Content of the invention
It is an object of the present invention to overcoming deficiency of the prior art, a kind of business trend forecasting method being provided and is
System, can provide business trend prediction for operator automatically based on history service data, help operator to understand that business is sent out in time
Exhibition trend.
For achieving the above object, the technical solution used in the present invention is: a kind of business trend forecasting method, comprising:
Obtain the statistical history business datum sequence of prediction term in units of predetermined period from business centring system;
Carry out periodization to described statistical history business datum sequence according to the service period of described prediction term to process, obtain
Statistical history business datum sequence after must processing;
According to the statistical history business datum sequence after processing, the method using time series forecasting exists to described prediction term
Business trend in selected predetermined period is predicted, and obtains tentative prediction data;
Described tentative prediction data is carried out go periodization to process corresponding reduction cycleization process with described, obtain choosing
Determine the prediction data of predetermined period.
Preferably, described statistical history business datum sequence is carried out with periodization and processes inclusion: adopt difference method
Periodization is gone to process described in described statistical history business datum sequence is carried out;
Described tentative prediction data is carried out with reduction cycleization process include:
The statistical history business datum of upper first-class sequence predetermined period is obtained from described statistical history business datum sequence, its
In, described upper first-class sequence predetermined period be in a upper service period sequentially in time with selected predetermined period be located at identical suitable
Predetermined period that tagmeme is put, a described upper service period is to select the service period that predetermined period is located;
Calculate described tentative prediction data and upper first-class sequence predetermined period statistical history business datum and as described
Prediction data.
Preferably, using difference method carry out described in go periodization process method include: if described service period
Comprise predetermined period of varying number, then choose two neighboring predetermined period of same sequence position from each little service period
Statistical history business datum carries out linear interpolation, to supply all little service period for big service period, is formed after supplying
Statistical history business datum sequence, to go at periodization described in the statistical history business datum sequence after described supplying is carried out
Reason, wherein, described big service period is the most service period of the quantity comprising predetermined period, and little service period is to comprise to predict
The quantity in cycle is less than the service period of big service period.
Preferably, described business trend forecasting method also includes: obtains the original of prediction term from business centring system
History service data sequence, obtains described prediction by described original history service data sequence is carried out with FFT
The service period of item;
The method obtaining described statistical history business datum sequence from business centring system is:
Described original history service data sequence is counted in units of predetermined period with generation in each predetermined period
Statistical history business datum, to form described statistical history business datum sequence.
Preferably, described business trend forecasting method also includes:
Obtaining described prediction term after the prediction data of selected predetermined period, judging whether the state of described prediction term goes out
Now abnormal;If it is, reporting to the police to user.
Preferably, whether the described state judging described prediction term abnormal inclusion:
Obtain the current statistic history service of described prediction term in units of predetermined period from described service center system
Data, described current statistic history service data is the business datum that described prediction term actually occurs in selected predetermined period;
Calculate the difference of the current statistic history service data selecting predetermined period and the prediction data of selected predetermined period;
If described difference exceedes default warning value it is determined that the state of described prediction term occurs extremely.
Another object of the present invention is to provide and a kind of operator can be helped timely to understand business development trend
Business trend predicting system.
The technical solution used in the present invention is: a kind of business trend predicting system, comprising:
Data acquisition module, for obtaining the statistical history of prediction term from business centring system in units of predetermined period
Business datum sequence;
Go periodization module, for entering according to the service period of described prediction term to described statistical history business datum sequence
Row goes periodization to process, the statistical history business datum after acquisition process;
Prediction module, for according to process after statistical history business datum, using time series forecasting method to institute
State business trend in selected predetermined period for the prediction term to be predicted, obtain tentative prediction data;
Data restoring module, for carrying out to described tentative prediction data going periodization to process corresponding reduction with described
Periodization is processed, and obtains the prediction data selecting predetermined period.
Preferably, described periodization module is gone to include:
Difference unit, for using difference method to described statistical history business datum sequence according to described prediction term industry
The business cycle goes periodization to process described in carrying out;
Described data restoring module includes:
Data capture unit, for obtaining the statistical history industry of upper first-class sequence predetermined period from described data acquisition module
Business data, wherein, described upper first-class sequence predetermined period be in a upper service period sequentially in time with selected predetermined period
Positioned at predetermined period of same sequence position, a described upper service period is to select the service period that predetermined period is located;
Reduction unit, for calculating the statistical history business datum of described tentative prediction data and upper first-class sequence predetermined period
And as described prediction data output.
Preferably, described periodization module is gone also to include:
Interpolating unit, for comprise predetermined period of varying number in described service period in the case of, from each little industry
The statistical history business datum choosing two neighboring predetermined period of same sequence position in the business cycle carries out linear interpolation, by institute
There is little service period to supply as big service period, form the statistical history business datum sequence after supplying, for difference unit pair
Described supply after statistical history business datum sequence carry out described in go periodization process, wherein, described big service period be bag
The most service period of quantity containing predetermined period, little service period is that the quantity comprising predetermined period is less than big service period
Service period.
Preferably, described data acquisition module includes:
Initial data acquiring unit, for obtaining original history service data sequence from business centring system;
Statistic unit, for being counted in units of predetermined period to described original history service data sequence in each prediction
The statistical history business datum occurring in cycle, forms described statistical history business datum sequence;
Described business trend predicting system also includes:
Service period computing module, for the described original history service data sequence that initial data acquiring unit is provided
Carry out FFT, to obtain the service period of described prediction term.
Preferably, described business trend predicting system also includes:
Judge module, for obtaining described prediction term after the prediction data of selected predetermined period, judges described prediction
Whether the state of item exception;
Alarm module, after judging that in judge module exception in the state of prediction term, reports to the police to user.
Preferably, described judge module includes:
Current data acquiring unit, for obtaining the current statistic history industry of described prediction term from described data acquisition module
Business data, described current statistic history service data is the business number that described prediction term actually occurs in selected predetermined period
According to;
Computing unit, for calculating the current statistic history service data of selected predetermined period and the pre- of selected predetermined period
Survey the difference of data;
Comparing unit, for being compared with default warning value to described difference, if described difference exceedes default warning
Value is it is determined that the state of described prediction term occurs extremely.
Business trend forecasting method and system that the present invention provides, pre- by carrying out to the development trend going up line service
Survey, and the early warning to ANOMALOUS VARIATIONS, help operator to understand the variation tendency of business in advance, find the abnormal conditions of business,
And then allow operator timely to find countermeasure, lift the service quality of oneself, improve the competitiveness of itself.
Specific embodiment
Embodiments of the invention are described below in detail, the example of described embodiment is shown in the drawings, wherein from start to finish
The element that same or similar label represents same or similar element or has same or like function.Below with reference to attached
The embodiment of figure description is exemplary, is only used for explaining the present invention, and is not construed as limiting the claims.
As shown in figure 1, business trend forecasting method provided in an embodiment of the present invention, comprise the following steps:
Step s1: obtain the statistical history business datum sequence of prediction term from business centring system in units of predetermined period
Row.
Illustrate, if predetermined period that user selectes is daily, that is, user wishes to predict that following a period of time is daily
Business datum, and the original history service data sequence of prediction term that service center system provides is also daily data, this
Sample statistics when be no need for being calculated, can directly using original history service data sequence as statistical history business datum sequence
Row, if predetermined period that user selectes is weekly, and the original history service data of the prediction term that service center system provides
Sequence is daily data, then need the data weekly according to the calculating of original history service data sequence, to be formed with one week be
The statistical history business datum sequence of unit;This step needs to all original history service when being predicted system first time
Data sequence is processed, and prediction after this only carries out process to untreated original history service data sequence and is
Can;Here, for telecommunications industry, this service center system is call center system, this prediction term is, for example, certain business
User's attention rate, service fulfillment quantity, business cancel one in quantity that quantity, business break down etc..
Step s2: according to the service period of described prediction term, periodization is carried out to described statistical history business datum sequence
Process, the statistical history business datum sequence after acquisition process.
Here, the data of business generally all assumes certain periodicity, the particularly business of telecommunications industry, the business of indication
Cycle is the periodicity that the original history service data of prediction term itself is presented;For different prediction term, business week
Phase is not quite similar, and directly cannot be carried out effectively predicting according to statistical history business datum sequence, accordingly, it would be desirable in prediction
The front periodicity removing statistical history business datum sequence.In addition, needing to all statistics when being predicted system first time
History service data carries out periodization and processes, and on this basis, prediction afterwards only needs to the new statistical history business obtaining
Data carries out periodization and processes.
Step s3: according to the statistical history business datum sequence after processing.
Using the e.g. time series forecasting of arma model method to described prediction term in selected predetermined period
Business trend is predicted, and obtains tentative prediction data;
Step s4: described tentative prediction data is carried out go periodization to process at corresponding reduction cycle with described
Reason, obtains the prediction data selecting predetermined period.
Here, the tentative prediction data due to obtaining in step s3 is according to the statistical history business after going periodization to process
Data sequence obtains, accordingly, it would be desirable to restore it as going the value before periodization process, just can obtain the reality of prediction term
Prediction data.
For above step s2, generally using difference method, statistical history business datum sequence can be carried out at periodization
Reason, corresponding carries out reduction cycle process as shown in Fig. 2 maing include: to tentative prediction data
Step s401: obtain the statistical history business number of upper first-class sequence predetermined period from statistical history business datum sequence
According to wherein, upper first-class sequence predetermined period is according to the time in a upper service period of selected predetermined period place service period
Order is located at predetermined period of same sequence position with selected predetermined period.
In order to make it easy to understand, providing following examples 1:
The prediction term that user selects is the quantity of opening of flow business, and predetermined period that user selects is daily, prediction term
Service period be one month, set today be November 15, if then user needs to know the flow business on November 16
Open quantity, then November 16 was selectes predetermined period, and upper first-class sequence predetermined period is October 16, if user needs
It is to be understood that the flow business on November 16,17 and 18 open quantity, upper first-class sequence predetermined period i.e. be respectively October 16,17
Day and 18 days.
Step s402: calculate the statistical history business datum and conduct of tentative prediction data and upper first-class sequence predetermined period
Prediction data.
During carry out periodization process using difference method, if service period comprises the prediction of varying number
Cycle, then generally supplied using the method for interpolation, to meet the condition of difference, particularly as follows: from each little service period
The statistical history business datum choosing two neighboring predetermined period of same sequence position carries out linear interpolation, by all little business
Cycle supplies as big service period, forms the statistical history business datum sequence after supplying, then to the statistical history industry after supplying
Business data sequence carries out n order difference, completes periodization and processes, wherein, n is the quantity of predetermined period that big service period comprises
(in this n round numbers), above big service period is the most service period of the quantity comprising predetermined period, and little service period is
The quantity comprising predetermined period is less than the service period of big service period.
In most cases, if service period comprises predetermined period of varying number, then big service period with little
Service period only differs from a predetermined period, if there is the situation of the little service period of difference difference predetermined period number, then can enter
The many wheels of row choose the process that same sequence position carries out linear interpolation, until supplying all little service period for big service period are
Only.
In order to make it easy to understand, still being illustrated using above example 1, service period is one month, then big business week
Phase just comprises 31 days (i.e. 31 predetermined period), and little service period then comprises 30 days (i.e. 30 predetermined period), then for example
Select to carry out linear interpolation to the statistical history business datum between 14 days and 15 days of little service period, that is, take two days average
Value is inserted to 14 and the statistical history business datum of 15 days between, forms the statistical history industry that each service period all comprises 31 days
The statistical history business datum sequence of business data, afterwards with the statistical history business datum on November 15 (because November comprises 30
My god, therefore, the actual statistical history business number becoming 16 in statistical history business datum sequence on the 15th after interpolation
According to) deduct the statistical history business datum on October 16, the like, periodization is completed to statistical history business datum sequence
Process.If November 14 carried out once predicting, then after the data before November 14 is all periodization
It is only necessary to deduct the statistical history business datum on October 16 with the statistical history business datum on November 15 just permissible,
So it is achieved that periodization.For 2 relatively more special months, the statistical history industry of 1 day or 2 days can be inserted in the middle of the month
Business data.
Service period for above prediction term is generally calculated when being predicted first time by system, if used
Family has learned that the service period of prediction term by other approach, then be no need for carrying out following calculating.For this reason, as Fig. 4 institute
Show, can obtain the service period of prediction term when being predicted first time by the following method:
Step s201: judge service period whether it is known that then executing above step s2 as known;
Step s202: the original history service data sequence of the prediction term obtaining in obtaining step s1;
Step s203: judge whether the original history service data sequence that step s202 obtains meets the industry calculating prediction term
The needs in business cycle?Then execution step s204 in this way, such as otherwise execution step s202, continue acquisition and have more original history industry
The original history service data sequence of business data;
Step s204: obtain the business of prediction term by original history service data sequence is carried out with FFT
In the cycle, execute above step s2 afterwards.
On the basis of above business trend forecasting method, also can carry out business trending early warning to prediction term, for this reason, the party
Method also includes:
Obtaining described prediction term after the prediction data of selected predetermined period, judging whether the state of described prediction term goes out
Now abnormal;If it is, reporting to the police to user.
As shown in figure 3, judging whether the state of prediction term abnormal method and can comprise the following steps that
Step s5: obtain the current statistic history service number of prediction term from business centring system in units of predetermined period
According to, current statistic history service data is the business datum that described prediction term actually occurs in selected predetermined period, here, by
Selecting predetermined period when carrying out early warning has been past tense, therefore, the statistical history business datum sequence of acquisition in step s1
Comprise above-mentioned current statistic history service data;
Step s6: calculate the prediction data of the current statistic history service data selecting predetermined period and selected predetermined period
Difference;
Step s7: judge whether described difference exceedes default warning value, then determine that the state of described prediction term occurs in this way
Abnormal, whether the state such as otherwise continuing monitoring and forecasting item exception.
The embodiment of the present invention additionally provides a kind of business trend prediction system that can realize above business trend forecasting method
System, as shown in figure 5, comprising:
Data acquisition module 1, the statistics for being obtained prediction term from business centring system in units of predetermined period is gone through
History business datum sequence;
Go periodization module 2, for described statistical history business datum sequence according to described prediction term service period
Carry out periodization to process, the statistical history business datum after acquisition process;
Prediction module 3, for according to process after statistical history business datum, using time series forecasting method to pre-
Survey business trend in selected predetermined period for the item to be predicted, obtain tentative prediction data;
Data restoring module 4, for carrying out to tentative prediction data processing corresponding reduction cycle with going periodization
Process, obtain the prediction data selecting predetermined period.
As shown in fig. 6, above data acquisition module 1 mays include:
Initial data acquiring unit 11, for obtaining original history service data sequence from business centring system;
Statistic unit 12, for being counted in units of predetermined period to original history service data sequence in each prediction week
The statistical history business datum occurring in phase, forms described statistical history business datum sequence.
As shown in fig. 6, may include difference unit 21 with the periodization module 2 that gets on, it adopts difference method to statistical history
Business datum sequence carries out periodization according to the service period of prediction term and processes.Accordingly, data above recovery module 4 can be wrapped
Include:
Data capture unit 41, first-class for (being specially in its statistic unit 12) from data acquisition module 1 in acquisition
The statistical history business datum of sequence predetermined period;
Reduction unit 42, for calculating tentative prediction data and the statistical history business datum of upper first-class sequence predetermined period
With as prediction data.
If the service period of prediction term comprises predetermined period of varying number, periodization is being carried out using difference method
During process, as shown in fig. 6, this goes periodization module also to include:
Interpolating unit 22 is for comprise predetermined period of varying number in described service period in the case of, little from each
The statistical history business datum choosing two neighboring predetermined period of same sequence position in service period carries out linear interpolation, will
All little service period are supplied as big service period, form the statistical history business datum sequence after supplying;
Above difference unit using difference method to the statistical history business datum sequence after described supplying according to described pre-
The service period of survey item goes periodization to process described in carrying out.
Voluntarily calculate the service period of prediction term if necessary to system when first time being predicted the prediction of item, then as schemed
Shown in 6, this goes periodization module also to include:
Service period computing unit 23, for the original history service data sequence that initial data acquiring unit 11 is provided
Carry out FFT, to obtain the service period of prediction term, if here, the current original history service data obtaining
Sequence is not enough to calculate service period, then continue acquisition from initial data acquiring unit 11 and comprise more original history service numbers
According to original history service data sequence.
As shown in fig. 7, this business trend predicting system may also include that
Judge module 5, for obtaining prediction term after the prediction data of selected predetermined period from data restoring module 4,
Judge whether the state of prediction term exception;
Alarm module 6, after judging that in judge module 5 exception in the state of prediction term, reports to the police to user.
As shown in figure 8, above judge module 5 mays include:
Current data acquiring unit 51, for being specially statistic unit 12 from data acquisition module 1() obtain prediction term
Current statistic history service data;
Computing unit 52, for calculating the current statistic history service data of selected predetermined period and selected predetermined period
The difference of prediction data;
Comparing unit 53, for being compared with default warning value to difference, if difference exceedes default warning value, really
Abnormal, notice alarm module warning in the state of fixed described prediction term, here, alarm module 6 can be in difference in setting
All occur in the section time reporting to the police to user again in the case of exception, carried out with avoiding the occurrence of because a secondary data once in a while there is error
The situation of false alarm.
Construction, feature and the action effect of the present invention, above institute are described in detail above according to the embodiment shown in schema
State only presently preferred embodiments of the present invention, but the present invention is not to limit practical range, every structure according to the present invention shown in drawing
Want made change, or be revised as the Equivalent embodiments of equivalent variations, when still covered with diagram without departing from specification spiritual,
All should be within the scope of the present invention.