CN104881704A - Telephone traffic index predicting method, apparatus and electronic equipment - Google Patents

Abstract

The invention provides a telephone traffic index predicting method, an apparatus and a piece of electronic equipment. The method comprises the following steps of determining the telephone traffic index seasonal component predicted value of a target time period according to the telephone traffic index seasonal component real value of each historic time period; and determining the telephone traffic index predicted value of the target time period according to the telephone traffic index trend component predicted value and the telephone traffic index seasonal component predicted value of the target time period. The method in the invention makes the telephone traffic index accurately predicted on the premise that the telephone traffic index seasonal component of the target time period is predictable.

Description

A kind of Forecasting Methodology of traffic indicator, device and electronic equipment

Technical field

The present invention relates to core network technology field, particularly relate to a kind of Forecasting Methodology of traffic indicator, device and electronic equipment.

Background technology

The number of users of present mobile communication and telephone traffic keep steady-state growth, and the operation that mobile network is stable for a long time, depends on network planning and optimization timely and effectively, and network operator successfully tackles change and the trend thereof of user behavior with the network change of initiative.When mobile communication telephone traffic exceedes certain capacity, very easily cause exchange system to transship, occur that circuit is congested, speech percent of call completed declines, than rising, even there is the phenomenon that switch large area is paralysed, all causing irretrievable loss to mobile communication carrier and mobile subscriber in traffic call drop.Therefore predict according to traffic measurement data and the variation tendency of other business information to mobile communication telephone traffic, the capacity that expanding capacity of project will reach is determined with this, and the useful capacity of the corresponding traffic model computing equipment set by various places, thus determine the device type and the quantity that meet design capacity, be necessary very much for mobile communication carrier.

For this reason, a kind of intelligent traffic predicting method based on different templates type has occurred.This intelligent traffic predicting method comprises following step: (1) is taken out parameter in recent years and trained, and determines initial parameter: growth factor, stencil value, each festivals or holidays traffic stencil value; (2) according to the prediction length of input, judgement is short-term or medium-term and long-term stage of stable development traffic forecast, and selected different stage of stable development traffic predicting method; (3) according to the result of step (2), stage of stable development traffic is predicted according to different algorithms; (4) according to the parameter that step (1) is determined, the length of location prediction interval each festivals or holidays and type, choose traffic forecast algorithm predicts traffic festivals or holidays corresponding festivals or holidays; (5) according to the result of step (3), (4), by stage of stable development traffic and festivals or holidays traffic carry out interface smooth process, synthesis prediction traffic, output to relevant device.

The feature of above-mentioned intelligent traffic predicting algorithm is divided into N section to predict the stage of stable development traffic of the whole year, and each section of stage of stable development traffic is considered as again being superposed by the simple addition of trend component and periodic component.Wherein, trend component is predicted by one-variable linear regression method and is obtained, and periodic component is predicted by template matching algorithm and obtained.

Meanwhile, have also appeared another kind of traffic predicting method.The method is set up based on ARMA (p, q) model, and arma modeling is obtained by autoregressive model and moving average model " mixing ", is a kind of important method of search time series model.The main flow process of the method is: the history value obtaining network traffic, as historical sample data, carries out pre-service to described historical sample data, obtains normal sample notebook data; Forecast model is utilized to carry out modeling and predict for described historical sample data; Utilize method of analysis of variance to described normal sample notebook data carry out red-letter day telephone traffic with telephone traffic significance test of difference at ordinary times; If red-letter day, telephone traffic was remarkable with the otherness of telephone traffic at ordinary times, then the predicted value of described forecast model is revised, obtain final predicted value; Otherwise the predicted value of described forecast model is final predicted value.

In addition, also have in " red-letter day traffic forecast and dilatation flow process " literary composition, to propose traffic predicting method a kind of festivals or holidays, contextual definition between telephone traffic is simple linear model by the method, utilizes linear prediction to add the method for modifying factor to obtain telephone traffic festivals or holidays in target time.

Summary of the invention

In view of this, the object of the embodiment of the present invention is to provide a kind of Forecasting Methodology of traffic indicator, device and electronic equipment, with in the predictable situation of traffic indicator seasonal component of object time section, realizes the Accurate Prediction of traffic indicator.

For solving the problems of the technologies described above, the embodiment of the present invention provides scheme as follows:

The embodiment of the present invention provides a kind of Forecasting Methodology of traffic indicator, comprising:

According to the traffic indicator seasonal component actual value of each historical time section, determine the traffic indicator seasonal component predicted value of object time section;

According to traffic indicator trend component predicted value and the traffic indicator seasonal component predicted value of object time section, determine the traffic indicator predicted value of object time section.

Preferably, the described traffic indicator seasonal component actual value according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, also comprises:

The traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

Preferably, the described traffic indicator seasonal component actual value according to each historical time section, determine that the traffic indicator seasonal component predicted value of object time section comprises:

Determine the weight of the traffic indicator seasonal component actual value of each historical time section;

According to described weight, the traffic indicator seasonal component actual value of each historical time section is weighted on average, obtains the traffic indicator seasonal component predicted value of object time section.

Preferably, each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, and the time number in each historical years and target time is n, and i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2 ..., n, n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njobtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}(u_k-\stackrel{\‾}{U_n})(u_k-\stackrel{\‾}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_j})}^2}}.$

Preferably, traffic indicator comprises registered user's number, and described method also comprises:

The telephone traffic of historical time section each in each historical time section and registered user's number are divided by, obtain user's liveness of each historical time section in each historical time section;

User's liveness that place vacation in each historical time section and object time section place have all historical time sections of same characteristic features vacation is averaged, obtains user's liveness predicted value of object time section;

Registered user's number predicted value of object time section is multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

Preferably, for each time period in each historical time section and object time section, the feature of place vacation this comprises time period: only have first red-letter day at this time period place in red-letter day that this place comprises vacation time period and be positioned at this first red-letter day place vacation this time period first or two days; Or first red-letter day at what this place comprised vacation time period only have red-letter day this time period place and this first red-letter day at the 3rd day of place vacation this time period; Or this time period place comprises the first red-letter day and second red-letter day at this time period place vacation.

Preferably, the first red-letter day comprised the Mid-autumn Festival, and the second red-letter day comprised National Day.

Preferably, the described traffic indicator trend component predicted value according to object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, also comprise:

According to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

Preferably, the described traffic indicator trend component actual value according to the first time period before object time section and object time section are relative to the traffic indicator trend component rate of growth predicted value of first time period, before determining the traffic indicator trend component predicted value of object time section, also comprise:

According to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately;

According to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

Preferably, described according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine that object time section comprises relative to the traffic indicator trend component rate of growth predicted value of first time period:

Utilize nonlinear regression model (NLRM) principle, relative to the traffic indicator trend component rate of growth actual value of the second time period before separately, data fitting is carried out to each historical time section, obtains rate of growth fitting function GR (Q)=ae ^bQ+ ε, wherein, a is constant, and b is constant, and ε is error term, and GR (Q) is for each historical time Duan Zhong Q historical time section is relative to the matching rate of growth of the traffic indicator trend component of the second time period before this historical time section;

GR (Q+1) is defined as the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

Preferably, time of belonging to respectively of each historical time section and object time section is continuous.

The embodiment of the present invention also provides a kind of prediction unit of traffic indicator, comprising:

First determination module, for the traffic indicator seasonal component actual value according to each historical time section, determines the traffic indicator seasonal component predicted value of object time section;

Second determination module, for according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, determines the traffic indicator predicted value of object time section.

Preferably, also comprise:

First computing module, for the traffic indicator seasonal component actual value of described first determination module according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, the traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

Preferably, described first determination module comprises:

First determining unit, for determining the weight of the traffic indicator seasonal component actual value of each historical time section;

Computing unit, for according to described weight, is weighted on average the traffic indicator seasonal component actual value of each historical time section, obtains the traffic indicator seasonal component predicted value of object time section.

Preferably, each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, and the time number in each historical years and target time is n, and i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2 ..., n, n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njobtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}(u_k-\stackrel{\‾}{U_n})(u_k-\stackrel{\‾}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_j})}^2}}.$

Preferably, traffic indicator comprises registered user's number, and described device also comprises:

Second computing module, for the telephone traffic of historical time section each in each historical time section and registered user's number being divided by, obtains user's liveness of each historical time section in each historical time section;

3rd computing module, for being averaged by user's liveness vacation place vacation in each historical time section and object time section place with all historical time sections of same characteristic features, obtains user's liveness predicted value of object time section;

4th computing module, for registered user's number predicted value of object time section being multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

Preferably, also comprise:

3rd determination module, for described second determination module according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, according to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

Preferably, also comprise:

Second determining unit, for described 3rd determination module according to the traffic indicator trend component actual value of the first time period before object time section and the object time section traffic indicator trend component rate of growth predicted value relative to first time period, before determining the traffic indicator trend component predicted value of object time section, according to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately,

3rd determining unit, for according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determines the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

The embodiment of the present invention also provides a kind of electronic equipment comprising the prediction unit of above-described traffic indicator.

As can be seen from the above, the embodiment of the present invention at least has following beneficial effect:

In the predictable situation of traffic indicator seasonal component of object time section, achieve the Accurate Prediction of traffic indicator.

Accompanying drawing explanation

Fig. 1 represents the flow chart of steps of the Forecasting Methodology of a kind of traffic indicator that the embodiment of the present invention provides;

Fig. 2 represents busy traffic forecast process flow diagram in red-letter day;

Fig. 3 represents busy registered user number prediction in red-letter day process flow diagram;

Fig. 4 represents the structured flowchart of the prediction unit of a kind of traffic indicator that the embodiment of the present invention provides.

Embodiment

For making the object of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawings and the specific embodiments the embodiment of the present invention is described in detail.

Fig. 1 represents the flow chart of steps of the Forecasting Methodology of a kind of traffic indicator that the embodiment of the present invention provides, and with reference to Fig. 1, the embodiment of the present invention provides a kind of Forecasting Methodology of traffic indicator, comprises the steps:

Step 101, according to the traffic indicator seasonal component actual value of each historical time section, determines the traffic indicator seasonal component predicted value of object time section;

Step 102, according to traffic indicator trend component predicted value and the traffic indicator seasonal component predicted value of object time section, determines the traffic indicator predicted value of object time section.

Visible, by the way, in the predictable situation of traffic indicator seasonal component of object time section, achieve the Accurate Prediction of traffic indicator.

Wherein, each historical time section and object time section all can be arranged in red-letter day, such as, can be whole day or the busy etc. in red-letter day in red-letter day.Red-letter day is such as: the Mid-autumn Festival or New Year's Eve etc.

Traffic indicator is such as: registered user's number or telephone traffic etc.

In the embodiment of the present invention, the described traffic indicator seasonal component actual value according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, can also comprise:

The traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

In the embodiment of the present invention, the described traffic indicator seasonal component actual value according to each historical time section, determine that the traffic indicator seasonal component predicted value of object time section can comprise:

Determine the weight of the traffic indicator seasonal component actual value of each historical time section;

According to described weight, the traffic indicator seasonal component actual value of each historical time section is weighted on average, obtains the traffic indicator seasonal component predicted value of object time section.

Wherein, each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, and the time number in each historical years and target time is n, and i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2 ..., n, n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njcan be obtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}(u_k-\stackrel{\‾}{U_n})(u_k-\stackrel{\‾}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{(u_k-\stackrel{\‾}{U_j})}^2}}.$

In the embodiment of the present invention, traffic indicator can comprise registered user's number, and described method can also comprise:

The telephone traffic of historical time section each in each historical time section and registered user's number are divided by, obtain user's liveness of each historical time section in each historical time section;

User's liveness that place vacation in each historical time section and object time section place have all historical time sections of same characteristic features vacation is averaged, obtains user's liveness predicted value of object time section;

Registered user's number predicted value of object time section is multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

Wherein, for each time period in each historical time section and object time section, the feature of place vacation this can comprise time period: only have first red-letter day at this time period place in red-letter day that this place comprises vacation time period and be positioned at this first red-letter day place vacation this time period first or two days; Or first red-letter day at what this place comprised vacation time period only have red-letter day this time period place and this first red-letter day at the 3rd day of place vacation this time period; Or this time period place comprises the first red-letter day and second red-letter day at this time period place vacation.

Wherein, the Mid-autumn Festival can be comprised the first red-letter day.

Second red-letter day can comprise National Day.

In the embodiment of the present invention, the described traffic indicator trend component predicted value according to object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, can also comprise:

According to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

Wherein, each historical time section and object time Duan Jun are arranged in the situation in red-letter day, first time period can be the first setting number month of individual month more Zao than object time section place month, then the traffic indicator trend component actual value of first time period can for meeting the mean value of the traffic indicator actual value on the date that imposes a condition in the first setting number month of individual month more Zao than object time section place month.

Wherein, the first setting number such as: the arbitrary number in natural number 1 ~ 11.

Meeting date that imposes a condition can comprise: except after the second setting number working day before place vacation in red-letter day and place vacation in red-letter day, the 3rd sets all working day except number working day.

Second sets number such as: 1,2 or 3.

3rd sets number such as: 1,2 or 3.

Further, the described traffic indicator trend component actual value according to the first time period before object time section and object time section are relative to the traffic indicator trend component rate of growth predicted value of first time period, before determining the traffic indicator trend component predicted value of object time section, can also comprise:

According to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately;

According to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

Wherein, each historical time section and object time Duan Jun are arranged in the situation in red-letter day, for historical time section each in each historical time section, the traffic indicator trend component actual value of this historical time section can for meeting the mean value of the traffic indicator actual value on the date that imposes a condition in this historical time section place month, this historical time section can be the first setting number month of individual month more Zao than this historical time section place month, the traffic indicator trend component actual value of the second time period before this historical time section can for meeting the mean value of the traffic indicator actual value on the date that imposes a condition in the first setting number month of individual month more Zao than this historical time section place month.

Like this, the traffic indicator mean value tendency meeting the date that imposes a condition in the month at each historical time section place separately tends to be steady and has downtrending, improve the accuracy that the traffic indicator mean value imposing a condition the date satisfied in object time section place month is predicted, thus improve the accuracy that the traffic indicator trend component of object time section is predicted.

The traffic indicator trend component actual value of the second time period before the described traffic indicator trend component actual value according to each historical time section and each historical time section, determine that in the step of each historical time section relative to the traffic indicator trend component rate of growth actual value of the second time period before separately, in each historical time section, an xth historical time section can be obtained by following formulae discovery relative to the traffic indicator trend component rate of growth actual value RRx of the second time period before it:

The traffic indicator trend component actual value * 100% of the second time period before RRx=(the traffic indicator trend component actual value of the second time period before traffic indicator trend component actual value-xth historical time section of an xth historical time section)/xth historical time section.

In addition, described according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine that object time section can comprise relative to the traffic indicator trend component rate of growth predicted value of first time period:

Utilize nonlinear regression model (NLRM) principle, relative to the traffic indicator trend component rate of growth actual value of the second time period before separately, data fitting is carried out to each historical time section, obtains rate of growth fitting function GR (Q)=ae ^bQ+ ε, wherein, a is constant, and b is constant, and ε is error term, and GR (Q) is for each historical time Duan Zhong Q historical time section is relative to the matching rate of growth of the traffic indicator trend component of the second time period before this historical time section;

GR (Q+1) is defined as the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

The time that each historical time section and object time section belong to respectively can be continuous.

< better embodiment >

For the embodiment of the present invention being set forth clearly clear, provide the better embodiment of the embodiment of the present invention below.

This better embodiment provides breakdown mobile network traffic predicting method festivals or holidays and system.

This better embodiment comprises three traffic forecast models (red-letter day is for the mid-autumn): the mid-autumn busy registered user number forecast model, the mid-autumn busy traffic forecast model and the mid-autumn whole day traffic forecast model.Detailed introduction is done below by the disposal route of three kinds of models.

Model 1: prediction busy registered user number in the mid-autumn

Because telephone traffic presents seasonal and steady growth property, accordingly, number of users also has identical character.Thus, also should consider seasonality and growth property when predicting busy number of users, the present invention supposes that busy number of users is the product model of user base number and seasonal factor, that is: simultaneously

Busy number of users U _busy=user base number U _basic× seasonal factor SF (1)

User base number U _basicfor remaining the number of users mean value of number of days after each month rejecting festivals or holidays, and the number of days of rejecting should extend 2 days respectively before and after vacation.As: calculate the user base number in September, suppose vacation in the Mid-autumn Festival be September 15 to September 17, then should by September 13,14 and September 18,19 number of users average again after rejecting in the lump.

1.1, user base number is predicted

Suppose T ₁moment prediction T ₂the user base number in moment, the historical data in total n cycle, wherein ${U_T}_1=\left(\begin{array}{c}{U}_{\mathrm{basic}\_11}\\ U_{\mathrm{basic}\_12}\\ .\\ .\\ .\\ U_{\mathrm{basic}\_1n}\end{array}\right)$ For T ₁the historic user radix vector in moment, ${U_T}_2=\left(\begin{array}{c}{U}_{\mathrm{basic}\_21}\\ U_{\mathrm{basic}\_22}\\ .\\ .\\ .\\ U_{\mathrm{basic}\_2n}\end{array}\right)$ For T ₂the historic user radix vector in moment; Δ U is T ₂moment relative T ₁number of users increase absolute number

$\mathrm{\&Delta;U}={U_T}_2-{U_T}_1=\left(\begin{array}{c}{U}_{\mathrm{basic}\_21}\\ U_{\mathrm{basic}\_22}\\ .\\ .\\ .\\ U_{\mathrm{basic}\_2n}\end{array}\right)-\left(\begin{array}{c}{U}_{\mathrm{basic}\_11}\\ U_{\mathrm{basic}\_12}\\ .\\ .\\ .\\ U_{\mathrm{basic}\_1n}\end{array}\right)=\left(\begin{array}{c}{U}_{\mathrm{basic}\_21}-U_{\mathrm{basic}\_11}\\ U_{\mathrm{basic}\_22}-U_{\mathrm{basic}\_12}\\ .\\ .\\ .\\ U_{\mathrm{basic}\_2n}-U_{\mathrm{basic}\_1n}\end{array}\right)---\left(2\right)$

The point division operation " ./" of definition matrix, A./B is the corresponding element of corresponding element divided by matrix B of matrix A.

If GR is T ₂moment relative T ₁user's rate of growth in moment, then:

$\mathrm{GR}=({U_T}_2-{U_T}_1)./{U_T}_1=\mathrm{\&Delta;U}./{U_T}_1\&times;100\%=\left(\begin{array}{c}\frac{U_{\mathrm{basic}\_21}-U_{\mathrm{basic}\_11}}{U_{\mathrm{basic}\_11}}\\ \frac{U_{\mathrm{basic}\_22}-U_{\mathrm{basic}\_12}}{U_{\mathrm{basic}\_12}}\\ .\\ .\\ .\\ \frac{U_{\mathrm{basic}\_2n}-U_{\mathrm{basic}\_1n}}{U_{\mathrm{basic}\_1n}}\end{array}\right)\&times;100\%---\left(3\right)$

GR is rate of growth vector, is designated as $\mathrm{GR}=\left(\begin{array}{c}{\mathrm{gr}}_1\\ {\mathrm{gr}}_2\\ .\\ .\\ .\\ {\mathrm{gr}}_n\end{array}\right),$ Wherein:

${\mathrm{gr}}_i=\frac{U_{\mathrm{basic}\_2i}-U_{\mathrm{basic}\_1i}}{U_{\mathrm{basic}\_1i}}\&times;100\%,i=\mathrm{1,2},...,n;$

Note time arrow is:

$T=\left(\begin{array}{c}{t}_1^{\&prime;}\\ t_2^{\&prime;}\\ .\\ .\\ .\\ t_n^{\&prime;}\end{array}\right)$

Use nonlinear regression model (NLRM) principle, the function about time T obtained to rate of growth data fitting, the form for following:

GR(t)＝ae ^bt+ε (4)

Wherein a, b are constant, and ε is error term

After obtaining the forecast model (4) of rate of growth, suppose that current point in time is t ₁, t ₁the user base number in moment is then t ₂moment is relative to t ₁user's rate of growth in moment is t ₂the user base number in moment can be obtained by following formula predictions:

${U_t}_2={U_t}_1(1+\mathrm{GR}\left(t_2\right))---\left(5\right)$

1.2, the process of seasonal factor

From 2008, country will split into 3 little long holidays " May Day " long holidays, i.e. 3 days Clear and Bright, 3 days Dragon Boat Festivals and in 3 days mid-autumns little long holidays.Due to the singularity of China's lunar calendar solar term, vacation in the Mid-autumn Festival may be little long holidays of independent 3 days, is also likely coincide with long holidays 7 day National Day, and at this moment the singularity of vacation in the mid-autumn will cause seasonal factor to occur larger fluctuation.So far having there is the phenomenon that 2 year vacation in the mid-autumn and vacation on National Day coincide in 2008 policies new vacation come into effect, is 2009 and 2012 respectively.Although be all the Mid-autumn Festival, normally the Mid-autumn Festival, busy number of users showed different seasonal characteristics from the busy number of users in the special Mid-autumn Festival (overlapping with National Day).Seasonal characteristic performance between normal Mid-autumn Festival in time busy number of users is comparatively close, and special Mid-autumn Festival in time, the seasonal characteristic performance of busy number of users was comparatively close.Therefore the prediction for seasonal factor can consider that use history seasonal factor weighted mean obtains predicting the seasonal factor in time.

1.1.1, the determination of weight

Weight is a relative concept, is for a certain index.The weight of a certain index refers to the relative importance of this index in the overall evaluation.Weight table is shown in evaluation procedure, is the rationed of the significance level of the not ipsilateral being evaluated object, treats with a certain discrimination the effect of each evaluation points in overall assessment.

Determine that the method for weight has a variety of, as investigation statistics method, sequence synthesis method, equation, Statistics Method, analytical hierarchy process, analysis of complexity method etc.Here to adopt the weight of correlation coefficient process determination different year to come how determining that weight is specifically described as an example, wherein, correlation coefficient process is the one of equation.

Suppose the history number of users every day data having n cycle, wherein the n-th issue is according to the data being the current prediction time, and every first phase historical data is designated as U _i=(u _i1, u _i2..., u _im) ^t, i=1,2 ..., n, wherein m is the data number of every first phase, and all historical datas can be designated as following matrix:

$D=\left(\begin{array}{cccc}{u}_{11}& u_{21}& ...& u_{n1}\\ u_{12}& u_{22}& ...& u_{n2}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ u_{1m}& u_{2m}& ...& u_{\mathrm{nm}}\end{array}\right)$

Related coefficient then between every first phase historical data can be obtained by following formulae discovery:

$r_{\mathrm{ij}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}(u_k-\stackrel{\&OverBar;}{U_i})(u_k-\stackrel{\&OverBar;}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_i})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_j})}^2}}i,j=\mathrm{1,2},...,n---\left(6\right)$

N-th phase user data and the 1st, 2 ..., the correlation matrix between n-1 phase user data is:

r＝(r _n1,r _n2,…,r _n(n-1))

So the weight of each phase can be obtained by following formulae discovery:

${\mathrm{\&omega;}}_j=\frac{r_{\mathrm{nj}}}{\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}{r}_{\mathrm{nj}}}---\left(7\right)$

ω＝(ω ₁,ω ₂,…,ω _n-1) (8)

1.1.2, the determination of seasonal factor

Note:

Suppose that vacation in the mid-autumn, history seasonal factor vector was designated as:

SF＝(a ₁,a ₂,…,a _n-1) (10)

Predict that the seasonal factor of the n-th phase is

${\mathrm{SF}}_n=\mathrm{\&omega;}\&CenterDot;{\mathrm{SF}}^T=\underset{j=1}{\overset{n-1}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_j{a}_j---\left(11\right)$

1.3, case test

Determine number of users radix: get 2007 to 2011 the whole network numbers of users in City in South China, T ₁moment is annual June, T ₂moment is annual September, then:

${U_T}_1=\left(\begin{array}{c}1141.13\\ 1314.85\\ 1430.98\\ 1615.77\\ 1673.35\end{array}\right),{U_T}_2=\left(\begin{array}{c}1187.10\\ 1347.23\\ 1476.82\\ 1639.83\\ 1689.19\end{array}\right)$

Obtain:

$\mathrm{GR}=\left(\begin{array}{c}4.03\%\\ 2.46\%\\ 3.20\%\\ 1.49\%\\ 0.95\%\end{array}\right),T=\left(\begin{array}{c}1\\ 2\\ 3\\ 4\\ 5\end{array}\right)$

Matching obtains the function of rate of growth about time t:

GR＝0.059e ^-0.34t

The rate of growth of prediction user base number in 2012 is GR (6)=0.059 × e ^{-0.34 × 6}=0.77%, the user base number in known in June, 2012 is 1736.13 ten thousand families, then the user base number in September, 2012 is families, 1736.13 × (1+0.77%)=1749.45 ten thousand.

Determine seasonal factor: little long holidays in the mid-autumn came into effect from 2008, get 2008 to 2012 1 to the average daily number of users data in June, calculate the related coefficient between each time number of users, as table 1:

Table 1

Time	2008	2009	2010	2011	2012
						2008	1
2009	0.63	1
						2010	0.82	0.41	1
2011	0.87	0.80	0.63	1
						2012	0.50	0.96	0.31	0.67	1

Correlation matrix then between number of users data in 2012 and 2008-2011 is r=(0.50,0.96,0.31,0.67), calculates weight matrix and is:

ω＝(0.20,0.39,0.13,0.27)

2008 to 2011 seasonal factor vectors are:

SF＝(0.96,0.89,0.96,0.94)

It is SF that prediction obtains the 2012 annual seasons factors ₂₀₁₂=ω SF ^t=0.9257.

The number of users then predicting the same day in the mid-autumn in 2012 is 0.9257 × 1749.45=1619.53 ten thousand family, and predicated error is

$\mathrm{ErrRate}=\frac{(1619.53-1608.44)}{1608.44}\&times;100\%=0.69\%$

Model 2: prediction busy telephone traffic on the same day in the mid-autumn

2.1, prediction principle

Suppose E _busyrepresent busy telephone traffic, U ₁₀for 10 registered user's numbers every day, Ap _busyfor busy user liveness, then busy telephone traffic can be expressed as:

E _busy＝U ₁₀×Ap _busy(12)

In model 1, set forth the Forecasting Methodology of busy number of users, the method for prediction 10 registered user's numbers is similar with the method for prediction busy registered user number, and therefore this section only need predict busy liveness Ap _busy.Mention the singularity due to China's lunar calendar solar term above, vacation in the Mid-autumn Festival may be little long holidays of independent 3 days, also may be coincide with long holidays 7 day National Day, and different features vacation makes the same day in mid-autumn busy user liveness present different rules.So far having there is the phenomenon that 2 year vacation in the mid-autumn and vacation on National Day coincide in 2008 policies new vacation come into effect, is 2009 and 2012 respectively.Little for mid-autumn long holidays are considered as normal vacation, and the mid-autumn overlaps with National Day and is considered as special vacation.For certain city, if be positioned at the first day of normal vacation and second day the Mid-autumn Festival, then there will be the situation that most of user roams into other districts and cities inside the province, the busy user liveness on now same day in the mid-autumn is defined as radix liveness Ap _basic; If be positioned at the 3rd day of normal vacation the Mid-autumn Festival, now due to the relation of vacation, the user roaming into other districts and cities is inside the province more on the low side than the situation of first, second day, but user speech call demand is larger, user's liveness is large, and busy user liveness is now designated as liveness Ap the 3rd day normal vacation ₃; If be positioned at special vacation the Mid-autumn Festival, busy user liveness is now designated as liveness Ap special vacation _special.

Suppose that n phase history radix liveness remembers into matrix A p _basic:

Ap _basic＝(Ap _{basic_1},Ap _{basic_2},…,Ap _{basic_n}) ^T

M phase history the 3rd day liveness is designated as matrix A p ₃:

Ap ₃＝(Ap _{3_1},Ap _{3_2},…,Ap _{3_m}) ^T

K phase history liveness special vacation is designated as matrix A p _special:

Ap _special＝(Ap _{special_1},Ap _{special_2},…,Ap _{special_k}) ^T

Note matrix I ₁, I ₂, I ₃be respectively:

I ₁＝(1,1,…,1) _1×n

I ₂＝(1,1,…,1) _1×m

I ₃＝(1,1,…,1) _1×k

Then:

${\stackrel{\&OverBar;}{\mathrm{AP}}}_{\mathrm{basic}}=\frac{1}{n}\&CenterDot;I_1\&CenterDot;A{p}_{\mathrm{basic}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{n}A{p}_{\mathrm{basic}\_i}$

${\stackrel{\&OverBar;}{\mathrm{AP}}}_3=\frac{1}{m}\&CenterDot;I_2\&CenterDot;A{p}_3=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}\frac{1}{n}A{p}_{3\_i}$

${\stackrel{\&OverBar;}{\mathrm{AP}}}_{\mathrm{special}}=\frac{1}{k}\&CenterDot;I_3\&CenterDot;A{p}_{\mathrm{special}}=\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\frac{1}{k}A{p}_{\mathrm{special}\_i}$

It is ω that note is positioned at the weight factor of the 3rd day normal vacation the Mid-autumn Festival ₃, the weight factor that the Mid-autumn Festival overlapped with National Day is ω _special, calculate ω ₃and ω _specialformula as follows:

$\begin{array}{c}{\mathrm{\&omega;}}_3=1+\frac{{\stackrel{\&OverBar;}{\mathrm{Ap}}}_3-{\stackrel{\&OverBar;}{\mathrm{Ap}}}_{\mathrm{basic}}}{{\stackrel{\&OverBar;}{\mathrm{Ap}}}_{\mathrm{basic}}}\\ {\mathrm{\&omega;}}_{\mathrm{special}}=1+\frac{{\stackrel{\&OverBar;}{\mathrm{Ap}}}_{\mathrm{special}}-{\stackrel{\&OverBar;}{\mathrm{Ap}}}_{\mathrm{basic}}}{{\stackrel{\&OverBar;}{\mathrm{Ap}}}_{\mathrm{basic}}}\end{array}---\left(13\right)$

Note set:

A＝{ω ₃,ω _special}

B={1,2,3; 1 expression is positioned at first and second sky little long holidays the Mid-autumn Festival,

2 expressions are positioned at the 3rd day little long holidays in Mid-autumn Festival, and 3 expressions overlap with National Day the Mid-autumn Festival }

Then busy user liveness anticipation function Ap on the same day in the mid-autumn (x, ω) is:

The mid-autumn busy liveness Ap _busy=Ap (x, ω).

2.2, case test

2008-2011 history feature vacation in the Mid-autumn Festival in City in South China is as shown in table 2:

Table 2

Predict that 10 registered user's numbers in 2012 are 1570.21 ten thousand families according to the method for model 1, obtaining the mid-autumn in 2012 busy user liveness according to the method prediction of model 2 is 0.02294, then the same day in the mid-autumn, busy traffic forecast value was 1570.21 × 0.02294=36.02 ten thousand Erl, the actual telephone traffic of the busy mid-autumn in 2012 is 34.06 ten thousand Erl, and predicated error is:

ErrRate=(36.02-34.06)/34.06=5.44%

Model 3: prediction telephone traffic on the same day in the mid-autumn

The same day in the mid-autumn telephone traffic Forecasting Methodology and the Forecasting Methodology of busy registered user number similar, be divided into two parts: one is predicting telephone traffic radix, two is predicting telephone traffic seasonal factors, product model prediction telephone traffic cardinal sum seasonal factor out being applied mechanically (12) obtains the traffic forecast value on the same day in the mid-autumn, and therefore not to repeat here.

3.1, case test

3.3.1, predicting telephone traffic radix

Get the day traffic data in June, 2008 to 2011 and September, T ₁moment is annual June, T ₂moment is annual September, then obtain after carrying out data scrubbing:

${E_T}_1=\left(\begin{array}{c}312.23\\ 349.48\\ 417.90\\ 436.71\end{array}\right),{E_T}_2=\left(\begin{array}{c}325.19\\ 374.56\\ 430.25\\ 440.18\end{array}\right)$

Obtain rate of growth and time matrix:

$\mathrm{GR}=\left(\begin{array}{c}3.99\%\\ 6.70\%\\ 2.87\%\\ 0.79\%\end{array}\right),T=\left(\begin{array}{c}1\\ 2\\ 3\\ 4\end{array}\right)$

Matching obtains the function of rate of growth about time t:

GR＝0.116e ^-0.57t

Telephone traffic rate of growth is GR (5)=0.67% to predict in September, 2012, and the telephone traffic radix in June, 2012 is 453.96 ten thousand Erl, then telephone traffic radix is 453.96 × (1+0.67%)=457.00 ten thousand Erl to predict September.

3.3.2, the determination of telephone traffic seasonal factor

Get 2008 to 2012 years 1 to day in June traffic data, after data scrubbing, calculate the related coefficient form between each year, as shown in table 3:

Table 3

Time	2008	2009	2010	2011	2012
						2008	1
2009	0.44	1
						2010	0.68	0.25	1
2011	0.78	0.65	0.52	1
						2012	0.35	0.85	0.16	0.53	1

Correlation matrix then between 2012 and 2008-2011 each year is: r=(0.35,0.85,0.16,0.53), and calculating weight matrix is:

ω＝(0.19,0.45,0.09,0.28)

2008-2011 telephone traffic seasonal factor in each mid-autumn in year vector is:

SF＝(1.173,0.927,1.122,1.087)

Seasonal factor is ω SF then to predict the same day in the mid-autumn in 2012 ^t=1.034.

Predict that the telephone traffic E on the same day in the mid-autumn in 2012 is 457.00 × 1.034=472.71 ten thousand Erl, predicated error is:

$\mathrm{ErrRate}=\frac{(472.71-447.79)}{447.79}\&times;100\%=5.57\%$

The Comparative result of the traffic predicting method that this better embodiment provides and additive method:

(1) adopt traffic predicting method described in this better embodiment to carry out case test to A city, B city, C city and four, D city districts and cities' data in the mid-autumn, experimental result is in shown in table 4 and table 5; Table 6 is the inspections to E city, F city and H city New Year's Eve result.Original predicated error can significantly reduce by result display methods described herein, and effectively improve traffic forecast precision, model has good versatility.

Table 4 aging method and new method predicted portions districts and cities whole day in mid-autumn telephone traffic Comparative result

Table 5 aging method and new method predicted portions districts and cities busy in mid-autumn registered user number Comparative result

Table 6 new method predicted portions districts and cities' New Year's Eve busy number of users and telephone traffic

This better embodiment relates to the traffic forecast model of three kinds of situations, and each model considers trend growth property and seasonality all simultaneously, and both are formed composite model with multiplication form.

This better embodiment takes nonlinear regression method to process to trend growth property, and seasonal factor takes method of weighted mean process.

This better embodiment is have also contemplated that single user liveness to during busy traffic forecast; what the prediction of single user liveness adopted is template matching method; is divided into normal vacation (3 days little long holidays) vacation in the mid-autumn and special vacation (overlapping with vacation on National Day) processes respectively, should give protection.

This better embodiment compensate for live traffic Forecasting Methodology by tendency and the seasonal deficiency being thought of as addition model, change the Forecasting Methodology of tendency into non-linear regression by linear regression simultaneously, weighted mean predicted method is taked in the prediction of seasonal factor, true traffic conditions pressed close to more by traffic forecast model festivals or holidays the present invention being set up by above-mentioned process, effectively raises traffic forecast precision.

There is following defect in existing traffic predicting method: telephone traffic is not merely the simple addition superposition of trend component and periodic component, when trend component and periodic component not only immediately, this stacking method likely can allow traffic forecast model can not truly reflect traffic actual conditions.

Domestic Carriers voice market is through more than ten years high speed development, current voice traffic has tended to be steady growth, the great outburst of nearly 2 years internets especially mobile Internet brings very large impact to the voice service of Incumbent, and this causes voice service to have a declining tendency.So live traffice trend component does not in most of the cases meet simple linear model, the practical development situation in use unitary linear prediction model reflection current voice market that anticipation trend component can not be correct.

This better embodiment achieve when trend component and periodic component not only immediately, telephone traffic Accurate Prediction when simultaneously trend component tends to be steady and has downtrending.

Fig. 4 represents the structured flowchart of the prediction unit of a kind of traffic indicator that the embodiment of the present invention provides, and with reference to Fig. 4, the embodiment of the present invention also provides a kind of prediction unit of traffic indicator, comprising:

First determination module 401, for the traffic indicator seasonal component actual value according to each historical time section, determines the traffic indicator seasonal component predicted value of object time section;

Second determination module 402, for according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, determines the traffic indicator predicted value of object time section.

Visible, by the way, in the predictable situation of traffic indicator seasonal component of object time section, achieve the Accurate Prediction of traffic indicator.

In the embodiment of the present invention, can also comprise:

First computing module, for the traffic indicator seasonal component actual value of described first determination module 401 according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, the traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

In the embodiment of the present invention, described first determination module 401 can comprise:

First determining unit, for determining the weight of the traffic indicator seasonal component actual value of each historical time section;

Computing unit, for according to described weight, is weighted on average the traffic indicator seasonal component actual value of each historical time section, obtains the traffic indicator seasonal component predicted value of object time section.

Particularly, Ke Yiyou:

Each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, the time number in each historical years and target time is n, i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2, n, the n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njobtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}(u_k-\stackrel{\&OverBar;}{U_n})(u_k-\stackrel{\&OverBar;}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_j})}^2}}.$

In the embodiment of the present invention, traffic indicator can comprise registered user's number, and described device can also comprise:

Second computing module, for the telephone traffic of historical time section each in each historical time section and registered user's number being divided by, obtains user's liveness of each historical time section in each historical time section;

3rd computing module, for being averaged by user's liveness vacation place vacation in each historical time section and object time section place with all historical time sections of same characteristic features, obtains user's liveness predicted value of object time section;

4th computing module, for registered user's number predicted value of object time section being multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

In the embodiment of the present invention, can also comprise:

3rd determination module, for described second determination module 402 according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, according to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

Further, can also comprise:

Second determining unit, for described 3rd determination module according to the traffic indicator trend component actual value of the first time period before object time section and the object time section traffic indicator trend component rate of growth predicted value relative to first time period, before determining the traffic indicator trend component predicted value of object time section, according to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately,

3rd determining unit, for according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determines the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

The embodiment of the present invention also provides a kind of electronic equipment, and described electronic equipment comprises the prediction unit of above-described traffic indicator.

The above is only the embodiment of the embodiment of the present invention; should be understood that; for those skilled in the art; under the prerequisite not departing from embodiment of the present invention principle; can also make some improvements and modifications, these improvements and modifications also should be considered as the protection domain of the embodiment of the present invention.

Claims (19)

1. a Forecasting Methodology for traffic indicator, is characterized in that, comprising:

According to the traffic indicator seasonal component actual value of each historical time section, determine the traffic indicator seasonal component predicted value of object time section;

According to traffic indicator trend component predicted value and the traffic indicator seasonal component predicted value of object time section, determine the traffic indicator predicted value of object time section.

2. method according to claim 1, is characterized in that, the described traffic indicator seasonal component actual value according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, also comprises:

The traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

3. method according to claim 1, is characterized in that, the described traffic indicator seasonal component actual value according to each historical time section, determines that the traffic indicator seasonal component predicted value of object time section comprises:

Determine the weight of the traffic indicator seasonal component actual value of each historical time section;

According to described weight, the traffic indicator seasonal component actual value of each historical time section is weighted on average, obtains the traffic indicator seasonal component predicted value of object time section.

4. method according to claim 3, it is characterized in that, each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, the time number in each historical years and target time is n, i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2, n, the n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njobtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}(u_k-\stackrel{\&OverBar;}{U_n})(u_k-\stackrel{\&OverBar;}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_j})}^2}}.$

5. method according to claim 1, is characterized in that, traffic indicator comprises registered user's number, and described method also comprises:

The telephone traffic of historical time section each in each historical time section and registered user's number are divided by, obtain user's liveness of each historical time section in each historical time section;

User's liveness that place vacation in each historical time section and object time section place have all historical time sections of same characteristic features vacation is averaged, obtains user's liveness predicted value of object time section;

Registered user's number predicted value of object time section is multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

6. method according to claim 5, it is characterized in that, for each time period in each historical time section and object time section, the feature of place vacation this comprises time period: only have first red-letter day at this time period place in red-letter day that this place comprises vacation time period and be positioned at this first red-letter day place vacation this time period first or two days; Or first red-letter day at what this place comprised vacation time period only have red-letter day this time period place and this first red-letter day at the 3rd day of place vacation this time period; Or this time period place comprises the first red-letter day and second red-letter day at this time period place vacation.

7. method according to claim 6, is characterized in that, the first red-letter day comprised the Mid-autumn Festival, and the second red-letter day comprised National Day.

8. method according to claim 1, is characterized in that, the described traffic indicator trend component predicted value according to object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, also comprise:

According to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

9. method according to claim 8, it is characterized in that, the described traffic indicator trend component actual value according to the first time period before object time section and object time section are relative to the traffic indicator trend component rate of growth predicted value of first time period, before determining the traffic indicator trend component predicted value of object time section, also comprise:

According to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately;

According to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

10. method according to claim 9, it is characterized in that, described according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determine that object time section comprises relative to the traffic indicator trend component rate of growth predicted value of first time period:

Utilize nonlinear regression model (NLRM) principle, relative to the traffic indicator trend component rate of growth actual value of the second time period before separately, data fitting is carried out to each historical time section, obtains rate of growth fitting function GR (Q)=ae ^bQ+ ε, wherein, a is constant, and b is constant, and ε is error term, and GR (Q) is for each historical time Duan Zhong Q historical time section is relative to the matching rate of growth of the traffic indicator trend component of the second time period before this historical time section;

GR (Q+1) is defined as the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

11. methods according to claim 9, is characterized in that, the time that each historical time section and object time section belong to respectively is continuous.

The prediction unit of 12. 1 kinds of traffic indicators, is characterized in that, comprising:

First determination module, for the traffic indicator seasonal component actual value according to each historical time section, determines the traffic indicator seasonal component predicted value of object time section;

Second determination module, for according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, determines the traffic indicator predicted value of object time section.

13. devices according to claim 12, is characterized in that, also comprise:

First computing module, for the traffic indicator seasonal component actual value of described first determination module according to each historical time section, before determining the traffic indicator seasonal component predicted value of object time section, the traffic indicator actual value of each historical time section and traffic indicator trend component actual value are divided by respectively, obtain the traffic indicator seasonal component actual value of each historical time section.

14. devices according to claim 12, is characterized in that, described first determination module comprises:

First determining unit, for determining the weight of the traffic indicator seasonal component actual value of each historical time section;

Computing unit, for according to described weight, is weighted on average the traffic indicator seasonal component actual value of each historical time section, obtains the traffic indicator seasonal component predicted value of object time section.

15. devices according to claim 14, it is characterized in that, each historical time section belongs to the setting red-letter day in each historical years respectively, object time section belongs to the setting red-letter day in the target time, the time number in each historical years and target time is n, i-th time period in each historical time section and object time section belongs to each historical years and i-th time in the target time, i=1,2, n, the n-th time was the target time, if the traffic indicator actual value set U on m setting date in i-th time _i=(u _i1, u _i2..., u _im) ^t, then describedly determine in the step of the weight of the traffic indicator seasonal component actual value of each historical time section, the weight r of the traffic indicator seasonal component actual value of a jth historical time section in each historical time section _njobtained by following formulae discovery:

$r_{\mathrm{nj}}=\frac{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}(u_k-\stackrel{\&OverBar;}{U_n})(u_k-\stackrel{\&OverBar;}{U_j})}{\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_n})}^2}\sqrt{\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{(u_k-\stackrel{\&OverBar;}{U_j})}^2}}.$

16. devices according to claim 12, it is characterized in that, traffic indicator comprises registered user's number, described device also comprises:

Second computing module, for the telephone traffic of historical time section each in each historical time section and registered user's number being divided by, obtains user's liveness of each historical time section in each historical time section;

3rd computing module, for being averaged by user's liveness vacation place vacation in each historical time section and object time section place with all historical time sections of same characteristic features, obtains user's liveness predicted value of object time section;

4th computing module, for registered user's number predicted value of object time section being multiplied with user's liveness predicted value of object time section, obtains the traffic forecast value of object time section.

17. devices according to claim 12, is characterized in that, also comprise:

3rd determination module, for described second determination module according to the traffic indicator trend component predicted value of object time section and traffic indicator seasonal component predicted value, before determining the traffic indicator predicted value of object time section, according to the traffic indicator trend component actual value of the first time period before object time section and object time section relative to the traffic indicator trend component rate of growth predicted value of first time period, determine the traffic indicator trend component predicted value of object time section.

18. devices according to claim 17, is characterized in that, also comprise:

Second determining unit, for described 3rd determination module according to the traffic indicator trend component actual value of the first time period before object time section and the object time section traffic indicator trend component rate of growth predicted value relative to first time period, before determining the traffic indicator trend component predicted value of object time section, according to the traffic indicator trend component actual value of the second time period before the traffic indicator trend component actual value of each historical time section and each historical time section, determine the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately,

3rd determining unit, for according to the traffic indicator trend component rate of growth actual value of each historical time section relative to the second time period before separately, determines the traffic indicator trend component rate of growth predicted value of object time section relative to first time period.

19. 1 kinds of electronic equipments, is characterized in that, comprise the prediction unit of the traffic indicator as described in claim arbitrary in claim 12 to 18.