CN107517481A - A kind of load of base station balanced management method and system - Google Patents
A kind of load of base station balanced management method and system Download PDFInfo
- Publication number
- CN107517481A CN107517481A CN201710860446.5A CN201710860446A CN107517481A CN 107517481 A CN107517481 A CN 107517481A CN 201710860446 A CN201710860446 A CN 201710860446A CN 107517481 A CN107517481 A CN 107517481A
- Authority
- CN
- China
- Prior art keywords
- base station
- load
- training sample
- prediction
- sample data
- 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
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W48/00—Access restriction; Network selection; Access point selection
- H04W48/02—Access restriction performed under specific conditions
- H04W48/06—Access restriction performed under specific conditions based on traffic conditions
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W48/00—Access restriction; Network selection; Access point selection
- H04W48/20—Selecting an access point
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Mobile Radio Communication Systems (AREA)
- Telephonic Communication Services (AREA)
Abstract
The invention provides the management method and system that a kind of load of base station is balanced, including：The training sample data of each base station are obtained respectively；It is respectively trained to obtain the load estimation model of each base station；The current load data of each base station is obtained respectively；The prediction load of each base station is calculated respectively；When needing accessing user, loaded according to the present load of each base station and prediction, a base station, the target BS to be accessed as access customer waiting are selected by the second preparatory condition.Historic load of the invention by using each base station, obtain the prediction load of each base station, when needing accessing user, the present load for covering all base stations of the user and prediction loading condition are considered, choose the suitable base station target BS to be accessed as access customer waiting, so as to realize the load balancing of the whole network, overall network performance is lifted.
Description
Technical field
The present invention relates to mobile communication technology field, espespecially a kind of load of base station balanced management method and system.
Background technology
With the fast development of mobile communication and Internet technology, it is a variety of to there is 2G, 3G, 4G, WiFi, WiMax etc.
Network coexisted phenomenon, that is, heterogeneous network, it is characterized in sending from any network in any place at any time
With obtain information, and freely can be switched in heterogeneous networks.
For the heterogeneous network formed in conventional cellular network because disposing small base station, the deployment of a large amount of small base stations,
On the one hand the network coverage and power system capacity are improved, improves the data receiver transmission rate of user so that user obtains more
Good network service, on the other hand, because the movable random of user changes, make the load of each base station different, some access use
Family is few and causes the wasting of resources, and some accessing users are excessive and cause congestion, influence Consumer's Experience.How Internet resources are lifted
Utilization rate, dynamically realizing the load balancing of small base station heterogeneous network turns into a hot issue.
Existing network load balancing method, the current load state in base station is only only accounted for, considering user and base station
Relevance when simply consider SNR (signal to noise ratio) index of user, the i.e. current signal quality of user, do not consider
The loading condition of each base station, also without the historic load for using for reference each base station, a kind of only technology of local optimum,
Certain performance loss can be integrally caused to network.
The content of the invention
It is an object of the invention to provide the management method and system that a kind of load of base station is balanced, by using each base station
Historic load, the prediction load of each base station is obtained, when needing accessing user, has considered the institute for covering the user
There are present load and the prediction loading condition of base station, a suitable base station is chosen as target BS to be accessed, so as to real
The load balancing of existing the whole network, lifts overall network performance.
Technical scheme provided by the invention is as follows：
A kind of balanced management method of load of base station, including：Step S100 obtains the number of training of each base station respectively
According to；The training sample data of each base station be chosen from historic load storehouse corresponding to the base station meet the first default bar
The historic load of part；Step S200 is respectively trained to obtain the load estimation model of each base station；The load of each base station is pre-
It is the training sample data according to the base station to survey model, is trained to obtain using preset model；Step S300 is obtained often respectively
The current load data of individual base station；Step S400 calculates the prediction load of each base station respectively；The prediction of each base station loads
According to load estimation model corresponding to the current load data of the base station and the base station, the base station being calculated is in preset time
Average load in section；Step S500 loads when needing accessing user according to the present load of each base station and prediction, by the
Two preparatory conditions select a base station, the target BS to be accessed as access customer waiting.
In the above-mentioned technical solutions, the present load of base station is not only allowed for, it is also contemplated that the historic load of base station,
According to historic load, the load estimation model of base station is obtained, the prediction for calculation base station loads；When needing accessing user
When, consider the present load for covering all base stations of the user and prediction loading condition, plan as a whole distribution, this contributes to reality
The load balancing of existing the whole network, so as to lift overall network performance.
Further, the process that load estimation model is trained to obtain in each base station in the step S200 includes：Step S210
According to the training sample data and GBDT models of the base station, obtain influenceing the combinations of features of the load of base station；Step S220 according to
Obtained combinations of features, using Multiple regression model, regression function is obtained, using each feature in combinations of features as recurrence
Each characteristic component in function；Step S230, using gradient descent method, obtains returning letter according to the training sample data of the base station
The value of each parameter in number；Step S240 will substitute into load estimation mould of the regression function as base station of the value of each parameter
Type.
Further, the regression function in the step S220 is：
hθ(x)=θTX=θ0x0+θ1x1+...+θnxn……………………………(1)
Wherein, x0For 1, x1,x2,...,xnEach feature in representative feature combination, x=[x0,x1,x2,...,xn]T；θ0,
θ1,θ2...,θnFor parameter, θ=[θ0,θ1,θ2,...,θn]T；
According to the training sample data of the base station in the step S230, using gradient descent method, obtain in regression function
Function is as follows used by the value of each parameter：
Wherein, i represents i-th of training sample in training sample data, and m is the total sample number in training sample data；
hθ(xi) loaded for prediction corresponding to i-th of training sample, yiRepresent the actual loading in i-th of training sample；When J (θ) is minimum
When, θ values corresponding to J (θ) are the value of each parameter in regression function.
In the above-mentioned technical solutions, describe and the prediction load module of each base station is obtained by GBDT+LR model trainings
Process, while further illustrate the present load that this programme not only allows for base station, it is also contemplated that the historic load of base station
Data.
Further, the step S500 includes：Step S510, according to the customer location, obtains when needing accessing user
Cover all base stations of the user；Step S520 obtain respectively cover the user each self-corresponding present load in all base stations and
Prediction load；Step S530 presses the second preparatory condition, from all base stations for covering the user, a base station is selected, as treating
Accessing user's target BS to be accessed.
In the above-mentioned technical solutions, according to the position of access customer waiting, all base stations for covering the user are obtained, synthesis is examined
The present load and prediction for considering these base stations load, then plan as a whole to distribute, and this contributes to the load balancing for realizing the whole network, so as to be lifted
Overall network performance.
Further, the step S530 further comprises：From cover the user all base stations in, select present load and
Prediction load is respectively less than default load threshold and predicts a minimum base station of load, the mesh to be accessed as access customer waiting
Mark base station.
In the above-mentioned technical solutions, present load and prediction load is selected to be respectively less than default load threshold and prediction load
A minimum base station, the target BS to be accessed as access customer waiting, be advantageous to the load balancing of the whole network.
Further, also include after the step S500：Step S600 is when in the presence of the arrival load estimation model modification time
Base station when, update the training sample data of the base station, and the load for updating according to the training sample data of renewal the base station is pre-
Survey model.
In the above-mentioned technical solutions, the load estimation model of base station is updated, new model can more accurately reflect
The present case of base station, can be closer to the present case of base station using the load of new model prediction.
The present invention also provides a kind of load of base station balanced management system, including：Acquisition module, it is each for obtaining respectively
The training sample data of base station, the training sample data of each base station are chosen from historic load storehouse corresponding to the base station
The historic load for meeting the first preparatory condition；Model training module, electrically connected with the acquisition module, for instructing respectively
The load estimation model of each base station is got, the load estimation model of each base station is the number of training according to the base station
According to being trained to obtain using preset model；The acquisition module, it is further used for obtaining the present load of each base station respectively
Data；Computing module, electrically connected with the model training module, the prediction for calculating each base station respectively loads, Mei Geji
The prediction load stood is the load estimation model according to corresponding to the current load data of the base station and the base station, and what is be calculated should
Average load of the base station in preset time period；The acquisition module, when being further used for needing accessing user, according to each base
The present load and prediction stood are loaded, and a base station, the target to be accessed as access customer waiting are selected by the second preparatory condition
Base station.
In the above-mentioned technical solutions, the present load of base station is not only allowed for, it is also contemplated that the historic load of base station,
According to historic load, the load estimation model of base station is obtained, the prediction for calculation base station loads；When needing accessing user
When, consider the present load for covering all base stations of the user and prediction loading condition, plan as a whole distribution, this contributes to reality
The load balancing of existing the whole network, so as to lift overall network performance.
Further, the process that load estimation model is trained to obtain in each base station in the model training module includes：According to
The training sample data and GBDT models of the base station, obtain influenceing the combinations of features of the load of base station；And according to obtained spy
Sign combination, using Multiple regression model, obtains regression function, using each feature in combinations of features as in regression function
Each characteristic component；And the training sample data according to the base station, using gradient descent method, obtain each parameter in regression function
Value；And will substitute into each parameter value regression function as base station load estimation model.
In the above-mentioned technical solutions, describe and the prediction load module of each base station is obtained by GBDT+LR model trainings
Process, while further illustrate the present load that the system not only allows for base station, it is also contemplated that the historic load of base station
Data.
Further, the acquisition module, it is further used for when needing accessing user, according to the customer location, acquisition is covered
Cover the base station of the user；And each self-corresponding present load in all base stations for covering the user and prediction load are obtained respectively；
And by the second preparatory condition, from all base stations for covering the user, a base station is selected, it is waiting as access customer waiting
The target BS entered.
In the above-mentioned technical solutions, according to the position of access customer waiting, all base stations for covering the user are obtained, synthesis is examined
The present load and prediction for considering these base stations load, then plan as a whole to distribute, and this contributes to the load balancing for realizing the whole network, so as to be lifted
Overall network performance.
Further, in addition to：The acquisition module, it is further used for when in the presence of the arrival load estimation model modification time
During base station, the training sample data of the base station are updated；The model training module, it is further used for the training sample according to renewal
Data update the load estimation model of the base station.
In the above-mentioned technical solutions, the load estimation model of base station is updated, new model can more accurately reflect
The present case of base station, can be closer to the present case of base station using the load of new model prediction.
By the management method and system that a kind of load of base station provided by the invention is balanced, following at least one can be brought
Beneficial effect：
1st, the present invention not only allows for the present load of base station, it is also contemplated that the historic load of base station, according to history
Load data, the load estimation model of base station is obtained, the prediction for calculation base station loads；It is comprehensive when needing accessing user
The present load for covering all base stations of the user and prediction loading condition are considered, plans as a whole distribution, this helps to realize the whole network
Load balancing, so as to lift overall network performance.
2nd, the present invention makes prediction load more accurately reflect working as base station by being updated to the load estimation model of base station
Preceding situation, so that the management method of load of base station equilibrium is capable of the change of adaptive environment.
Brief description of the drawings
Below by a manner of clearly understandable, preferred embodiment is described with reference to the drawings, balanced to a kind of load of base station
Above-mentioned characteristic, technical characteristic, advantage and its implementation of management method and system are further described.
Fig. 1 is a kind of flow chart of one embodiment of the balanced management method of load of base station of the present invention；
Fig. 2 is a kind of flow chart of another embodiment of the balanced management method of load of base station of the present invention；
Fig. 3 is a kind of flow chart of another embodiment of the balanced management method of load of base station of the present invention；
Fig. 4 is a kind of structural representation of one embodiment of the balanced management system of load of base station of the present invention.
Drawing reference numeral explanation：
10. acquisition module, 20. model training modules, 30. computing modules.
Embodiment
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, control is illustrated below
The embodiment of the present invention.It should be evident that drawings in the following description are only some embodiments of the present invention, for
For those of ordinary skill in the art, on the premise of not paying creative work, other can also be obtained according to these accompanying drawings
Accompanying drawing, and obtain other embodiments.
To make simplified form, part related to the present invention is only schematically show in each figure, they are not represented
Its practical structures as product.In addition, so that simplified form readily appreciates, there is identical structure or function in some figures
Part, one of those is only symbolically depicted, or only marked one of those.Herein, "one" is not only represented
" only this ", the situation of " more than one " can also be represented.
In one embodiment of the invention, as shown in figure 1, a kind of balanced management method of load of base station, including：
Step S100 obtains the training sample data of each base station respectively；
The training sample data of each base station be chosen from historic load storehouse corresponding to the base station meet first
The historic load of preparatory condition；
Step S200 is respectively trained to obtain the load estimation model of each base station；
The load estimation model of each base station is the training sample data according to the base station, is trained using preset model
Obtain；
Step S300 obtains the current load data of each base station respectively；
Step S400 calculates the prediction load of each base station respectively；
The prediction load of each base station is the load estimation mould according to corresponding to the current load data of the base station and the base station
Type, average load of the base station being calculated in preset time period；
Step S500 loads when needing accessing user according to the present load of each base station and prediction, default by second
Condition selects a base station, the target BS to be accessed as access customer waiting.
Specifically, the historic load storehouse of base station is made up of historic load, i.e., gathered in historical time
Load of base station data form, wherein, load of base station data contain the value for the various features for influenceing load of base station.Influence base station
The factor of load is a lot, for example time, load of base station, base station user number, base station use the dynamic variable quantity such as power, and base station
Clear and definite static amount during planning, as the affiliated place in base station, ultimate load, the maximum transmission power allowed, frequency range, with
And lobe width etc., these factors can all as the feature for influenceing load of base station, can also empirically selected part, from
And reduce the complexity of training.
First preparatory condition can be to choose data volume to meet default training sample duration requirement, and occur in current time
Load data before, as meet the historic load of the first preparatory condition.Each base station is from respective historic load number
According in storehouse, historic load is chosen, as training sample data.During training sample in the first preparatory condition of each base station
Length can be with identical, can also be different, for the sake of simplicity, suggesting identical.For example current time is August 31, the training of all base stations
Sample duration is set to 3 months, can be from the historic load of each base station in order to predict September 1 day and later load of base station
The load of base station data in May-July are chosen in storehouse, can also choose the load of base station data of June-August, the instruction as each base station
Practice sample data.Than the former more preferably, the nearer data of chosen distance current time can more reflect the present case of base station to the latter.Instruction
Practice sample duration nor the longer the better, data remote influence smaller on the load estimation of current base station, accurate to prediction
The lifting of true rate and unobvious.First preparatory condition can also be to choose that bearing before current time occurs by preset data amount
Data are carried, as meet the historic load of the first preparatory condition.The preset data amount of different base stations can also may be used with identical
With difference, for the sake of simplicity, suggesting identical.Preset data amount is nor be the bigger the better, setting of the reason with training sample duration.
Preset model, various classification in machine learning, regression model can be fully used for reference, such as, lifted using gradient
(i.e. GBDT) and logistic regression (i.e. LR) model combination are set, maximum first is influenceed on load of base station with GBDT model extractions
Combinations of features, reuse LR models, obtain regression function, but contain some unknown parameters in the function, it is necessary to by using
Training sample data learn, and estimate the optimal solution of these parameters, the value using this optimal solution as parameter, are updated to former return
Return function, so far model is just successfully established；Preset model, the combination of random forest and LR models can also be used, first with random
Forest extraction influences maximum combinations of features on load of base station, reuses LR models, obtains regression function, recycles gradient to decline
Method iteration obtains the optimal solution of the parameter in the regression function, the value using this optimal solution as parameter, is updated to former recurrence
Function, so far model be successfully established；GBDT models in foregoing can also be replaced with other models, such as XGBoost；In foregoing
LR models can also be replaced with other regression models, such as, linear regression, but LR models compared to be easier realize, so pushing away
Recommend LR models；Preset model can be with identical used by each base station, can also be different, for the sake of simplicity, suggesting identical.Each base
Stand, according to the training sample data and preset model of the base station, obtain load estimation model corresponding to the base station；Follow-up basis should
Load estimation model carries out load estimation.
During training obtains the load estimation model of each base station, multigroup window data can also be taken to instruct parallel
Practice, by cross validation, selecting the high model of accuracy rate, this can be with the general Huaneng Group power of lift scheme as final mask.Such as
According to following one day load of base station of trimestral historic load prediction, it is assumed that the load of base station of prediction September 1, from 5
Month No. 1 arrives the nearest load data of four months of August 31, and data are divided into 3 windows, sliding window duration two weeks, successively
Obtain the data window (corresponding window 1) of 1 day-August in May 1, or the data window (corresponding window 2) of 15 days-August in May 15, or 5
The data window (corresponding window 3) of month No. 30-August 30, it is trained to obtain pair according to trimestral data respectively per group window
The temporary pattern answered, the mode for reusing cross validation are verified the accuracy rate of model, such as tested with all or part of data of window 2
The model of the corresponding window 1 of card, with the model of the data verification window 2 of window 3, with the model of the data verification window 3 of window 1,
The high temporary pattern of accuracy rate is chosen, as final load estimation model.
The prediction load of each base station is calculated respectively, wherein, the prediction load of each base station is according to the current of the base station
Load estimation model corresponding to load data and the base station, average load of the base station being calculated in preset time period.
For example preset time period is 3 days, inputs the August data of 31 days, it is desirable to the 1 day-September average load of 3 days in September is predicted, if
Load estimation model, it is the load of base station data prediction load of next day according to one day, then needs the negative of continuous prediction 3 days
Carry, this 3 days load is averaging, obtain required prediction load；Loaded according to required prediction, selection target base
Stand, it is assumed that the second preparatory condition is " selection present load and prediction load are respectively less than default load threshold, and predict that load is minimum
Base station ", then show not to be intended merely to select present load light, and following 3 days average load also lighter base station as mesh
Mark base station.The preset time period of each base station can be with identical, can also be different, it is proposed that and it is identical, be advantageous to subsequently select target base
When standing, referenced parameter is obtained according to consistent mode.
If current time needs accessing user, loaded according to the present load of each base station and prediction, it is pre- by second
If condition selects a base station, the target BS to be accessed as access customer waiting.Each base station, refer to cover the user
All base stations；Under heterogeneous network, overlapping covering be present；In the cellular network that tradition is made up of macro base station,
Overlapping covering be present；The present embodiment is applied to the scene that overlapping covering be present.Second preparatory condition, can set as needed
Put, such as, a kind of mode is to select present load and prediction to load and be respectively less than default load threshold, and predicts load minimum
Base station；Another way is that preferred present load and prediction load are respectively less than the base station for presetting load threshold, then therefrom selection is used
The best base station of family channel condition；The former is load minimum principle, and the latter is to load within the specific limits, and priority channel condition is good
Base station.The user is probably new access user, it is also possible to the user that migration comes from other base stations.
The prediction load of each base station is calculated respectively, can be when needing accessing user, then perform, can also be only for covering
All base stations for covering the user perform；For example needing accessing user 1, base station 1,2,3,4 covers the user, can detect
When needing accessing user, then go to obtain the present load of base station 1,2,3,4 and calculate corresponding prediction load, then it is default by second
Condition selects target BS；If current time need accessing user 2, base station 1,2,5,6 covers the user, in view of base station 1,
2 present load obtained with it is corresponding prediction load has calculated, then need to only be directed to base station 5,6, then go obtain present load with
Prediction load corresponding to calculating.
The present embodiment is loaded by the prediction for first calculating all base stations, then when needing accessing user, then go pre- using this
The mode for surveying load describes, advantage of this is that when needing accessing user, is directly loaded using the prediction calculated, can
To be rapidly that user selects target BS, Consumer's Experience may be more preferable.
In another embodiment of the present invention, as shown in Fig. 2 a kind of balanced management method of load of base station, including：
Step S100 obtains the training sample data of each base station respectively；
The training sample data of each base station be chosen from historic load storehouse corresponding to the base station meet first
The historic load of preparatory condition；
Step S200 is respectively trained to obtain the load estimation model of each base station；
The load estimation model of each base station is the training sample data according to the base station, is trained using preset model
Obtain；
Step S300 obtains the current load data of each base station respectively；
Step S400 calculates the prediction load of each base station respectively；
The prediction load of each base station is the load estimation mould according to corresponding to the current load data of the base station and the base station
Type, average load of the base station being calculated in preset time period；
Step S500 loads when needing accessing user according to the present load of each base station and prediction, default by second
Condition selects a base station, the target BS to be accessed as access customer waiting；
The process that load estimation model is trained to obtain in each base station in the step S200 includes：
Step S210 obtains influenceing the feature group of the load of base station according to the training sample data and GBDT models of the base station
Close；
Step S220 using Multiple regression model, obtains regression function, by feature group according to obtained combinations of features
Each feature in conjunction is as each characteristic component in regression function；
Step S230, using gradient descent method, obtains each parameter in regression function according to the training sample data of the base station
Value；
Step S240 will substitute into load estimation model of the regression function as base station of the value of each parameter；
Regression function in the step S220 is：
hθ(x)=θTX=θ0x0+θ1x1+...+θnxn……………………………(1)
Wherein, x0For 1, x1,x2,...,xnEach feature in representative feature combination, x=[x0,x1,x2,...,xn]T；θ0,
θ1,θ2...,θnFor parameter, θ=[θ0,θ1,θ2,...,θn]T；
According to the training sample data of the base station in the step S230, using gradient descent method, obtain in regression function
Function is as follows used by the value of each parameter：
Wherein, i represents i-th of training sample in training sample data, and m is the total sample number in training sample data；
hθ(xi) loaded for prediction corresponding to i-th of training sample, yiRepresent the actual loading in i-th of training sample；When J (θ) is minimum
When, θ values corresponding to J (θ) are the value of each parameter in regression function；
The step S500 includes：
Step S510, according to the customer location, obtains all base stations for covering the user when needing accessing user；Step
S520 obtains each self-corresponding present load in all base stations for covering the user respectively and prediction loads；Step S530 is pre- by second
If condition, from all base stations for covering the user, a base station, the target BS to be accessed as access customer waiting are selected；
The step S530 further comprises：
From all base stations for covering the user, select present load and prediction load be respectively less than default load threshold and
A minimum base station of prediction load, the target BS to be accessed as access customer waiting.
Specifically, relatively previous embodiment, adds step S210-S240, GBDT is employed to each base station respectively
With the combination of LR models, it is trained to obtain each self-corresponding load estimation model.Train to obtain only for single base station below
The process of load estimation model is described, other base station similar operations.
GBDT (Gradient Boost Decision Tree) is a kind of conventional nonlinear model, and it is based on integrated learn
Boosting thoughts in habit, each iteration all newly establish a decision tree, iteration how many times in the gradient direction for reducing residual error
How many decision trees will be generated.GBDT thought make it have inherent advantage can be found that a variety of features for having a distinction and
Combinations of features, the path of decision tree can be defeated directly as LR (Logistic Regression, Multiple regression model)
Enter feature use, eliminate artificial the step of finding feature, combinations of features.LR is a kind of linear fit model, can be utilized
Logistic functions become grader.
Maximum combinations of features is first influenceed on load of base station with GBDT model extractions, by this combinations of features, is input to logic
This base of a fruit regression model, the regression function for reflecting load of base station is obtained, but contain some unknown parameters, i.e. θ in the function；Pass through
Gradient descent method, according to the training sample data of base station, find the θ values for making the function of formula (2) minimum, i.e. θ optimal solution；
Value using this θ optimal solution as each parameter, is updated to former regression function, so far the just success of the load estimation model of base station
Establish；Load estimation model of the fixed regression function of parameter as base station.
Step S510-S530 is added, when needing accessing user, according to the customer location, obtains and covers the user's
All base stations；For these base stations, each self-corresponding present load is obtained respectively and prediction loads；The second preparatory condition is pressed again,
User's target BS to be accessed is selected from these base stations.
The second preparatory condition in the present embodiment is, select present load and prediction load be respectively less than default load threshold,
And the base station that prediction load is minimum.The setting of default load threshold, is in order to which base station is divided into heavily loaded base station and underloading base station two
Class.If the present load of base station is less than default load threshold, show that base station currently belongs to underloading；If the prediction load of base station
Less than default load threshold, show to fall within underloading under base station for a period of time；Preferably select current time and lower a period of time all
It is the base station of underloading, is so advantageous to the load balancing of whole net, so as to lifts the utilization rate of whole net resource.
In another embodiment of the present invention, as shown in figure 3, a kind of management method of load of base station equilibrium includes：
Step S100 obtains the training sample data of each base station respectively；
The training sample data of each base station be chosen from historic load storehouse corresponding to the base station meet first
The historic load of preparatory condition；
Step S200 is respectively trained to obtain the load estimation model of each base station；
The load estimation model of each base station is the training sample data according to the base station, is trained using preset model
Obtain；
Step S300 obtains the current load data of each base station respectively；
Step S400 calculates the prediction load of each base station respectively；
The prediction load of each base station is the load estimation mould according to corresponding to the current load data of the base station and the base station
Type, average load of the base station being calculated in preset time period；
Step S500 loads when needing accessing user according to the present load of each base station and prediction, default by second
Condition selects a base station, the target BS to be accessed as access customer waiting；
The process that load estimation model is trained to obtain in each base station in the step S200 includes：
Step S210 obtains influenceing the feature group of the load of base station according to the training sample data and GBDT models of the base station
Close；
Step S220 using Multiple regression model, obtains regression function, by feature group according to obtained combinations of features
Each feature in conjunction is as each characteristic component in regression function；
Step S230, using gradient descent method, obtains each parameter in regression function according to the training sample data of the base station
Value；
Step S240 will substitute into load estimation model of the regression function as base station of the value of each parameter；
Regression function in the step S220 is：
hθ(x)=θTX=θ0x0+θ1x1+...+θnxn……………………………(1)
Wherein, x0For 1, x1,x2,...,xnEach feature in representative feature combination, x=[x0,x1,x2,...,xn]T；θ0,
θ1,θ2...,θnFor parameter, θ=[θ0,θ1,θ2,...,θn]T；
According to the training sample data of the base station in the step S230, using gradient descent method, obtain in regression function
Function is as follows used by the value of each parameter：
Wherein, i represents i-th of training sample in training sample data, and m is the total sample number in training sample data；
hθ(xi) loaded for prediction corresponding to i-th of training sample, yiRepresent the actual loading in i-th of training sample；When J (θ) is minimum
When, θ values corresponding to J (θ) are the value of each parameter in regression function；
The step S500 includes：
Step S510, according to the customer location, obtains all base stations for covering the user when needing accessing user；
Step S520 obtains each self-corresponding present load in all base stations for covering the user respectively and prediction loads；
Step S530 presses the second preparatory condition, from all base stations for covering the user, a base station is selected, as waiting
Access customer target BS to be accessed；
The step S530 further comprises：
From all base stations for covering the user, select present load and prediction load be respectively less than default load threshold and
A minimum base station of prediction load, the target BS to be accessed as access customer waiting；
Also include after the step S500：
Step S600 updates the number of training of the base station when the base station for reaching the load estimation model modification time be present
According to, and update according to the training sample data of renewal the load estimation model of the base station.
Specifically, relatively previous embodiment, embodiment adds step S600, introduces load estimation model more
Newly.
The load of base station data that each time is gathered are updated into the historic load storehouse of base station, over time
Elapse, there are many new data in the historic load storehouse of base station, therefore when arrival load estimation model modification be present
Between base station when, the training sample data of the base station may be updated, and the negative of the base station is updated according to the training sample data of renewal
Carry forecast model.The load estimation model modification time of each base station can be set to different, can also be set to identical, some base stations
It can also be not provided with.
The update frequency of load estimation model is relatively low, but if the environment around base station is changed, then model
Need to refresh.Such as a new resident living area, occupied less initial stage from inhabitation, to gradually occupying, then to cell into
It is ripe, there are many people to live, in the meantime, the load of base station is more big changes, and corresponding load of base station forecast model should in time more
Newly.
In another embodiment of the present invention, as shown in figure 4, a kind of balanced management system of load of base station, including：
Acquisition module 10, for obtaining the training sample data of each base station, the training sample data of each base station respectively
It is that to be chosen from historic load storehouse corresponding to the base station meet the historic load of the first preparatory condition；
Model training module 20, electrically connected with the acquisition module 10, for being respectively trained to obtain the load of each base station
Forecast model, the load estimation model of each base station are the training sample data according to the base station, are instructed using preset model
Get；
The acquisition module 10, it is further used for obtaining the current load data of each base station respectively；
Computing module 30, being electrically connected with the model training module 20, the prediction for calculating each base station respectively loads,
The prediction load of each base station is the load estimation model according to corresponding to the current load data of the base station and the base station, is calculated
Average load of the base station arrived in preset time period；
The acquisition module 10, when being further used for needing accessing user, according to the present load and prediction of each base station
Load, a base station, the target BS to be accessed as access customer waiting are selected by the second preparatory condition.
Specifically, the historic load storehouse of base station is made up of historic load, i.e., gathered in historical time
Load of base station data form, wherein, load of base station data contain the value for the various features for influenceing load of base station.Influence base station
The factor of load is a lot, for example time, load of base station, base station user number, base station use the dynamic variable quantity such as power, and base station
Clear and definite static amount during planning, as the affiliated place in base station, ultimate load, the maximum transmission power allowed, frequency range, with
And lobe width etc., these factors can all as the feature for influenceing load of base station, can also empirically selected part, from
And reduce the complexity of training.
First preparatory condition can be to choose data volume to meet default training sample duration requirement, and occur in current time
Load data before, as meet the historic load of the first preparatory condition.Each base station is from respective historic load number
According in storehouse, historic load is chosen, as training sample data.During training sample in the first preparatory condition of each base station
Length can be with identical, can also be different, for the sake of simplicity, suggesting identical.For example current time is August 31, the training of all base stations
Sample duration is set to 3 months, can be from the historic load of each base station in order to predict September 1 day and later load of base station
The load of base station data in May-July are chosen in storehouse, can also choose the load of base station data of June-August, the instruction as each base station
Practice sample data.Than the former more preferably, the nearer data of chosen distance current time can more reflect the present case of base station to the latter.Instruction
Practice sample duration nor the longer the better, data remote influence smaller on the load estimation of current base station, accurate to prediction
The lifting of true rate and unobvious.First preparatory condition can also be to choose that bearing before current time occurs by preset data amount
Data are carried, as meet the historic load of the first preparatory condition.The preset data amount of different base stations can also may be used with identical
With difference, for the sake of simplicity, suggesting identical.Preset data amount is nor be the bigger the better, setting of the reason with training sample duration.
Preset model, various classification in machine learning, regression model can be fully used for reference, such as, lifted using gradient
(i.e. GBDT) and logistic regression (i.e. LR) model combination are set, maximum first is influenceed on load of base station with GBDT model extractions
Combinations of features, reuse LR models, obtain regression function, but contain some unknown parameters in the function, it is necessary to by using
Training sample data learn, and estimate the optimal solution of these parameters, the value using this optimal solution as parameter, are updated to
Former regression function, so far model be just successfully established；Preset model, the combination of random forest and LR models can also be used, is first used
Random forest extraction influences maximum combinations of features on load of base station, reuses LR models, obtains regression function, recycles gradient
Descent method iteration obtains the optimal solution of the parameter in the regression function, the value using this optimal solution as parameter, is updated to original
Regression function, so far model be successfully established；GBDT models in foregoing can also be replaced with other models, such as XGBoost；Before
LR models in stating can also be replaced with other regression models, such as, linear regression, but LR models are compared to easily realization, institute
To recommend LR models；Preset model can be with identical used by each base station, can also be different, for the sake of simplicity, suggesting identical.Often
Individual base station, according to the training sample data and preset model of the base station, obtain load estimation model corresponding to the base station；Follow-up root
Load estimation is carried out according to the load estimation model.
During training obtains the load estimation model of each base station, multigroup window data can also be taken to instruct parallel
Practice, by cross validation, selecting the high model of accuracy rate, this can be with the general Huaneng Group power of lift scheme as final mask.Such as
According to following one day load of base station of trimestral historic load prediction, it is assumed that the load of base station of prediction September 1, from 5
Month No. 1 arrives the nearest load data of four months of August 31, and data are divided into 3 windows, sliding window duration two weeks, successively
The data window (corresponding window 1) of 1 day-August in May 1, or the data window (corresponding window 2) of 15 days-August in May 15 are obtained, or
May 30 days-August 30 data window (corresponding window 3), be trained to obtain according to trimestral data respectively per group window
Corresponding temporary pattern, the mode for reusing cross validation verify the accuracy rate of model, such as use all or part of data of window 2
The model of the corresponding window 1 of checking, with the model of the data verification window 2 of window 3, with the mould of the data verification window 3 of window 1
Type, the high temporary pattern of accuracy rate is chosen, as final load estimation model.
The prediction load of each base station is calculated respectively, wherein, the prediction load of each base station is according to the current of the base station
Load estimation model corresponding to load data and the base station, average load of the base station being calculated in preset time period.
For example preset time period is 3 days, inputs the August data of 31 days, it is desirable to the 1 day-September average load of 3 days in September is predicted, if
Load estimation model, it is the load of base station data prediction load of next day according to one day, then needs the negative of continuous prediction 3 days
Carry, this 3 days load is averaging, obtain required prediction load；Loaded according to required prediction, selection target base
Stand, it is assumed that the second preparatory condition is " selection present load and prediction load are respectively less than default load threshold, and predict that load is minimum
Base station ", then show not to be intended merely to select present load light, and following 3 days average load also lighter base station as mesh
Mark base station.The preset time period of each base station can be with identical, can also be different, it is proposed that and it is identical, be advantageous to subsequently select target base
When standing, referenced parameter is obtained according to consistent mode.
If current time needs accessing user, loaded according to the present load of each base station and prediction, it is pre- by second
If condition selects a base station, the target BS to be accessed as access customer waiting.Each base station, refer to cover the user
All base stations；Under heterogeneous network, overlapping covering be present；In the cellular network that tradition is made up of macro base station,
Overlapping covering be present；The present embodiment is applied to the scene that overlapping covering be present.Second preparatory condition, can set as needed
Put, such as, a kind of mode is to select present load and prediction to load and be respectively less than default load threshold, and predicts load minimum
Base station；Another way is that preferred present load and prediction load are respectively less than the base station for presetting load threshold, then therefrom selection is used
The best base station of family channel condition；The former is load minimum principle, and the latter is to load within the specific limits, and priority channel condition is good
Base station.The user is probably new access user, it is also possible to the user that migration comes from other base stations.
The prediction load of each base station is calculated respectively, can be when needing accessing user, then perform, can also be only for covering
All base stations for covering the user perform；For example needing accessing user 1, base station 1,2,3,4 covers the user, can detect
When needing accessing user, then go to obtain the present load of base station 1,2,3,4 and calculate corresponding prediction load, then it is default by second
Condition selects target BS；If current time need accessing user 2, base station 1,2,5,6 covers the user, in view of base station 1,
2 present load obtained with it is corresponding prediction load has calculated, then need to only be directed to base station 5,6, then go obtain present load with
Prediction load corresponding to calculating.
The present embodiment is loaded by the prediction for first calculating all base stations, then when needing accessing user, then go pre- using this
The mode for surveying load describes, advantage of this is that when needing accessing user, is directly loaded using the prediction calculated, can
To be rapidly that user selects target BS, Consumer's Experience may be more preferable.
In another embodiment of the present invention, as shown in figure 4, a kind of balanced management system of load of base station, including：
Acquisition module 10, for obtaining the training sample data of each base station, the training sample data of each base station respectively
It is that to be chosen from historic load storehouse corresponding to the base station meet the historic load of the first preparatory condition；
Model training module 20, electrically connected with the acquisition module 10, for being respectively trained to obtain the load of each base station
Forecast model, the load estimation model of each base station are the training sample data according to the base station, are instructed using preset model
Get；
The acquisition module 10, it is further used for obtaining the current load data of each base station respectively；
Computing module 30, being electrically connected with the model training module 20, the prediction for calculating each base station respectively loads,
The prediction load of each base station is the load estimation model according to corresponding to the current load data of the base station and the base station, is calculated
Average load of the base station arrived in preset time period；
The acquisition module 10, when being further used for needing accessing user, according to the present load and prediction of each base station
Load, a base station, the target BS to be accessed as access customer waiting are selected by the second preparatory condition；
The process that load estimation model is trained to obtain in each base station in the model training module 20 includes：
According to the training sample data and GBDT models of the base station, obtain influenceing the combinations of features of the load of base station；And
According to obtained combinations of features, using Multiple regression model, obtain regression function, using each feature in combinations of features as
Each characteristic component in regression function；And the training sample data according to the base station, using gradient descent method, returned
The value of each parameter in function；And will substitute into each parameter value regression function as base station load estimation model；
The acquisition module 10, it is further used for when needing accessing user, according to the customer location, obtains and cover the use
The base station at family；And each self-corresponding present load in all base stations for covering the user and prediction load are obtained respectively；And
By the second preparatory condition, from all base stations for covering the user, a base station, the mesh to be accessed as access customer waiting are selected
Mark base station；
The acquisition module 10, it is further used for from all base stations for covering the user, selects present load and prediction
Load is respectively less than default load threshold and prediction loads a minimum base station, the target base to be accessed as access customer waiting
Stand；
The acquisition module 10, it is further used for when the base station for reaching the load estimation model modification time be present, renewal
The training sample data of the base station；
The model training module 20, the load for being further used for updating the base station according to the training sample data of renewal are pre-
Survey model.
Specifically, relatively previous embodiment, has done further to the function of acquisition module 10, model training module 20
Enhancing.
The model training module 20, it is further used for the combination respectively to each base station using GBDT and LR models, enters
Row training obtains each self-corresponding load estimation model；
The acquisition module 10, it is further used for when needing accessing user, according to the customer location, obtains and cover the use
All base stations at family；For these base stations, each self-corresponding present load is obtained respectively and prediction loads；Again by the second default bar
Part, user's target BS to be accessed is selected from these base stations.The second preparatory condition in the present embodiment is to select current bear
Carry and prediction loads the base station for being respectively less than default load threshold and prediction load minimum.The setting of default load threshold, be in order to
Base station is divided into heavily loaded base station and the class of underloading base station two.If the present load of base station is less than default load threshold, show base station
Currently belong to underloading；If the prediction load of base station is less than default load threshold, show to fall within underloading under base station for a period of time；
Preferably select current time and lower a period of time be all underloading base station, be so advantageous to the load balancing of whole net, so as to be lifted
The utilization rate of whole net resource.
The acquisition module 10 and the model training module 20, are further used for the renewal of load estimation model；Will be every
The load of base station data that the individual time is gathered are updated into the historic load storehouse of base station, over time, base station
There are many new data in historic load storehouse, therefore when the base station for reaching the load estimation model modification time be present,
The training sample data of the base station may be updated, and the load estimation model of the base station is updated according to the training sample data of renewal.
The load estimation model modification time of each base station can be set to different, can also be set to identical, some base stations can not also be set
Put.
The update frequency of load estimation model is relatively low, but if the environment around base station is changed, then model
Need to refresh.Such as a new resident living area, occupied less initial stage from inhabitation, to gradually occupying, then to cell into
It is ripe, there are many people to live, in the meantime, the load of base station is more big changes, and corresponding load of base station forecast model should in time more
Newly.
It should be noted that above-described embodiment can independent assortment as needed.Described above is only the preferred of the present invention
Embodiment, it is noted that for those skilled in the art, do not departing from the premise of the principle of the invention
Under, some improvements and modifications can also be made, these improvements and modifications also should be regarded as protection scope of the present invention.
Claims (10)
- A kind of 1. balanced management method of load of base station, it is characterised in that including：Step S100 obtains the training sample data of each base station respectively；The training sample data of each base station be chosen from historic load storehouse corresponding to the base station to meet first default The historic load of condition；Step S200 is respectively trained to obtain the load estimation model of each base station；The load estimation model of each base station is the training sample data according to the base station, is trained using preset model Arrive；Step S300 obtains the current load data of each base station respectively；Step S400 calculates the prediction load of each base station respectively；The prediction load of each base station is the load estimation model according to corresponding to the current load data of the base station and the base station, is counted Average load of the obtained base station in preset time period；Step S500 is loaded, by the second preparatory condition when needing accessing user according to the present load of each base station and prediction Select a base station, the target BS to be accessed as access customer waiting.
- 2. the balanced management method of load of base station according to claim 1, it is characterised in that each in the step S200 The process that load estimation model is trained to obtain in base station includes：Step S210 obtains influenceing the combinations of features of the load of base station according to the training sample data and GBDT models of the base station；Step S220 using Multiple regression model, obtains regression function, by combinations of features according to obtained combinations of features Each feature as each characteristic component in regression function；Step S230 is according to the training sample data of the base station, using gradient descent method, obtains taking for each parameter in regression function Value；Step S240 will substitute into load estimation model of the regression function as base station of the value of each parameter.
- 3. the balanced management method of load of base station according to claim 2, it is characterised in that：Regression function in the step S220 is：hθ(x)=θTX=θ0x0+θ1x1+...+θnxn……………………………(1)Wherein, x0For 1, x1,x2,...,xnEach feature in representative feature combination, x=[x0,x1,x2,...,xn]T；θ0,θ1, θ2...,θnFor parameter, θ=[θ0,θ1,θ2,...,θn]T；According to the training sample data of the base station in the step S230, using gradient descent method, obtain respectively joining in regression function Function is as follows used by several values：Wherein, i represents i-th of training sample in training sample data, and m is the total sample number in training sample data；hθ(xi) To predict load, y corresponding to i-th of training sampleiRepresent the actual loading in i-th of training sample；When J (θ) is minimum, J θ values corresponding to (θ) are the value of each parameter in regression function.
- 4. the balanced management method of load of base station according to claim 1, it is characterised in that the step S500 includes：Step S510, according to the customer location, obtains all base stations for covering the user when needing accessing user；Step S520 obtains each self-corresponding present load in all base stations for covering the user respectively and prediction loads；Step S530 presses the second preparatory condition, a base station is selected from all base stations for covering the user, as use to be accessed Family target BS to be accessed.
- 5. the balanced management method of load of base station according to claim 4, it is characterised in that the step S530 is further Including：From all base stations for covering the user, present load and prediction load is selected to be respectively less than default load threshold and predict Load a minimum base station, the target BS to be accessed as access customer waiting.
- 6. the balanced management method of load of base station according to claim 1, it is characterised in that after the step S500 also Including：Step S600 updates the training sample data of the base station when the base station for reaching the load estimation model modification time be present, And the load estimation model of the base station is updated according to the training sample data of renewal.
- 7. a kind of load of base station equilibrium of the management method balanced using any described load of base station of the claims 1-6 Management system, it is characterised in that including：Acquisition module, for obtaining the training sample data of each base station respectively, the training sample data of each base station are from this That is chosen in historic load storehouse corresponding to base station meets the historic load of the first preparatory condition；Model training module, electrically connected with the acquisition module, for being respectively trained to obtain the load estimation model of each base station, The load estimation model of each base station is the training sample data according to the base station, is trained to obtain using preset model；The acquisition module, it is further used for obtaining the current load data of each base station respectively；Computing module, electrically connected with the model training module, the prediction for calculating each base station respectively loads, each base station Prediction load be the load estimation model according to corresponding to the current load data of the base station and the base station, the base being calculated Stand the average load in preset time period；The acquisition module, when being further used for needing accessing user, loaded, pressed according to the present load of each base station and prediction Second preparatory condition selects a base station, the target BS to be accessed as access customer waiting.
- 8. the balanced management system of load of base station according to claim 7, it is characterised in that in the model training module The process that load estimation model is trained to obtain in each base station includes：According to the training sample data and GBDT models of the base station, obtain influenceing the combinations of features of the load of base station；And according to Obtained combinations of features, using Multiple regression model, regression function is obtained, using each feature in combinations of features as recurrence Each characteristic component in function；And the training sample data according to the base station, using gradient descent method, obtain regression function In each parameter value；And will substitute into each parameter value regression function as base station load estimation model.
- 9. the balanced management system of load of base station according to claim 7, it is characterised in that including：The acquisition module, it is further used for when needing accessing user, according to the customer location, obtains the base for covering the user Stand；And each self-corresponding present load in all base stations for covering the user and prediction load are obtained respectively；And by second Preparatory condition, from all base stations for covering the user, select a base station, the target base to be accessed as access customer waiting Stand.
- 10. the balanced management system of load of base station according to claim 7, it is characterised in that：The acquisition module, it is further used for, when the base station for reaching the load estimation model modification time be present, updating the base station Training sample data；The model training module, it is further used for updating the load estimation mould of the base station according to the training sample data of renewal Type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710860446.5A CN107517481B (en) | 2017-09-21 | 2017-09-21 | Base station load balancing management method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710860446.5A CN107517481B (en) | 2017-09-21 | 2017-09-21 | Base station load balancing management method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107517481A true CN107517481A (en) | 2017-12-26 |
| CN107517481B CN107517481B (en) | 2021-06-04 |
Family
ID=60725982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710860446.5A Active CN107517481B (en) | 2017-09-21 | 2017-09-21 | Base station load balancing management method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107517481B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108834079A (en) * | 2018-09-21 | 2018-11-16 | 北京邮电大学 | A kind of load balance optimization method based on mobility prediction in heterogeneous network |
| CN109559146A (en) * | 2018-09-25 | 2019-04-02 | 国家电网有限公司客户服务中心 | Electricity customer service center accesses data center's optimization method based on the provinces and cities of logistic model prediction potential user's quantity |
| WO2019144401A1 (en) | 2018-01-29 | 2019-08-01 | Telefonaktiebolaget Lm Ericsson (Publ) | Methods, apparatuses, computer programs and computer program products for load balancing |
| CN110190909A (en) * | 2019-06-06 | 2019-08-30 | 北京邮电大学 | A kind of signal equalizing method and device for optic communication |
| CN111225382A (en) * | 2018-11-27 | 2020-06-02 | 中国移动通信集团上海有限公司 | LTE functional characteristic selection method and system |
| CN111225392A (en) * | 2018-11-27 | 2020-06-02 | 中国移动通信集团辽宁有限公司 | Cell load balancing method, device, equipment and computer storage medium |
| CN111385128A (en) * | 2018-12-29 | 2020-07-07 | 中兴通讯股份有限公司 | Method and device for predicting burst load, storage medium, and electronic device |
| CN112422443A (en) * | 2020-10-27 | 2021-02-26 | 鹏城实验室 | Adaptive control method, storage medium, equipment and system of congestion algorithm |
| CN112637904A (en) * | 2019-09-24 | 2021-04-09 | 中国移动通信集团陕西有限公司 | Load balancing method and device and computing equipment |
| CN114189475A (en) * | 2021-12-15 | 2022-03-15 | 中国电信股份有限公司 | Network load optimization method and device, computer storage medium and electronic equipment |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023224576A1 (en) * | 2022-05-17 | 2023-11-23 | Telefonaktiebolaget Lm Ericsson (Publ) | Managing unit and method in a communications network |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101860885A (en) * | 2010-06-11 | 2010-10-13 | 西安电子科技大学 | Access network selection method based on neural network and fuzzy logic |
| CN102711177A (en) * | 2012-04-26 | 2012-10-03 | 北京邮电大学 | Service prediction based load balancing method |
| CN103533577A (en) * | 2013-08-09 | 2014-01-22 | 国家电网公司 | Judgment control method and system for vertical switching among heterogeneous wireless networks |
| CN103763747A (en) * | 2014-02-21 | 2014-04-30 | 重庆邮电大学 | Method for achieving dynamic load balancing in heterogeneous network |
| CN103781120A (en) * | 2014-02-25 | 2014-05-07 | 重庆邮电大学 | Mobility load balancing method combined with hierarchical dynamic resource distribution |
| CN103778474A (en) * | 2012-10-18 | 2014-05-07 | 华为技术有限公司 | Resource load capacity prediction method, analysis prediction system and service operation monitoring system |
| CN104519528A (en) * | 2014-12-12 | 2015-04-15 | 京信通信系统（中国）有限公司 | Method and device for determining unloading neighbor area |
| CN104751374A (en) * | 2015-03-27 | 2015-07-01 | 深圳供电局有限公司 | Metering automation system wireless communication failure prediction method |
| CN105246103A (en) * | 2015-09-23 | 2016-01-13 | 广东工业大学 | Load balancing access method for wireless local area network |
| WO2017035519A1 (en) * | 2015-08-27 | 2017-03-02 | Skytree, Inc. | Supervised learning based recommendation system |
| CN106602575A (en) * | 2015-10-14 | 2017-04-26 | 中国电力科学研究院 | Classified combined regulation and control method of user load equipments |
| CN106933650A (en) * | 2017-03-03 | 2017-07-07 | 北方工业大学 | load management method and system of cloud application system |
| US20170251046A1 (en) * | 2016-02-26 | 2017-08-31 | Pearson Education, Inc. | System and method for remote interface alert triggering |
| US20170264679A1 (en) * | 2016-03-11 | 2017-09-14 | International Business Machines Corporation | Load balancing based on user behavior prediction |
-
2017
- 2017-09-21 CN CN201710860446.5A patent/CN107517481B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101860885A (en) * | 2010-06-11 | 2010-10-13 | 西安电子科技大学 | Access network selection method based on neural network and fuzzy logic |
| CN102711177A (en) * | 2012-04-26 | 2012-10-03 | 北京邮电大学 | Service prediction based load balancing method |
| CN103778474A (en) * | 2012-10-18 | 2014-05-07 | 华为技术有限公司 | Resource load capacity prediction method, analysis prediction system and service operation monitoring system |
| CN103533577A (en) * | 2013-08-09 | 2014-01-22 | 国家电网公司 | Judgment control method and system for vertical switching among heterogeneous wireless networks |
| CN103763747A (en) * | 2014-02-21 | 2014-04-30 | 重庆邮电大学 | Method for achieving dynamic load balancing in heterogeneous network |
| CN103781120A (en) * | 2014-02-25 | 2014-05-07 | 重庆邮电大学 | Mobility load balancing method combined with hierarchical dynamic resource distribution |
| CN104519528A (en) * | 2014-12-12 | 2015-04-15 | 京信通信系统（中国）有限公司 | Method and device for determining unloading neighbor area |
| CN104751374A (en) * | 2015-03-27 | 2015-07-01 | 深圳供电局有限公司 | Metering automation system wireless communication failure prediction method |
| WO2017035519A1 (en) * | 2015-08-27 | 2017-03-02 | Skytree, Inc. | Supervised learning based recommendation system |
| CN105246103A (en) * | 2015-09-23 | 2016-01-13 | 广东工业大学 | Load balancing access method for wireless local area network |
| CN106602575A (en) * | 2015-10-14 | 2017-04-26 | 中国电力科学研究院 | Classified combined regulation and control method of user load equipments |
| US20170251046A1 (en) * | 2016-02-26 | 2017-08-31 | Pearson Education, Inc. | System and method for remote interface alert triggering |
| US20170264679A1 (en) * | 2016-03-11 | 2017-09-14 | International Business Machines Corporation | Load balancing based on user behavior prediction |
| CN106933650A (en) * | 2017-03-03 | 2017-07-07 | 北方工业大学 | load management method and system of cloud application system |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3747214A4 (en) * | 2018-01-29 | 2021-10-06 | Telefonaktiebolaget LM Ericsson (publ) | Methods, apparatuses, computer programs and computer program products for load balancing |
| WO2019144401A1 (en) | 2018-01-29 | 2019-08-01 | Telefonaktiebolaget Lm Ericsson (Publ) | Methods, apparatuses, computer programs and computer program products for load balancing |
| US11330494B2 (en) | 2018-01-29 | 2022-05-10 | Telefonaktiebolaget Lm Ericsson (Publ) | Methods, apparatuses, computer programs and computer program products for load balancing |
| CN108834079A (en) * | 2018-09-21 | 2018-11-16 | 北京邮电大学 | A kind of load balance optimization method based on mobility prediction in heterogeneous network |
| CN109559146A (en) * | 2018-09-25 | 2019-04-02 | 国家电网有限公司客户服务中心 | Electricity customer service center accesses data center's optimization method based on the provinces and cities of logistic model prediction potential user's quantity |
| CN109559146B (en) * | 2018-09-25 | 2022-11-04 | 国家电网有限公司客户服务中心 | Provincial and municipal access data center optimization method for predicting number of potential users by electric power customer service center based on logistic model |
| CN111225382A (en) * | 2018-11-27 | 2020-06-02 | 中国移动通信集团上海有限公司 | LTE functional characteristic selection method and system |
| CN111225392A (en) * | 2018-11-27 | 2020-06-02 | 中国移动通信集团辽宁有限公司 | Cell load balancing method, device, equipment and computer storage medium |
| CN111225382B (en) * | 2018-11-27 | 2023-04-25 | 中国移动通信集团上海有限公司 | LTE functional characteristic selection method and system |
| CN111385128A (en) * | 2018-12-29 | 2020-07-07 | 中兴通讯股份有限公司 | Method and device for predicting burst load, storage medium, and electronic device |
| CN110190909A (en) * | 2019-06-06 | 2019-08-30 | 北京邮电大学 | A kind of signal equalizing method and device for optic communication |
| CN112637904A (en) * | 2019-09-24 | 2021-04-09 | 中国移动通信集团陕西有限公司 | Load balancing method and device and computing equipment |
| CN112637904B (en) * | 2019-09-24 | 2024-04-12 | 中国移动通信集团陕西有限公司 | Load balancing method and device and computing equipment |
| CN112422443A (en) * | 2020-10-27 | 2021-02-26 | 鹏城实验室 | Adaptive control method, storage medium, equipment and system of congestion algorithm |
| CN114189475A (en) * | 2021-12-15 | 2022-03-15 | 中国电信股份有限公司 | Network load optimization method and device, computer storage medium and electronic equipment |
| CN114189475B (en) * | 2021-12-15 | 2024-03-19 | 中国电信股份有限公司 | Network load optimization method and device, computer storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107517481B (en) | 2021-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107517481A (en) | A kind of load of base station balanced management method and system | |
| JP3888954B2 (en) | CDMA wireless network planning and evaluation method and system | |
| CN107426759B (en) | Method and system for predicting data traffic of newly added base station | |
| CN103491544B (en) | The choosing method and device of sharing site | |
| CN109245840A (en) | Spectrum prediction method in cognitive radio system based on convolutional neural networks | |
| CN110033312A (en) | Generation method, device, equipment and the storage medium of room rate prediction model | |
| CN102640545A (en) | Telecommunications network node and methods | |
| US7844698B2 (en) | Wireless network modification support system and radio network modification support method | |
| CN107609583A (en) | Parameter optimization method, device, computer equipment and the storage medium of disaggregated model | |
| CN113498137A (en) | Method and device for obtaining cell relation model and recommending cell switching guide parameters | |
| CN108462965A (en) | Share planing method in a kind of steel tower demand site | |
| CN103813411A (en) | Game theory based multi-access bandwidth allocation method | |
| CN105790866B (en) | Base station rankings method and device | |
| CN106454857A (en) | Evaluation method and device for network planning | |
| CN102970163A (en) | Method and system for upgrading nodes in power communication backbone network | |
| CN109038605A (en) | A kind of electric network reactive-load planing method considering voltage responsive characteristic | |
| CN109145932A (en) | User's gender prediction's method, device and equipment | |
| Pervaiz et al. | Game theoretical formulation of network selection in competing wireless networks: An analytic hierarchy process model | |
| CN105744592A (en) | Service Access Method, Device and System Based on Heterogeneous Wireless Network | |
| CN104349326B (en) | Network plan method | |
| CN107197467A (en) | A kind of NB IoT wireless network deployment methods based on LTE | |
| CN107689884A (en) | A kind of method of half network plan simulation in kind of power communication | |
| CN109561436B (en) | Physical layer cell identity (PCI) optimization method, equipment, computer readable storage medium and device | |
| CN103002458B (en) | Base station selection method and device | |
| Arifin et al. | The prediction of mobile data traffic based on the ARIMA model and disruptive formula in industry 4.0: A case study in Jakarta, Indonesia |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20201124 Address after: 318015 no.2-3167, zone a, Nonggang City, no.2388, Donghuan Avenue, Hongjia street, Jiaojiang District, Taizhou City, Zhejiang Province Applicant after: Taizhou Jiji Intellectual Property Operation Co.,Ltd. Address before: 201616 Shanghai city Songjiang District Sixian Road No. 3666 Applicant before: Phicomm (Shanghai) Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210726 Address after: 518000 4th floor, Tsinghua Ziguang science and Technology Park, No. 13, Langshan Road, songpingshan community, Xili street, Nanshan District, Shenzhen, Guangdong Patentee after: Shenzhen Aibo Communication Co.,Ltd. Address before: 318015 no.2-3167, area a, nonggangcheng, 2388 Donghuan Avenue, Hongjia street, Jiaojiang District, Taizhou City, Zhejiang Province Patentee before: Taizhou Jiji Intellectual Property Operation Co.,Ltd. |