unit, for by front 3% big handling capacity in the handling capacity sequence by each acquired base station
The meansigma methodss of the handling capacity are replaced with, the new handling capacity sequence of each base station is obtained, and by each base
The time serieses of the new handling capacity stood are normalized, and obtain the Normalized throughput sequence of each base station.
9. device according to claim 8, it is characterised in that the structure module includes：
Correlation coefficient unit is calculated, for the Normalized throughput sequence according to resulting each base station, is counted respectively
Calculate correlation coefficient two-by-two between base station in N number of base station；
Generate nonoriented edge unit, for when the correlation coefficient be more than correlation coefficient threshold when, then it is described two-by-two
A nonoriented edge is generated between base station；
Construction unit, for the nonoriented edge by being generated between base station in N number of base station two-by-two, builds described
The base station relationship network of N number of base station.
10. device according to claim 9, it is characterised in that the searching modul includes：
Acquiring unit, for the undirected edge strip number by counting each base station in the base station relationship network, obtains
The degree of each base station, and the big base station of front m degree in N number of base station is chosen successively, according to supporting vector
Machine assesses the handling capacity of N-m base station, obtains the throughput evaluation effect of m kind base station relationship networks；
Searching unit, for, in the throughput evaluation effect of resulting m kind base station relationship networks, choosing most
Good throughput evaluation effect, and using m selected best throughput evaluation effect corresponding base station as
The M base station played an important role by base station relationship network throughput Evaluated effect；
Wherein, m, M, N be positive integer, M<=m, M<N, m<N.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510686017.1A CN106612511B (en) | 2015-10-21 | 2015-10-21 | Wireless network throughput evaluation method and device based on support vector machine |
| PCT/CN2016/084549 WO2016188498A1 (en) | 2015-10-21 | 2016-06-02 | Wireless network throughput evaluating method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510686017.1A CN106612511B (en) | 2015-10-21 | 2015-10-21 | Wireless network throughput evaluation method and device based on support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106612511A true CN106612511A (en) | 2017-05-03 |
| CN106612511B CN106612511B (en) | 2020-03-27 |
Family
ID=57392550
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510686017.1A Active CN106612511B (en) | 2015-10-21 | 2015-10-21 | Wireless network throughput evaluation method and device based on support vector machine |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN106612511B (en) |
| WO (1) | WO2016188498A1 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111901206A (en) * | 2020-08-28 | 2020-11-06 | 杭州安恒信息技术股份有限公司 | Network card testing method, device and related equipment |
| CN112839342A (en) * | 2020-12-31 | 2021-05-25 | 国网吉林省电力有限公司长春供电公司 | Disaster relief mobile emergency base station site selection method based on support vector machine |
| TWI780822B (en) * | 2021-07-19 | 2022-10-11 | 國立陽明交通大學 | Network throughput evaluation device and method |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112600728B (en) * | 2020-12-07 | 2022-06-14 | 南昌交通学院 | 5G mobile base station flow prediction analysis system based on big data |
| CN113709794B (en) * | 2021-08-23 | 2024-03-19 | Oppo广东移动通信有限公司 | Wireless network communication method and related device |
| CN117494908B (en) * | 2023-12-29 | 2024-03-22 | 宁波港信息通信有限公司 | Port cargo throughput prediction method and system based on big data |
| CN117880856B (en) * | 2024-03-11 | 2024-05-31 | 武汉众诚华鑫科技有限公司 | Telecommunication base station safety early warning method and system based on artificial intelligence |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101541030A (en) * | 2009-05-06 | 2009-09-23 | 华为技术有限公司 | Method for predicting data based on support vector machine and equipment thereof |
| CN101998476A (en) * | 2009-08-31 | 2011-03-30 | 中国移动通信集团设计院有限公司 | Method and device for determining cell throughput |
| CN102685766A (en) * | 2012-05-13 | 2012-09-19 | 西华大学 | Wireless network flow prediction method based on local minimax probability machine |
| US20120278130A1 (en) * | 2011-04-28 | 2012-11-01 | Empire Technology Development Llc | Mobile traffic forecasting using public transportation information |
| CN104394538A (en) * | 2014-11-28 | 2015-03-04 | 重庆大学 | Mobile network data flow analysis and prediction method |
-
2015
- 2015-10-21 CN CN201510686017.1A patent/CN106612511B/en active Active
-
2016
- 2016-06-02 WO PCT/CN2016/084549 patent/WO2016188498A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101541030A (en) * | 2009-05-06 | 2009-09-23 | 华为技术有限公司 | Method for predicting data based on support vector machine and equipment thereof |
| CN101998476A (en) * | 2009-08-31 | 2011-03-30 | 中国移动通信集团设计院有限公司 | Method and device for determining cell throughput |
| US20120278130A1 (en) * | 2011-04-28 | 2012-11-01 | Empire Technology Development Llc | Mobile traffic forecasting using public transportation information |
| CN102685766A (en) * | 2012-05-13 | 2012-09-19 | 西华大学 | Wireless network flow prediction method based on local minimax probability machine |
| CN104394538A (en) * | 2014-11-28 | 2015-03-04 | 重庆大学 | Mobile network data flow analysis and prediction method |
Non-Patent Citations (1)
| Title |
|---|
| CLAUDINA RATTARO等: "Throughput prediction in wireless networks using statistical learning", 《LAWDN - LATIN-AMERICAN WORKSHOP ON DYNAMIC NETWORKS》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111901206A (en) * | 2020-08-28 | 2020-11-06 | 杭州安恒信息技术股份有限公司 | Network card testing method, device and related equipment |
| CN112839342A (en) * | 2020-12-31 | 2021-05-25 | 国网吉林省电力有限公司长春供电公司 | Disaster relief mobile emergency base station site selection method based on support vector machine |
| CN112839342B (en) * | 2020-12-31 | 2022-02-08 | 国网吉林省电力有限公司长春供电公司 | Disaster relief mobile emergency base station site selection method based on support vector machine |
| TWI780822B (en) * | 2021-07-19 | 2022-10-11 | 國立陽明交通大學 | Network throughput evaluation device and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106612511B (en) | 2020-03-27 |
| WO2016188498A1 (en) | 2016-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106612511A (en) | Wireless network throughput evaluation method and device based on support vector machine | |
| CN108520357B (en) | Method and device for judging line loss abnormality reason and server | |
| CN105207821B (en) | A kind of network synthesis performance estimating method of service-oriented | |
| CN106548357B (en) | Client satisfaction evaluation method and system | |
| CN104318482A (en) | Comprehensive assessment system and method of smart distribution network | |
| CN106228389A (en) | Network potential usage mining method and system based on random forests algorithm | |
| CN105893637A (en) | Link prediction method in large-scale microblog heterogeneous information network | |
| CN102722577B (en) | The determination method and device of index changeable weight | |
| CN108304853A (en) | Acquisition methods, device, storage medium and the electronic device for the degree of correlation of playing | |
| CN107729519A (en) | Appraisal procedure and device, terminal based on multi-source multidimensional data | |
| CN106991577A (en) | A kind of method and device for determining targeted customer | |
| CN109711424A (en) | A kind of rule of conduct acquisition methods, device and equipment based on decision tree | |
| CN109345109A (en) | A kind of Stakeholder Evaluation method and terminal device based on classification prediction model | |
| CN104331613B (en) | The evaluation method of the communication equipment antijamming capability of multiple types | |
| CN106447075A (en) | Industry electricity utilization demand prediction method and industry electricity utilization demand prediction system | |
| CN108243435A (en) | Parameter optimization method and device in a kind of LTE cell scenarios division | |
| CN109428760B (en) | User credit evaluation method based on operator data | |
| CN115545514A (en) | Health degree evaluation-based differentiated operation and maintenance method and device for power distribution fusion equipment | |
| CN111652403A (en) | Feedback correction-based work platform task workload prediction method | |
| CN104821854B (en) | A kind of many primary user's multidimensional frequency spectrum sensing methods based on random set | |
| CN108985595A (en) | The move transaction service evaluation method and device mutually commented based on counterparty | |
| CN111105127B (en) | Modular product design evaluation method based on data driving | |
| CN104008493A (en) | Data acquisition method and device | |
| CN106980872A (en) | K arest neighbors sorting techniques based on polling committee | |
| CN109948908A (en) | A kind of electric power enterprise develops efficiency rating method, apparatus, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20171016 Address after: 201203 Shanghai Free Trade Zone pilot blue wave Road No. 889 B block 205 Applicant after: Shanghai Zhongxing Software Co., Ltd. Address before: 518057 Nanshan District science and Technology Industrial Park, Guangdong high tech Industrial Park, ZTE building Applicant before: ZTE Corporation |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |