CN106921507A - The method and apparatus being predicted to customer complaint within a wireless communication network - Google Patents
The method and apparatus being predicted to customer complaint within a wireless communication network Download PDFInfo
- Publication number
- CN106921507A CN106921507A CN201510994227.7A CN201510994227A CN106921507A CN 106921507 A CN106921507 A CN 106921507A CN 201510994227 A CN201510994227 A CN 201510994227A CN 106921507 A CN106921507 A CN 106921507A
- Authority
- CN
- China
- Prior art keywords
- data
- history
- alarm
- complaint
- time
- 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
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
According to the invention it is proposed that a kind of method being predicted to customer complaint within a wireless communication network, including：History Performance Data, history alarm data and the history that each network element for cordless communication network is extracted from historical data base complain data；Based on set pre-warning time length, input time window size and predicted time window size, in units of convergence grid, to convergence grid in the History Performance Data of each network element, history alarm data complain data to be combined and associate with history；History Performance Data, history alarm data and history after combining and associating complain the set of data samples of data to train and generate complaint forecast model as training dataset；And the real-time performance data and the set of data samples of real-time alarm data in time window to be predicted are input to the complaint forecast model, produce complaint to predict the outcome.
Description
Technical field
Customer complaint is carried out within a wireless communication network the present invention relates to a kind of
The method and apparatus of prediction, can be according to convergence grid, using history
Performance data, history alarm data and history complain data to complaining prediction mould
Type is trained, so as to obtain complaint prediction using the complaint forecast model
As a result.
Background technology
During the operation of cordless communication network, equipment event often occurs
Barrier.Now, the cellphone subscriber in a certain position cannot be led to base station
Letter or it is not smooth with base station communication, and be embodied as user mobile phone without
Signal, phone cannot be put through, the phenomenon such as call drop.Generally, wireless communication networks
The operator of network provides complaint service for cellphone subscriber.
By the complaint from cellphone subscriber, Virtual network operator it can be found that net
Produced problem in network, and solve.Currently, Virtual network operator is usual
The complaint of user is received by phone call for appeal, and the complaint of user is entered into
In system in case backstage net dimension network optimization personnel further processed.
It sometimes appear that lasting or more serious failure problems, lead in network
It is burst complaint to cause a large amount of complaint of the generation in short time period.It is right that burst is complained
It is very harmful for Consumer's Experience.But, received with current complaint
Cannot be effectively treated for this kind of burst complaint with processing mode.Therefore,
The new mechanism that needs are predicted to the complaint that such as happens suddenly, so that root
It is predicted that result does sth. in advance to find and process the network that burst may be caused to complain
Side problem, the generation for preventing burst from complaining, so as to improve the experience of user.
Chinese patent discloses CN103188705A and discloses a kind of customer complaint
" inquiry " system.System is primarily based on the calling information of history and signaling is lost
Information etc. is lost, service fail information bank is set up.Then in real time execution
Each complaint, system will run into network based on information library inquiry user
Failed signaling information during problem；Then according to these calling informations and mistake
The combination for losing signaling goes to be compared with some threshold values to produce warning.Based on these
Warning information, can improve great batch complaint handling promptness rate, reduce visitor
Clothes and network maintenance staff artificial judgment link, reduce information distortion and
The time of circulation.But it is substantially still the place for following " detecting " afterwards
Reason pattern, can not reach the effect of " anticipating ".
Chinese patent is disclosed CN103428741A and is abided by using the portfolio of base station
Follow specific probability distribution it is assumed that and being found using threshold value " bad
Base station " and " badly perceiving user ", and calculate " influence coefficient "
The generation for indicating ensuing operating procedure to prevent burst from complaining.The method
Judgment rule is relatively simple, therefore accuracy turns into larger problem.
There is the substantial amounts of real time information collected, example within a wireless communication network
Such as network exception event information, network performance information, these information and user
Complaint has more close relation.But, in the prior art simultaneously
Associating between the generation to performance data, alarm data with complaint is not carried out
Research, therefore, it is difficult to be effectively predicted to burst complaint etc..
The content of the invention
In order to the drawbacks described above for overcoming prior art proposes the present invention.Therefore,
An object of the present invention be to provide it is a kind of within a wireless communication network to
The method and apparatus that are predicted are complained at family, can according to convergence grid,
Data are complained to throwing using History Performance Data, history alarm data and history
Tell that forecast model is trained, so as to be obtained using the complaint forecast model
Complaint predicts the outcome.
To achieve these goals, according to the invention it is proposed that one kind is in nothing
The method being predicted to customer complaint in line communication network, including：From going through
The history of each network element for cordless communication network is extracted in history database
Can data, history alarm data and history complaint data；Based on set
Pre-warning time length, input time window size and predicted time window size, with
Convergence grid is unit, to convergence grid in each network element go through
History performance data, history alarm data and history complain data to be combined simultaneously
Association；History Performance Data, history alarm data after combining and associating
The set of data samples for complaining data with history is trained as training dataset
And generate complaint forecast model；And by the real-time in time window to be predicted
The set of data samples of energy data and real-time alarm data is input to the complaint
Forecast model, produces complaint to predict the outcome.
Preferably, it is described to combine and the History Performance Data after associating, go through
History alarm data and history complain the set of data samples of data as training number
The step of according to collection, also includes：Data are complained in data sample set pair burst
It is identified.
Preferably, it is described to combine and the History Performance Data after associating, go through
History alarm data and history complain the set of data samples of data as training number
The step of according to collection, also includes：Invalid data sample is cleaned from the training dataset
This.
Preferably, the convergence grid is by close many of physical location
What individual network element was constituted.
Preferably, network element included in the convergence grid be can be more
New.
Preferably, the History Performance Data has different performance categories,
And the history alarm data have different alarm species.
In addition, according to the present invention, it is also proposed that one kind is in cordless communication network
In the device that is predicted to customer complaint, including：From historical data base
Extract History Performance Data, the history of each network element for cordless communication network
Alarm data and history complain the data import modul of data；Based on set
Pre-warning time length, input time window size and predicted time window size,
In units of convergence grid, to convergence grid in each network element
History Performance Data, history alarm data and history complain data to be combined
And the data aggregation module for associating；Historical performance number after combining and associating
The set of data samples of data is complained as instruction according to, history alarm data and history
Practice data set to train and generate the training module for complaining forecast model；With
And by the real-time performance data and real-time alarm data in time window to be predicted
Set of data samples be input to the complaint forecast model, produce and complain prediction
The burst prediction module of result.
Preferably, the data aggregation module is prominent in the data sample set pair
Hair complains data to be identified.
Preferably, described device also includes：
The data cleansing mould of invalid data sample is cleaned from the training dataset
Block.
In accordance with the invention it is possible to according to convergence grid, using history
Energy data, history alarm data and history complain data to complaining forecast model
It is trained, so as to obtain complaint prediction knot using the complaint forecast model
Really.
Brief description of the drawings
Fig. 1 shows the side that forecast model is complained in generation of the invention
The flow chart of method.
Fig. 2 shows of the invention according to the real-time performance number for obtaining
Data are complained according to, real-time alarm data and in real time, online to convergence
The flow chart of the process being updated.
Fig. 3 is of the invention pre- using complaining when prediction is complained in triggering
Survey the flow chart that model complains the process for predicting the outcome to generate.
Fig. 4 is showed and for base station (network element) to be divided into corresponding convergence
The flow chart of the process of grid.
Fig. 5 show for will occur customer complaint position at base station with
The example table that affiliated convergence grid is mapped.
Fig. 6 is showed according to convergence grid, based on time window to alarm
Data, performance data be combined after set of data samples example.
Fig. 7 shows the schematic diagram of the form of alarm information.
Fig. 8 shows the schematic diagram of the form of capability message.
Fig. 9 shows the customer complaint message that the network control center is received
Form schematic diagram.
Figure 10 shows throwing of each convergence grid in time windows
Tell several example sample graphs.
Figure 11 is the schematic configuration diagram of complaint forecasting system of the invention.
Specific embodiment
The preferred embodiments of the present invention are described below with reference to the accompanying drawings.In accompanying drawing
In, identical element will be represented by identical reference symbol or numeral.Additionally,
In description below of the invention, the tool to known function and configuration will be omitted
Body is described, to avoid making subject of the present invention unclear.
Fig. 1 shows the side that forecast model is complained in generation of the invention
The flow chart of method.
As shown in figure 1, in step 101, extracting wireless from historical data base
The History Performance Data of each network element of communication network, history alarm data and go through
History complains data.The network element of cordless communication network includes base station, switched wireless
Center (MSC) etc..So-called performance data include be characterize communication network in
Different network elements at runnability data, such as the telephone traffic at network element,
Resources occupation rate etc..So-called alarm data refers to that generation is all as before at network element
Alarm information reported to the network control center produced during barrier etc..
So-called complaint data are to extract existing for channel radio from Network Management Equipment
Believe complaint message of failure etc..
According to the present invention, each network element of communication network is divided into different numbers
According to converge grid, in units of the convergence grid by convergence
The all types of convergences of all network elements in grid are together.As an example,
Convergence grid, example can be divided according to the physical relationship of network element
Such as, the close multiple network elements of physical location can be divided into same data to converge
Poly- grid.Specific dividing mode is not limited thereto.
Division for convergence grid should be variable, such as at certain
New network element etc. is laid in a little places.Therefore, by convergence grid
Before the data of all network elements are converged, it is thus necessary to determine that whether there occurs number
According to the change (step 103) of the division for converging grid.If it is determined that there occurs
The change (step 103 be) of the division of convergence grid, then in step
In 121 base station resource data, such as base station name are extracted from corresponding database
Title and base station geographic position etc..Then, in step 123, according to what is extracted
Base station resource data, update the data the division for converging grid.
If the division of convergence grid is without updating, in step 105,
According to convergence grid to extracting History Performance Data, the history of each network element
Together with alarm data complains convergence with history.
In step 107, pre-warning time length is set, input time window size and
The parameters such as predicted time window size.The pre-warning time length is predicted time
Difference between the starting point of window and the end point of input time window, that is, predict
After making how long, the content predicted just occurs.During the input
Between window refer in order to be predicted, it is necessary to collect prediction occur time point it is many forward
In time range long, all types of data that network is produced.During the prediction
Between window be used to refer to show in following time range how long, if can occur
Burst is complained.
In step 109, based on set pre-warning time length, input time
The parameter such as window size and predicted time window size, according to convergence grid,
Data are complained to History Performance Data, history alarm data and history carries out group
Merge association, thus can generate the number of mapping form as shown in Figure 6
According to sample set.
In step 111, History Performance Data, history alarm number from after combination
Carried out according to data are complained in the mapping form with history complaint data to burst
Mark, to provide learning objective (prediction for follow-up machine-learning process
Burst complain) foundation, so as to form training dataset.
Complaining data to be identified to burst can be thrown by comparing in time window
The mode of total number and threshold size is told to realize.For example, data will be complained
The number of mapping form as shown in Figure 10 is generated after being combined for each time window
Passing through according to the complaint total number in each time window of sample set, i.e. Statistics Division will
The numerical value of last row of the mapping form shown in Figure 10 is compared with threshold value
Relatively come judge its whether be burst complain data.Then, would indicate that complaint number
According to whether be burst complain data mark according to time window and mesh fitting
Mode, with History Performance Data and the set of data samples of history alarm data
Joint composing training data set.
Complaining data to be identified burst can also be by the side of time-domain filtering
Formula is realized, for example, the time series of data will be complained to pass through certain wave filter
Filtered result be compared with threshold value set in advance.In the time
On the premise of window is longer than filter length, as long as occurring one in certain time window
The secondary filter result higher than threshold value, then its corresponding complaint data is to be determined
For data are complained in burst.Then, would indicate that whether complaint data are that burst is thrown
The mark of data is told according to time window and the mode of mesh fitting, with history
The set of data samples joint composing training number of energy data and history alarm data
According to collection.
Then, in step 113, to by History Performance Data, history alarm number
According to the training dataset constituted with history complaint data, invalid data sample is cleaned
This.Invalid data sample is to represent in sample certain or some feature values
Invalid or missing situation.For example in figure 6, if " the performance kind of grid 1
Class P value _ 1 hour " numerical value fails to get for NULL, i.e. this numerical value, then
This numerical value is invalid, causes this sample equally invalid, it should to be cleaned.Or
Concentrating the time range for including when historical data can not provide certain sample institute
During the All Eigenvalues for needing, such as certain sample needs 23 points of December 31 day
Only comprising the institute so far of January 1 in performance and alarm data, but historical data
There are performance and alarm data, then the sample occurs the missing of character numerical value,
It is invalid to cause to be judged as.Need to provide input time window in itself due to model
Size, therefore, when the data at a time point be judged as it is missing or invalid
When, the sample comprising the time point can be determined in each input time window
Directly deleted for invalid, that is, complete data cleansing function.
In step 115, using the historical performance number cleaned after invalid data sample
The training data set pair for complaining data according to, history alarm data and history is complained
Forecast model is trained and generates.
In step 117, the complaint forecast model to generating is preserved.
In step 119, the pre-warning time length set in step 107 is preserved,
The parameter such as input time window size and predicted time window size.
Fig. 2 shows of the invention according to the real-time performance number for obtaining
Data are complained according to, real-time alarm data and in real time, online to convergence
The flow chart of the process being updated.
As shown in Fig. 2 in step 201, nothing is obtained from data collection module
The real-time performance data of each network element of line communication network, real-time alarm data and
Data are complained in real time.
In step 203, real-time performance data, the real-time alarm data that will be obtained
With complaint data in real time, converged according to convergence grid.
In step 205, based on set pre-warning time length, input time
The parameter such as window size and predicted time window size, more new system is directed to each data
Converge the data sample of grid.
Finally, in step 207, according to the data sample being updated come the more new calendar
History database.
Fig. 3 is of the invention pre- using complaining when prediction is complained in triggering
Survey the flow chart that model complains the process for predicting the outcome to generate.
In step 301, complain prediction to be triggered, be set as cyclic forecast
When, complaint prediction can be by upper predetermined period end trigger；In setting
When being predicted for event-triggered, prediction is complained to be touched by a class particular event
Hair, for example, newly receive one group of complaint data or performance data etc..
In step 303, updated in form from the real time data shown in Fig. 6 and obtained
The performance data of each convergence grid in predicted time section and alarm
The set of data samples of data.
In step 305, to performance data and the set of data samples of alarm data,
Cleaning invalid data sample.Invalid data sample is to represent certain in real-time sample
Individual or some feature values it is invalid or missing situation.For example in figure 6,
If " performance categories P value _ 1 hour " numerical value of grid 1 be NULL, i.e., this
Numerical value fails to get, then this numerical value is invalid, causes this sample equally invalid,
Should be deleted.
In step 307, by performance data and the data of alarm data after cleaning
Forecast model is complained in sample set input, so as to produce complaint to predict the outcome.
In step 309, produced predicting the outcome is entered from output communication module
Row output.
In step 311, predicted the outcome to historical data using resulting complaint
Storehouse is updated.
In step 313, resulting complaint is predicted the outcome and is input to statistics mould
Block, can obtain complaining the statistics for predicting the outcome.
Fig. 4 is showed and is divided into accordingly base station (network element) using the method for exhaustion
Convergence grid process flow chart.
In step 401, the size K of convergence grid, i.e. each net are set
Comprising base station nearest each other on K position in lattice.
In step 403, from base station set { 1,2,3 ... NBSOne base of middle selection
Stand i.
In step 405 and step 407, initialize grid search angle a and
Apart from d.
In step 409, according to existing angle a and apart from d values, find a little
(BS0+ dcos (a), BS1+ dsin (a)), wherein BS0And BS1Respectively base station i
The coordinate value that is converted into of latitude and longitude.Then based on the point, look for
To the K base station away from its nearest neighbours, and the grid mark that they are formed is n.
Subsequently into step 411.
In step 411, whether judge in the grid comprising base station i itself.
If comprising, illustrate to continue to increase detection range along this direction, still having can
Other grids comprising i can be obtained, into step 413；If do not wrapped
Contain, then will not again produce other grids comprising i in this direction, enter
Enter step 423.
In step 413, on the premise of base station i itself is included in grid,
To each base station in grid n, the list of the grid in query graph 5 belonging to it.
If not including n, n is added.
In step 415, d is increased on d parameter basis existingstep, weight
Multiple step 409.
In step 423, the direction of more new search, in existing angle a ginsengs
Increase a on the basis of numberstep。
In step 421, judge the direction after updating whether more than 360 degree.
If it is, show the base station that direction search procedure has terminated, and should more renew,
Scanned for for other points, into step 419.Otherwise, repeat step
409。
In step 419, next base station i+1 of current base station is selected, and
Into step 417.
In step 417, judge whether that all base stations all have stepped through to finish.
If it is, terminating whole flow process.If it is not, then repeat step 405.
Fig. 5 show for will occur customer complaint position at base station with
The example table that affiliated convergence grid is mapped.
As shown in figure 5, three columns in the left side of form respectively illustrate generation user
Base station IDs, base station name at the position of complaint, base station physical position (longitude and latitude
Degree).One column of the rightmost side of the form of Fig. 5 shows the number belonging to the base station
According to the ID for converging grid.As shown in figure 5, same base station may belong to multiple
Different pieces of information converges grid.For example, the base station that base station IDs are 1 can be simultaneously
Belong to convergence grid 1 and convergence grid 3.
Fig. 6 is showed according to convergence grid, based on time window to alarm
Data, performance data be combined after set of data samples example.
When the data for being converged are performance data and alarm data, for example,
Can be according to field " the base station name included in performance data or alarm data
Claim ", the performance data or alarm data are mapped to the base station name
Base station corresponding to convergence grid, further, with reference to shown in Fig. 5
The base station complained and the mapping relations of convergence grid, Ke Yisheng
Into as shown in Figure 6 according to convergence grid, based on time window to accusing
Set of data samples after warning data, performance data and complaining data to be combined.
In the form of Fig. 6, as an example, alarm data is accused for each difference
The quantity of the alarm information of categories within police force class, performance data is each different performance kind
The performance value of class, and it is the complaint quantity in time window to complain data.
In the present invention, as an example, K to being included in convergence grid
The quantity of all alarm informations of a certain alarm species of individual base station is tired out
Plus to form the alarm information of the alarm species for convergence grid
Number.Certain performance categories of the K base station to being included in convergence grid
Performance value make average (if there is part in the K performance value of base station
Missing, i.e., for the performance categories, should be subject to correspond to K songs base station
K performance value, but the numerical value quantity for actually receiving is less than K, then not
Update corresponding field) form the performance for the convergence grid
The performance value of species.
For complaining data, can at once be associated complaining due to lacking at present
The method of user's own base station when complaining problem to occur, therefore can be with root
According to complaining the detailed place for occurring to describe, such as " so-and-so street number is attached
Closely " etc., the place is converted by generalized information system (GIS-Geographic Information System)
It is the longitude and latitude of corresponding base station, is then based on the form shown in above-mentioned Fig. 5,
It is mapped to the convergence grid belonging to the base station.Further, according to Fig. 5
Involved complaint related data, can form data as shown in Figure 6 and converge
The complaint quantity of poly- grid.
" T hours " described in Fig. 6 refers to (prediction object time window
Beginning time point-pre-warning time length-T) to (during the window starting of prediction object time
Point-pre-warning time length-T+1) time window in the range of, all kinds of alarms disappear
The quantity of breath or all kinds of performance values.Here, pre-warning time length is model
The parameter selected during training.It is (prediction object time window that history complains quantity
Initial time point-pre-warning time length-W) to (prediction object time window is initial
Time point-pre-warning time length) time window in the convergence that is counted
The complaint sum that grid occurs.Here, W is input time window size, together
Selected parameter when sample is model training.
In the form of Fig. 6, unassignable field after initialization is deposited, with
NULL is represented.For example 12：13 can not possibly be obtained when 00：The property of 00 collection
Energy data, so being represented with NULL.
Fig. 7 shows the schematic diagram of the form of alarm information.
As shown in fig. 7, alarm information includes：The base station name that alerts,
Alarm time of origin and alarm type.Each network element is usual in modern network system
All have log recording function, the unusual condition that will occur in running and
Timestamp is recorded, and is sent to NMS, i.e. network management system
In common warning information.Common alarm type such as base station CPRI is alerted,
Base station flash alarm etc., is type specified in NMS
Fig. 8 shows the schematic diagram of the form of capability message.
As shown in figure 8, capability message includes：Send the base station of the capability message
Base station name, performance statistics time and for different performance species
Value.The wireless network of operation conditions in to(for) part of eating dishes without rice or wine passes through periodically
The mode of daily record is recorded, i.e., often putting will collect in the cycle for a period of time
Wireless network various operations parameter index and acquisition time record
Come, send to NMS.The kind of common 2G network performance indexes
Class includes, TCH voice channel telephone traffics, the equivalent telephone traffic of EDGE data,
Voice channel congestion ratio etc., is standard specified in NMS
Pointer type.
Fig. 9 shows the customer complaint message that the network control center is received
Form schematic diagram.
As shown in figure 9, customer complaint message includes：User complained when
Between, the concern category of the place that user is complained and customer complaint.
Common concern category such as mobile phone has signal but cannot use, city dweller
Community's indoor network coverage etc., is specified in complaint handling system and complains
Type.
Figure 10 shows throwing of each convergence grid in time windows
Tell several example sample graphs.
As shown in Figure 10, difference can be counted according to convergence grid
Time window in complaint sum, and be indicated in a tabular form.
Figure 11 is the schematic configuration diagram of complaint forecasting system of the invention.
As shown in figure 11, complaint forecasting system 10 of the invention includes：
Output communication module 1101, historical data base 1103, data collection module
1105th, controller 1107, burst prediction module 1109, training module 1111,
Statistical module 1113, data aggregation module 1115, data cleansing module 1117,
And data import modul 1119.
Data collection module 1105 collects the reality of each network element of cordless communication network
When performance data, real-time alarm data and complain data in real time.
Data import modul 1119 is by collected by data collection module 1105
Real-time performance data, the real-time alarm data of each network element of cordless communication network
Historical data base 1203 is imported into complaint data in real time.
Data aggregation module 1115 in units of the convergence grid, by number
According to the performance data, alarm data and complaint number that converge each network element in grid
According to being gathered together.
Data cleansing module 1117 is used in historical data or real time data group
Into sample in invalid sample remove.
Training module 1111 is utilized performance data, alarm data and complains number
Carried out according to the training data set pair complaint forecast model for being gathered together composition
Training.
Burst prediction module 1109 is utilized when prediction is complained and complains forecast model
Predicted the outcome to obtain complaint.
Statistical module 1113 pairs is complained to predict the outcome and counted, and generation is complained
The statistical value for predicting the outcome.
Complaint is predicted the outcome or complains prediction to tie by output communication module 1101
Outside is arrived in the statistical value output of fruit.
Controller 1107 is used to control the operation of above-mentioned each module of system
System.
Some specific embodiments are enumerated above to elaborate the present invention, this
A few examples are merely to illustrate principle of the invention and its implementation, rather than right
Limitation of the invention, without departing from the spirit and scope of the present invention,
Those skilled in the art can also make various modifications and improvement.Therefore,
The present invention should not be limited by above-described embodiment, and should be by appended claims
And its equivalent is limited.
Claims (9)
1. a kind of side being predicted to customer complaint within a wireless communication network
Method, including：
Extraction is for each network element of cordless communication network from historical data base
History Performance Data, history alarm data and history complain data；
Based on set pre-warning time length, input time window size and pre-
Time window size is surveyed, in units of convergence grid, to convergence net
The History Performance Data of each network element in lattice, history alarm data and history are thrown
Tell that data are combined and associate；
History Performance Data, history alarm data after combining and associating and
History complains the set of data samples of data to be trained simultaneously as training dataset
Forecast model is complained in generation；And
By real-time performance data and real-time alarm data in time window to be predicted
Set of data samples be input to the complaint forecast model, produce and complain prediction
As a result.
2. method according to claim 1, it is characterised in that：
It is described to combine and History Performance Data, history alarm number after associating
The step of according to the set of data samples of data is complained as training dataset with history
Also include：
Data are complained to be identified in data sample set pair burst.
3. method according to claim 1, it is characterised in that：
It is described to combine and History Performance Data, history alarm number after associating
The step of according to the set of data samples of data is complained as training dataset with history
Also include：
Invalid data sample is cleaned from the training dataset.
4. method according to claim 1, it is characterised in that：
The convergence grid is by the close multiple network element structures of physical location
Into.
5. method according to claim 4, it is characterised in that：
Included network element is renewable in the convergence grid.
6. method according to claim 1, it is characterised in that：
The History Performance Data has different performance categories, and described goes through
History alarm data has different alarm species.
7. a kind of dress being predicted to customer complaint within a wireless communication network
Put, including：
Extraction is for each network element of cordless communication network from historical data base
History Performance Data, history alarm data and history complain the data of data to lead
Enter module；
Based on set pre-warning time length, input time window size and pre-
Time window size is surveyed, in units of convergence grid, to convergence net
The History Performance Data of each network element in lattice, history alarm data and history are thrown
Tell the data aggregation module that data are combined and associate；
History Performance Data, history alarm data after combining and associating and
History complains the set of data samples of data to be trained simultaneously as training dataset
The training module of forecast model is complained in generation；And
By real-time performance data and real-time alarm data in time window to be predicted
Set of data samples be input to the complaint forecast model, produce and complain prediction
The burst prediction module of result.
8. device according to claim 7, it is characterised in that：
The data aggregation module complains number in data sample set pair burst
According to being identified.
9. device according to claim 7, it is characterised in that also include：
The data cleansing mould of invalid data sample is cleaned from the training dataset
Block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510994227.7A CN106921507B (en) | 2015-12-25 | 2015-12-25 | Method and apparatus for predicting user complaints in a wireless communication network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510994227.7A CN106921507B (en) | 2015-12-25 | 2015-12-25 | Method and apparatus for predicting user complaints in a wireless communication network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106921507A true CN106921507A (en) | 2017-07-04 |
| CN106921507B CN106921507B (en) | 2020-10-09 |
Family
ID=59454739
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510994227.7A Active CN106921507B (en) | 2015-12-25 | 2015-12-25 | Method and apparatus for predicting user complaints in a wireless communication network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106921507B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019062710A1 (en) * | 2017-09-30 | 2019-04-04 | 华为技术有限公司 | Communication method and communication device thereof |
| CN109963292A (en) * | 2017-12-25 | 2019-07-02 | 中国移动通信集团广东有限公司 | Complain method, apparatus, electronic equipment and the storage medium of prediction |
| CN110399995A (en) * | 2018-04-20 | 2019-11-01 | 顺丰科技有限公司 | Waybill complaint handling method, apparatus, equipment and its storage medium |
| CN110602652A (en) * | 2019-10-15 | 2019-12-20 | 中移信息技术有限公司 | Complaint model training method, and user complaint prediction method, device and equipment |
| CN110891283A (en) * | 2019-11-22 | 2020-03-17 | 超讯通信股份有限公司 | Small base station monitoring device and method based on edge calculation model |
| CN111210057A (en) * | 2019-12-25 | 2020-05-29 | 广东飞企互联科技股份有限公司 | Method for predicting complaints of mobile phone internet users |
| CN111292094A (en) * | 2018-12-07 | 2020-06-16 | 中国移动通信集团重庆有限公司 | Complaint early warning method, device, equipment and medium based on big data technology |
| CN111669281A (en) * | 2019-03-08 | 2020-09-15 | 华为技术有限公司 | Alarm analysis method, device, equipment and storage medium |
| CN113495909A (en) * | 2020-04-06 | 2021-10-12 | 中信百信银行股份有限公司 | Customer complaint single quantity early warning method and device, electronic equipment and storage medium |
| CN113822454A (en) * | 2020-06-18 | 2021-12-21 | 中国移动通信集团浙江有限公司 | Prediction method and device for slice complaint processing resources |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101110699A (en) * | 2007-08-07 | 2008-01-23 | 广州诚予国际市场信息研究有限公司 | System with network satisfaction degree estimation and early warning function and implementing method thereof |
| US20130198767A1 (en) * | 2012-01-30 | 2013-08-01 | Board Of Regents, The University Of Texas System | Method and apparatus for managing quality of service |
| CN105095588A (en) * | 2015-08-05 | 2015-11-25 | 中国联合网络通信集团有限公司 | Mobile internet user complaint prediction method and device |
-
2015
- 2015-12-25 CN CN201510994227.7A patent/CN106921507B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101110699A (en) * | 2007-08-07 | 2008-01-23 | 广州诚予国际市场信息研究有限公司 | System with network satisfaction degree estimation and early warning function and implementing method thereof |
| US20130198767A1 (en) * | 2012-01-30 | 2013-08-01 | Board Of Regents, The University Of Texas System | Method and apparatus for managing quality of service |
| CN105095588A (en) * | 2015-08-05 | 2015-11-25 | 中国联合网络通信集团有限公司 | Mobile internet user complaint prediction method and device |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019062710A1 (en) * | 2017-09-30 | 2019-04-04 | 华为技术有限公司 | Communication method and communication device thereof |
| US11451286B2 (en) | 2017-09-30 | 2022-09-20 | Huawei Technologies Co., Ltd. | Communication method and communications device thereof |
| CN109963292A (en) * | 2017-12-25 | 2019-07-02 | 中国移动通信集团广东有限公司 | Complain method, apparatus, electronic equipment and the storage medium of prediction |
| CN109963292B (en) * | 2017-12-25 | 2022-04-01 | 中国移动通信集团广东有限公司 | Complaint prediction method, complaint prediction device, electronic apparatus, and storage medium |
| CN110399995A (en) * | 2018-04-20 | 2019-11-01 | 顺丰科技有限公司 | Waybill complaint handling method, apparatus, equipment and its storage medium |
| CN111292094A (en) * | 2018-12-07 | 2020-06-16 | 中国移动通信集团重庆有限公司 | Complaint early warning method, device, equipment and medium based on big data technology |
| CN111669281B (en) * | 2019-03-08 | 2021-11-30 | 华为技术有限公司 | Alarm analysis method, device, equipment and storage medium |
| CN111669281A (en) * | 2019-03-08 | 2020-09-15 | 华为技术有限公司 | Alarm analysis method, device, equipment and storage medium |
| CN110602652A (en) * | 2019-10-15 | 2019-12-20 | 中移信息技术有限公司 | Complaint model training method, and user complaint prediction method, device and equipment |
| CN110602652B (en) * | 2019-10-15 | 2021-07-27 | 中移信息技术有限公司 | Complaint model training method, and user complaint prediction method, device and equipment |
| CN110891283A (en) * | 2019-11-22 | 2020-03-17 | 超讯通信股份有限公司 | Small base station monitoring device and method based on edge calculation model |
| CN111210057A (en) * | 2019-12-25 | 2020-05-29 | 广东飞企互联科技股份有限公司 | Method for predicting complaints of mobile phone internet users |
| CN113495909A (en) * | 2020-04-06 | 2021-10-12 | 中信百信银行股份有限公司 | Customer complaint single quantity early warning method and device, electronic equipment and storage medium |
| CN113822454A (en) * | 2020-06-18 | 2021-12-21 | 中国移动通信集团浙江有限公司 | Prediction method and device for slice complaint processing resources |
| CN113822454B (en) * | 2020-06-18 | 2023-08-18 | 中国移动通信集团浙江有限公司 | Method and device for predicting slice complaint processing resources |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106921507B (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106921507A (en) | The method and apparatus being predicted to customer complaint within a wireless communication network | |
| CN107547633B (en) | User constant standing point processing method and device and storage medium | |
| US12063528B2 (en) | Anomaly detection method and device, terminal and storage medium | |
| CN102097004B (en) | Mobile phone positioning data-based traveling origin-destination (OD) matrix acquisition method | |
| CN107466103A (en) | A kind of method of locating terminal and the network equipment | |
| CN103619036B (en) | A kind of method and apparatus with assessing for data acquisition | |
| CN105764059B (en) | A method of pseudo-base station is identified based on terminal positional information | |
| CN108156626A (en) | Rail traffic quality of wireless network appraisal procedure, device and medium | |
| CN108632746A (en) | A kind of method of determining region flow of the people | |
| CN104349366B (en) | A kind of radio network problems association analysis method based on GIS | |
| CN107548082A (en) | The method, apparatus and system of one germplasm difference regional analysis | |
| CN107580337A (en) | Hot spot area identification method and device | |
| CN106060920B (en) | Hot spot region determines method and system | |
| CN103763149B (en) | Real-time statistical method for network user number | |
| CN105873113A (en) | Method and system for positioning wireless quality problem | |
| CN103108343B (en) | Set up method and device, network performance optimizing method and the device of decision tree | |
| CN104202719B (en) | Number measure and crowd situation monitoring method and system based on position credibility | |
| CN108133001A (en) | MR indoor and outdoors separation method, device and medium | |
| CN103686818A (en) | Simulation test method and equipment | |
| CN106504524A (en) | A kind of method that express highway section is divided based on mobile signaling protocol dynamic | |
| WO2017063532A1 (en) | Method and device for identifying address type | |
| CN106488467A (en) | Network capacity transfinites determination methods, device and network capacity management method, system | |
| CN106780248A (en) | City internet of things equipment operation control system and method based on big data technology | |
| CN112601247B (en) | Method, device and system for monitoring service quality of base station cell | |
| CN113902583B (en) | Distribution network side operation and maintenance method and system using low-voltage network equipment data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |