CN108230049A - The Forecasting Methodology and system of order - Google Patents
The Forecasting Methodology and system of order Download PDFInfo
- Publication number
- CN108230049A CN108230049A CN201810136485.5A CN201810136485A CN108230049A CN 108230049 A CN108230049 A CN 108230049A CN 201810136485 A CN201810136485 A CN 201810136485A CN 108230049 A CN108230049 A CN 108230049A
- Authority
- CN
- China
- Prior art keywords
- order
- history
- prediction
- type
- prediction model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Abstract
The invention discloses the Forecasting Methodology and system of a kind of order, wherein, method includes：Obtain the data of History Order, the sequence information of quantity and each History Order including History Order；Classified according to the sequence information of each History Order to History Order, obtain at least a kind of order type；The sequence information of the quantity of History Order based on each order type and History Order therein, the multiple prediction models of training, weighting coefficient obtaining each prediction model, carrying out for each order type quantitative forecast；By multiple prediction models and the weighting coefficient of the quantitative forecast of each prediction model, the quantity on order of any order type in preset time period is predicted, obtains prediction result.This method can greatly improve the accuracy of the quantity on order prediction to any order type in preset time period, so as to reasonable arrangement attendant in advance, achieve the purpose that cost-effective and provide better service to client.
Description
Technical field
The present invention relates to field of computer technology, the Forecasting Methodology and system of more particularly to a kind of order.
Background technology
Nowadays, more and more manufacturers provide family life and make house calls, such as：Household electrical appliances are repaired, household is installed etc., largely
Order also occurs therewith.If the order volume that each service type generates can be predicted, can with arranged rational attendant, so as to
It is cost-effective and give client provide better service.
In the relevant technologies, Forecasting Methodology relative precision is poor, in order to improve the accuracy of prediction, has very to prediction model
High request, and different models have different application conditions, it is more using limiting.Such as：Gray level model is used for predicting fluctuating range
Little data；Bp (back propagation, backpropagation) neural network is suitble to predict the data of various dimensions；Seasonal index number
Exponential smoothing is suitble to have periodically variable data.And family life is made house calls the order of class, is had spontaneous and when predicting
Prediction fluctuating range variation differs, if still using of the prior art according to different time different scenes, it is different pre- to use
It surveys model or weight coefficient is predicted, it is very cumbersome.
Invention content
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, an object of the present invention is to provide a kind of any orders relatively accurately predicted in preset time period
The Forecasting Methodology of type quantity on order.
It is another object of the present invention to propose a kind of forecasting system of order.
In order to achieve the above objectives, the embodiment of one aspect of the present invention discloses a kind of Forecasting Methodology of order, including following
Step：The data of History Order are obtained, wherein, the data of the History Order include the quantity of the History Order and each go through
The sequence information of history order；Classified according to the sequence information of each History Order to the History Order, obtain to
Few a kind of order type；The sequence information of the quantity of History Order based on each order type and History Order therein,
The multiple prediction models of training, with weighting that obtain each prediction model, to carry out for each order type quantitative forecast
Coefficient；By the multiple prediction model and each prediction model, for each order type carry out quantity
The weighting coefficient of prediction predicts the quantity on order of any order type in preset time period, obtains prediction result.
The Forecasting Methodology of the order of the embodiment of the present invention, by the information of the History Order to being collected into History Order into
Row classification, establishes multiple prediction models, and obtains each prediction model, adding for each order type progress quantitative forecast
Weight coefficient, and then obtain according to the weighting coefficient of multiple prediction models any order type quantity on order in preset time period
Prediction result substantially increases the accuracy of prediction result, so as to reasonable arrangement attendant in advance, reach saving into
The purpose of more good service originally and to client is provided.
In some instances, the sequence information according to each History Order divides the History Order
Class, including：The property of each History Order is determined according to the sequence information of each History Order；According to described each
The property of History Order classifies to the History Order, and obtains at least a kind of order type.
In some instances, the property of each History Order includes periodical, sudden and relevance.
In some instances, the multiple prediction model includes gray scale prediction model, seasonal index smoothing model and god
Through network model.
In some instances, the quantity of the History Order based on each order type and History Order therein are ordered
Single information, the multiple prediction models of training, with obtain each prediction model, for each order type carry out quantitative forecast
Weighting coefficient, including：It is predicted according to prediction model each in predetermined amount of time for the History Order of each order type
The actual quantity of the order type quantity and the order type determines the prediction error of each prediction model；According to each pre-
The prediction error for surveying model obtains error sum of squares；It is true using optimal weighted method according to the error sum of squares of each prediction model
The weighting coefficient of fixed corresponding prediction model.
In some instances, the Forecasting Methodology of the order further includes：According to the prediction result respectively to described more
A prediction model is trained and optimizes.
The embodiment of another aspect of the present invention discloses a kind of forecasting system of order, including：Acquisition module, for obtaining
The data of History Order, wherein, quantity and each History Order of the data including the History Order of the History Order
Sequence information；Sort module for being classified according to the sequence information of each History Order to the History Order, obtains
To at least a kind of order type；Weighting coefficient acquisition module, for the History Order according to each order type quantity and
The sequence information of History Order therein, the multiple prediction models of training, with obtain each prediction model, order for each
Single type carries out the weighting coefficient of quantitative forecast；Prediction module, for passing through the multiple prediction model and each institute
State prediction model, for each order type carry out quantitative forecast weighting coefficient, to any order in preset time period
The quantity on order of type is predicted, obtains prediction result.
The forecasting system of the order of the embodiment of the present invention, by the information of the History Order to being collected into History Order into
Row classification, establishes multiple prediction models, and obtains each prediction model, adding for each order type progress quantitative forecast
Weight coefficient, and then obtain according to the weighting coefficient of multiple prediction models any order type quantity on order in preset time period
Prediction result substantially increases the accuracy of prediction result, so as to reasonable arrangement attendant in advance, reach saving into
The purpose of more good service originally and to client is provided.
In some instances, institute's sort module is further used for：According to the sequence information of each History Order
Determine the property of each History Order；Classified according to the property of each History Order to the History Order,
And obtain at least a kind of order type.
In some instances, the property of each History Order includes periodical, sudden and relevance.
In some instances, the multiple prediction model includes gray scale prediction model, seasonal index smoothing model and god
Through network model.
In some instances, the weighting coefficient acquisition module is further used for：It is each predicted according in predetermined amount of time
Model is directed to the quantity of the order type and the actual quantity of the order type that the History Order of each order type is predicted,
Determine the prediction error of each prediction model；According to the prediction error of each prediction model obtain error sum of squares and according to
The error sum of squares of each prediction model determines the weighting coefficient of corresponding prediction model using optimal weighted method.
In some instances, the forecasting system of the order, further includes：Model optimization module, for being tied according to prediction
Fruit is trained and optimizes to the multiple prediction model respectively.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
It obtains significantly or is recognized by the practice of the present invention.
Description of the drawings
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Significantly and it is readily appreciated that, wherein：
Fig. 1 is the Forecasting Methodology flow chart according to the order of the embodiment of the present invention；With
Fig. 2 is the forecasting system structure diagram according to the order of the embodiment of the present invention.
Specific embodiment
The embodiment of the present invention is described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
The Forecasting Methodology and system of order according to embodiments of the present invention are described below in conjunction with attached drawing, introduces the present invention first
The Forecasting Methodology of the order of embodiment.
Fig. 1 is the Forecasting Methodology flow chart of order according to embodiments of the present invention, as shown in Figure 1, according to one of the invention
The Forecasting Methodology of the order of embodiment, includes the following steps：
S1：The data of History Order are obtained, wherein, the data of History Order include the quantity of History Order and each history
The sequence information of order.
In specific example, the data of History Order can be from each different types of service in the city to be predicted
It obtains and summarizes, and real-time update in standing.
In specific example, the data of History Order can be any time period in the city past few years to be predicted
Data in (such as 3 months).The sequence information of History Order can include the letters such as order time, order position, order price
Breath.
S2：Classified according to the sequence information of each History Order to History Order, obtain at least a kind of order class
Type.
Specifically, sequence information can include the information such as order time, order position, order price, according to these letters
Breath classifies History Order.
Specifically, the property of each History Order is determined according to the sequence information of each History Order for the first time；Further according to every
The property of a History Order classifies to History Order, at least obtains one kind that is following but not only including following order type
Order type.
Specifically, the property of History Order can be divided into according to the sequence information of History Order but is not limited only to be divided into
Periodical order (such as summer circuit system is safeguarded), sudden order (gas pipeline maintenance), relevance order (such as lower water
Pipe-dredging, piping foundation, pipeline cleaning).History Order is divided into, and according to difference by different order types according to above-mentioned property
Order type, all History Orders are summarized, and the quantity of the History Order of each order type can be counted.
S3：The sequence information of the quantity of History Order based on each order type and History Order therein, training are more
A prediction model, with obtain each prediction model, for each order type carry out quantitative forecast weighting coefficient.
Specifically, for each order property (such as：Periodical order, sudden order, relevance order etc.) institute
The order type of division, using the History Order data of each order type as mode input data, to the multiple predictions of training
Model.After completing model training, each prediction model, quantitative forecast for each order type weighting coefficient is obtained.
Wherein, the data of each order history order include the quantity of each order history order and the order letter of each History Order
Breath.
In specific example, multiple prediction models include but are not limited to the smooth mould of gray scale prediction model, seasonal index
Type and neural network model.
It is understood that in a specific example, gray scale prediction model, seasonal index smoothing model can be utilized
It is predicted respectively according to the data of History Order that get with neural network model, and by prediction result Macro or mass analysis, with
Reach comprehensive and accurate purpose.
Wherein, grey forecasting model is suitble to predict the system containing uncertain factor, so being predicted using gray scale
Model carries out quantitative forecast to each order type.Mould is predicted according to the data configuration gray scale of the History Order of each order type
Type, prediction is sometime and/or the quantity that places an order of certain scene.Wherein, the data of each order history order include each ordering
The sequence information of the quantity of single History Order and each History Order.
This step can include with realizing process example：Construction grey forecasting model, testing model process and it is related to
Algorithm.Its process is as follows：
In a specific example, equipped with initial data row x (0)=(x (0) (1), x (0) (2) ..., x (0) (n)),
In, n is above-mentioned historical data number, and initial data is classified as the data of the quantity of each order history order being collected into.Certainly,
The method for increasing data amount check n can be used to increase the credibility of prediction.It can be arranged according to x (0) data and establish grey forecasting model
Realize forecast function, then basic step is as follows：
(1) initial data is added up so as to the fluctuation and randomness that weaken random sequence, such as abnormal number can be eliminated
According to the influence to prediction result, new data sequence is obtained：
X (1)={ x (1) (1), x (1) (2) ..., x (1) (n) },
Wherein, each data represent the cumulative of corresponding former item datas in x (1) (t), such as：Two item numbers before x (1) (2) is represented
According to it is cumulative, as shown in following equation.
(2) following linear first-order differential equations are established to x (1) (t)：That is GM (1,1) (Grey Model gray models), such as
Shown in lower：
Wherein, a, u are undetermined coefficient, are referred to as development coefficient and grey actuating quantity, represent the following predetermined time respectively
The development trend of pre-generated quantity of such order and data variation relation in section.Further, the valid interval of a be (- 2,
2), and remember a, the matrix that u is formed is grey parameterIt is corresponding with gray model.As long as parameter a, u is obtained, with regard to that can be obtained
X (1) (t), and then pre-manufactured quantity in the following predetermined amount of time of x (0) is obtained, it is as follows, it is calculating process：
(3) Accumulating generation data are done with average generation B and constant item vector Yn, i.e.,：
(4) grey parameter is solved with least square methodThen
(5) by grey parameterIt substitutes intoIt is and rightIt is solved, is obtained
It should be noted that due toIt is the approximation being obtained by least square method, soIt is one near
Like expression formula, in order to be distinguished with former sequence x (1) (t+1), therefore it is denoted asT is time series in above-mentioned formula, can
To take year, season or moon etc..Wherein, forCalculating there is existing formula to realize, be not repeated herein.
(6) to function expressionAndIt carries out discrete, and the two is made the difference to restore x (0) originals
Sequence, that is to say, that obtain the likelihood data sequence of History Order quantity x (0) (t+1)Its process is as follows：
(7) it tests to the gray model of foundation, specifically checking procedure is divided into：Calculate x (0) (t) with(0)(t)
Between residual error e (0) (t) and calculate relative error q (0) (t), calculation formula is as follows：
Q (0) (t)=e (0) (t)/x (0) (t)
Judge whether the Precision of Grey Model of the present embodiment can reach prediction requirement according to the size of residual sum relative error,
If can, the pre-generated quantity of such order in following predetermined amount of time is predicted.
(8) the pre-generated quantity of order is predicted using gray model：
Specifically, according to different forecast demands, take different time sequence that can obtain corresponding prediction result.
Then, in a specific example, seasonal exponential smoothing method is suitble to have periodically variable data, with history
The data of order, the quantity including sorted each order history order, the time series of order construction seasonal index are put down
Sliding formwork type carrys out the pre-generated quantity of such order predetermined amount of time Nei.It specifically includes：Construct the mistake of seasonal index smoothing model
Journey and the algorithm being related to, formula are as follows：
In a specific example, if the time series period is l,, it is known that observation x1, x2..., x2l, wherein, observation
It is worth the History Order quantity and sequence information for period each in each period, using the prediction process of seasonal exponential smoothing method
It is as follows：
1) average horizontal value in each period in the first two period is calculated respectively：
In this embodiment it is possible to it is every in each period l that is understood, which is the average horizontal value in each period in the period,
The History Order quantity in a period.
2) increment in average each period in two periods is calculated, i.e., the History Order in each period in each period l
The increment of quantity：
3) initial exponential smooth value is calculated：
It is understood that the development trend in the model sequential can be found out according to initial exponential smooth value.
4) indexes of seasonal variation C in each period the first two period Nei is determinedt', each period in a cycle
The indexes of seasonal variation are：
During wherein t=1, when m=1, t=2, when m=2 ..., t=l, m=l.The season in each period in second period
Variation index is：
Wherein, as t=l+1, when m=1 ..., t=2l, m=l.
5) indexes of seasonal variation in average each period in the first two period are calculated：
The one shared l average indexes of seasonal variation.It is understood that the indexes of seasonal variation are to find out seasonal factor
Influence to History Order quantity is separated carry out subsequent operation.
6) by above-mentioned indexes of seasonal variation normal state：
L "=C "l+1+C″l+2+…+C″2l,
7) according to above-mentioned formula, each period in the third period is predicted：
Ft+m=(S+mB) Ct-l+m,
During wherein t=2l, m takes 1,2 ..., l.That is tentative prediction 2l+1,2l+2 ..., the value in 3l periods, thus may be used
To predict the pre-generated quantity of such order in each period in the third period.
8) exponential smoothing value is modified：(the x after observation in first period in third period is obtainedt=
x2l+1), the smoothing factor α determined with one group, beta, gamma modified index smooth value, trend and the indexes of seasonal variation, correction formula：
Bt=γ (St- S)+(1- γ) B,
It can predict the numerical value in remaining (l-1) a period third period Nei again in this way：
Ft+m=(St+mBt)Ct-l+m,
Wherein m=1,2 ..., l-1.
It is understood that with constantly being corrected to exponential smoothing value, which can be with
More close to forecast demand at this stage, prediction result is with real-time and more accurate.
9) hereafter, work as the observation x in t periods before obtaining every timetWhen, it is possible to it calculates single index respectively with the following formula and puts down
Sliding value, trend and the indexes of seasonal variation：
Bt=γ (St-St-1)+(1-γ)Bt-1,
Wherein m=1,2 ..., l. are whenever having been calculated a cycle, after obtaining the l indexes of seasonal variation, by 6 method,
They are subject to normal state again.
10) smoothing factor selects：
It should be noted that smoothing factor determine smooth level and between predicted value and actual result difference response
Speed.The value of smoothing factor should determine that iteration is since 0.01, step-length 0.01, directly according to the characteristics of known time sequence data
To 1, minimum prediction error value is selected.
Wherein, the seasonal index smoothing model reference value of the present embodiment is：
α：When more stable level trend is presented in former ordered series of numbers (History Order quantity)：0.1~0.3；It fluctuates larger：0.3
~0.5；It is also obvious to fluctuate very main trend：0.6~0.8.β：It is handled by actual conditions.γ：When predicted time is 3 units
0.01 is selected when selecting 0.1,6 units.
In a specific example, BP neural network then is trained with data after processing, achievees the purpose that predict order.
In specific example, according to order position, the predetermined time, order category, the order type history service times and
History Order quantity of the order type unit interval etc. instructs BP neural network as the training sample of BP neural network
Practice.These data can be carried out before training as managed, to improve training effectiveness.Such as：Data are normalized
Deng.
In specific example, BP neural network algorithmic procedure is as follows：
(1) sample is divided into training sample and test sample, and be normalized：
Such as：End value is made to be mapped between [0-1].Transfer function is as follows：
Wherein max is the maximum value of sample data, and min is the minimum value of sample data.X be initial data input value, X*
For output valve after normalization.
(2) input layer of BP networks, the number of nodes of output layer, hidden layer and its number of nodes are determined.
(3) hidden layer, output layer transmission function relationship and training function are determined.
(4) training network is fitted training sample, its mean square deviation is made to reach minimum.
(5) trained network model is examined to predict future using test sample.To training test and prediction
The result of output carries out renormalization processing, i.e., output valve is reduced to commercial weight guiding principle value.
Wherein, BP neural network is multilayer feedforward formula error backward propagation method, can simulate arbitrary non-linear input
Output relation.
To any neuron i, input/output relation can be described as：
Wherein, xj- j-th input, Yi- i-th neuron exports, wij- j-th power being connected with i-th of neuron
Value, θiThe threshold values of-i-th neuron.
In a specific example, in specific example, process and target to BP neural network training are：Training net
Network is fitted training sample, its mean square deviation is made to reach minimum.
Further, in an embodiment of the present invention, weighting system is obtained according to the quantity predicted for each order type
Number.It can be optimized in this way according to different scenes, the order of different category properties by different weights coefficient, tie prediction
Fruit is more accurate.Wherein weighting coefficient is calculated according to different application scene.
In specific example, forecasting effective measure algorithm, that is, optimal weighted method is substantially according to " combined prediction error in the past
Quadratic sum is minimum " principle seeks the weighting coefficient of each prediction model.Its process is as follows：
I-th kind of prediction model is in the prediction error e of t momentI, t；
eI, t=yt-yI, t,
Y in formulatThe actual value of-the t moment, t=1,2,3 ..., n；The e in the specific example of the present inventionI, tIt is i-th kind
The pre-generated quantity of such order and the error of actual value that prediction model is predicted in t moment.
The prediction error vector for establishing i-th kind of prediction model is：
Fi=[eI, 1, eI, 2..., eI, n]T,
Then error matrix e is：
E=[F1, F2..., Fr],
Obtain control information matrix ErFor：
Enable Rr=[1,1,1 ..., 1]TFor the r dimensional vectors of element power and position 1, W=[w1, w2..., wr]TFor r kind prediction models
Weight vectors.
Then the error sum of squares S of combined prediction is：
Then obtain linear programming：
MinS=WTErW,
Then its optimal solution is：
Since the optimal weighting coefficients vector W that above formula obtains is likely to occur negative component, in practical applications in order to avoid going out
Existing negative weight, increases nonnegativity restrictions, so as to find out non-negative optimal weighting coefficients, sees below formula：
MinS=WTErW,
W >=0,
In specific example, weighting coefficient W is obtained, to obtain subsequent prediction as a result, and above-mentioned calculating process do not repeat.
S4：By multiple prediction models and each prediction model, for each order type carry out quantitative forecast
Weighting coefficient predicts the quantity on order of any order type in preset time period, obtains prediction result.
Specifically, obtain each prediction model, for each order type carry out quantitative forecast weighting coefficient, can
It is predicted, and then obtain prediction result with the quantity on order to any order type in preset time period.In specific example
In, the data that three kinds of Forecasting Methodologies obtain are brought into weighting coefficient W, obtain final result.
Such as：Table 1 carries out above-mentioned for Bengbu replacement Gas Pipe order on April 29, -2017 years on the 23rd April in 2017 situation
The prediction result table of method, as shown in table 1：
Table 1
As can be seen from the table, using the forecasting effective measure algorithm of the present embodiment, optimal weighting coefficients can be calculated, are made
The error rate for obtaining prediction result is only 2.14%, is significantly smaller than the error of single prediction model institute prediction result, it was demonstrated that this implementation
The prediction result obtained by the weighting coefficient of the multiple models of synthesis of example can improve predictablity rate, be allowed to more use
Value.
In addition, user (such as service station) multiple prediction models can also be trained respectively according to prediction result with it is excellent
Change, obtain more accurately prediction result, and according to result arranged rational attendant, so as to cost-effective and client is given to provide
Better service.
The Forecasting Methodology of the order of the embodiment of the present invention, by the information of the History Order to being collected into History Order into
Row classification, establishes multiple prediction models, and obtains each prediction model, adding for each order type progress quantitative forecast
Weight coefficient, and then obtain according to the weighting coefficient of multiple prediction models any order type quantity on order in preset time period
Prediction result substantially increases the accuracy of prediction result, so as to reasonable arrangement attendant in advance, reach saving into
The purpose of more good service originally and to client is provided.
Fig. 2 is the forecasting system structure diagram of order according to embodiments of the present invention, as shown in Fig. 2, the prediction of order
System 10 includes：Acquisition module 101, sort module 102, weighting coefficient acquisition module 103 and prediction module 104.Next it is right
It is described in detail：
Acquisition module 101, for obtaining the data of History Order, wherein, the data of History Order include History Order
The sequence information of quantity and each History Order.
Sort module 102 for being classified according to the sequence information of each History Order to History Order, obtains at least
A kind of order type.
In specific example, sequence information includes order category, and sort module 102 is additionally operable to according to each History Order
Sequence information determines the property of each History Order；Classified according to the property of each History Order to History Order, and
To at least a kind of order type.
Wherein, the property of History Order includes periodical, sudden and relevance.
Weighting coefficient acquisition module 103, data and history therein for the History Order according to each order type
The sequence information of order, the multiple prediction models of training, with obtain each prediction model, for each order type carry out quantity
The weighting coefficient of prediction.
Wherein, multiple prediction models include gray scale prediction model, seasonal index smoothing model and neural network model.
In specific example, weighting coefficient acquisition module 103 is used for：It is directed to according to prediction model each in predetermined amount of time
The quantity of the order type and the actual quantity of the order type of the History Order prediction of each order type, determine each
The prediction error of prediction model；Error sum of squares is obtained and according to each prediction according to the prediction error of each prediction model
The error sum of squares of model determines the weighting coefficient of corresponding prediction model using optimal weighted method.
Prediction module 104, for pass through multiple prediction models and each prediction model, for each order type
The weighting coefficient of quantitative forecast is carried out, the quantity on order of any order type in preset time period is predicted, is obtained pre-
Survey result.
Further, it in a specific example of the forecasting system of order of the present invention, further includes：Model optimization module,
For multiple prediction models to be trained and are optimized respectively according to prediction result.
It should be noted that the explanation of the aforementioned Forecasting Methodology embodiment to order is also applied for the prediction of the order
The embodiment of system, details are not described herein again.
The forecasting system of the order of the embodiment of the present invention, by the information of the History Order to being collected into History Order into
Row classification, establishes multiple prediction models, and obtains each prediction model, adding for each order type progress quantitative forecast
Weight coefficient, and then obtain according to the weighting coefficient of multiple prediction models any order type quantity on order in preset time period
Prediction result substantially increases the accuracy of prediction result, so as to reasonable arrangement attendant in advance, reach saving into
The purpose of more good service originally and to client is provided.
In the description of the present invention, term " first ", " second " are only used for description purpose, and it is not intended that instruction or dark
Show relative importance or the implicit quantity for indicating indicated technical characteristic.The feature of " first ", " second " is defined as a result,
It can express or implicitly include at least one this feature.In the description of the present invention, " multiple " are meant that at least two,
Such as two, three etc., unless otherwise specifically defined.
In the present invention unless specifically defined or limited otherwise, the terms such as term " connected ", " connection ", " fixation " should
It is interpreted broadly, for example, it may be being fixedly connected or being detachably connected or integral；Can be mechanical connection,
It can be electrical connection；It can be directly connected, can also be indirectly connected by intermediary, can be the company inside two elements
Logical or two elements interaction relationship, unless otherwise restricted clearly.For the ordinary skill in the art, may be used
To understand the concrete meaning of above-mentioned term in the present invention as the case may be.
In the description of this specification, reference term " one embodiment ", " example ", " is specifically shown " some embodiments "
The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description
Point is contained at least one embodiment of the present invention or example.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It is combined in an appropriate manner in one or more embodiments or example.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the different embodiments or examples described in this specification and the feature of different embodiments or examples
It closes and combines.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is impossible to limitation of the present invention is interpreted as, those of ordinary skill in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (12)
1. a kind of Forecasting Methodology of order, which is characterized in that include the following steps：
The data of History Order are obtained, wherein, the data of the History Order include the quantity of the History Order and each go through
The sequence information of history order；
Classified according to the sequence information of each History Order to the History Order, obtain at least a kind of order class
Type；
The sequence information of the quantity of History Order based on each order type and History Order therein, the multiple prediction moulds of training
Type, with obtain each prediction model, for each order type carry out quantitative forecast weighting coefficient；
By the multiple prediction model and each prediction model, for each order type carry out quantity
The weighting coefficient of prediction predicts the quantity on order of any order type in preset time period, obtains prediction result.
2. the Forecasting Methodology of order according to claim 1, which is characterized in that described according to each History Order
Sequence information classifies to the History Order, obtains at least a kind of order type, including：
The property of each History Order is determined according to the sequence information；
Classified according to the property of each History Order to the History Order, and obtain at least a kind of order class
Type.
3. the Forecasting Methodology of order according to claim 2, which is characterized in that the property of each History Order includes
Periodical, sudden and relevance.
4. the Forecasting Methodology of order according to claim 1, which is characterized in that it is pre- that the multiple prediction model includes gray scale
Survey model, seasonal index smoothing model and neural network model.
5. according to the Forecasting Methodology of claim 1-4 any one of them orders, which is characterized in that described to be based on each order class
The sequence information of the quantity of the History Order of type and History Order therein, the multiple prediction models of training are each described to obtain
Prediction model, for each order type carry out quantitative forecast weighting coefficient, including：
The order type number predicted according to prediction model each in predetermined amount of time for the History Order of each order type
The actual quantity of amount and the order type determines the prediction error of each prediction model；
Error sum of squares is obtained according to the prediction error of each prediction model；
According to the error sum of squares of each prediction model, the weighting coefficient of corresponding prediction model is determined using optimal weighted method.
6. the Forecasting Methodology of order according to claim 1, which is characterized in that further include：According to the prediction result point
It is other that the multiple prediction model is trained and optimized.
7. a kind of forecasting system of order, which is characterized in that including：
Acquisition module, for obtaining the data of History Order, wherein, the data of the History Order include the History Order
The sequence information of quantity and each History Order；
Sort module, for being classified according to the sequence information of each History Order to the History Order, obtain to
Few a kind of order type；
Weighting coefficient acquisition module, for the quantity of the History Order according to each order type and ordering for History Order therein
Single information, the multiple prediction models of training, with obtain each prediction model, for each order type carry out quantitative forecast
Weighting coefficient；
Prediction module, for pass through the multiple prediction model and each prediction model, for each order
Type carries out the weighting coefficient of quantitative forecast, and the quantity on order of any order type in preset time period is predicted, is obtained
To prediction result.
8. the forecasting system of order according to claim 7, which is characterized in that the sort module is further used for：
The property of each History Order is determined according to the sequence information of each History Order；
Classified according to the property of each History Order to the History Order, and obtain at least a kind of order class
Type.
9. the forecasting system of order according to claim 8, which is characterized in that the property of each History Order includes
Periodical, sudden and relevance.
10. the forecasting system of order according to claim 7, which is characterized in that the multiple prediction model includes gray scale
Prediction model, seasonal index smoothing model and neural network model.
11. according to the forecasting system of claim 7-10 any one of them orders, which is characterized in that the weighting coefficient obtains
Module is further used for：
According to prediction model each in predetermined amount of time for the order type of the History Order prediction of each order type
The actual quantity of quantity and the order type determines the prediction error of each prediction model；
Error sum of squares and the square-error according to each prediction model are obtained according to the prediction error of each prediction model
With the weighting coefficient of corresponding prediction model is determined using optimal weighted method.
12. the forecasting system of order according to claim 7, which is characterized in that further include：Model optimization module, is used for
The multiple prediction model is trained and optimized respectively according to the prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810136485.5A CN108230049A (en) | 2018-02-09 | 2018-02-09 | The Forecasting Methodology and system of order |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810136485.5A CN108230049A (en) | 2018-02-09 | 2018-02-09 | The Forecasting Methodology and system of order |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108230049A true CN108230049A (en) | 2018-06-29 |
Family
ID=62661442
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810136485.5A Pending CN108230049A (en) | 2018-02-09 | 2018-02-09 | The Forecasting Methodology and system of order |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108230049A (en) |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636013A (en) * | 2018-11-27 | 2019-04-16 | 拉扎斯网络科技（上海）有限公司 | Dispense generation method, device, electronic equipment and the storage medium of range |
| CN109816442A (en) * | 2019-01-16 | 2019-05-28 | 四川驹马科技有限公司 | A kind of various dimensions freight charges prediction technique and its system based on feature tag |
| CN110084438A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110109800A (en) * | 2019-04-10 | 2019-08-09 | 网宿科技股份有限公司 | A kind of management method and device of server cluster system |
| CN110188926A (en) * | 2019-05-10 | 2019-08-30 | 重庆天蓬网络有限公司 | A kind of order information forecasting system and method |
| CN110766184A (en) * | 2018-07-25 | 2020-02-07 | 北京京东尚科信息技术有限公司 | Order quantity prediction method and device |
| CN110956480A (en) * | 2018-09-26 | 2020-04-03 | 北京嘀嘀无限科技发展有限公司 | User structure estimation method and device and server |
| CN111046909A (en) * | 2019-11-05 | 2020-04-21 | 新奥数能科技有限公司 | Load prediction method and device |
| CN111125951A (en) * | 2019-12-16 | 2020-05-08 | 新奥数能科技有限公司 | Optimization method and device of evaporator scaling prediction model |
| CN111242747A (en) * | 2020-01-21 | 2020-06-05 | 北京爱数智慧科技有限公司 | Order quotation method and device |
| CN111260101A (en) * | 2018-11-30 | 2020-06-09 | 北京嘀嘀无限科技发展有限公司 | Information processing method and device |
| CN111260424A (en) * | 2018-11-30 | 2020-06-09 | 北京嘀嘀无限科技发展有限公司 | Information processing method and device |
| CN111292105A (en) * | 2018-12-06 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Service demand determination method and device |
| CN111292106A (en) * | 2018-12-06 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Method and device for determining business demand influence factors |
| WO2020124977A1 (en) * | 2018-12-19 | 2020-06-25 | 平安科技（深圳）有限公司 | Method and apparatus for processing production data, computer device, and storage medium |
| CN111612197A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Security event order detection method and device and electronic equipment |
| WO2020186380A1 (en) * | 2019-03-15 | 2020-09-24 | State Street Corporation | Techniques to forecast future orders using deep learning |
| CN111768093A (en) * | 2020-06-23 | 2020-10-13 | 中国工商银行股份有限公司 | Network point configuration method and device based on time sequence |
| CN112561551A (en) * | 2019-09-26 | 2021-03-26 | 富士通株式会社 | Method, apparatus, and storage medium for optimizing object prediction |
| CN112583610A (en) * | 2019-09-27 | 2021-03-30 | 顺丰科技有限公司 | System state prediction method, system state prediction device, server and storage medium |
| CN112749821A (en) * | 2019-10-29 | 2021-05-04 | 顺丰科技有限公司 | Express delivery quantity prediction method and device, computer equipment and storage medium |
| CN112990624A (en) * | 2019-12-13 | 2021-06-18 | 顺丰科技有限公司 | Task allocation method, device, equipment and storage medium |
| CN113326983A (en) * | 2021-05-28 | 2021-08-31 | 重庆能源大数据中心有限公司 | Natural gas consumption prediction system and method |
| CN113393279A (en) * | 2021-07-08 | 2021-09-14 | 北京沃东天骏信息技术有限公司 | Order quantity estimation method and system |
| CN113657916A (en) * | 2020-05-12 | 2021-11-16 | 北京市天元网络技术股份有限公司 | Method and device for acquiring budget of buried pipe |
| CN113781156A (en) * | 2021-05-13 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Malicious order recognition method, malicious order model training method, malicious order recognition equipment and malicious order model training storage medium |
| CN116467645A (en) * | 2023-06-20 | 2023-07-21 | 中通服建设有限公司 | Pollution source collection monitoring system |
| CN116882598A (en) * | 2023-09-08 | 2023-10-13 | 四川丝路易购科技有限公司 | Import and export goods trade order management method and system |
| CN117436709A (en) * | 2023-12-20 | 2024-01-23 | 四川宽窄智慧物流有限责任公司 | Cross-region order data overall warning method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184678A (en) * | 2015-09-18 | 2015-12-23 | 齐齐哈尔大学 | Method for constructing photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms |
| CN106408341A (en) * | 2016-09-21 | 2017-02-15 | 北京小米移动软件有限公司 | Goods sales volume prediction method and device, and electronic equipment |
| CN106971249A (en) * | 2017-05-05 | 2017-07-21 | 北京挖玖电子商务有限公司 | A kind of Method for Sales Forecast and replenishing method |
| CN107220764A (en) * | 2017-05-25 | 2017-09-29 | 北京中电普华信息技术有限公司 | A kind of electricity sales amount Forecasting Methodology compensated based on preamble analysis and factor and device |
| CN107515898A (en) * | 2017-07-22 | 2017-12-26 | 复旦大学 | Based on data diversity and the multifarious tire enterprise sales forecasting method of task |
| US20180005253A1 (en) * | 2016-06-30 | 2018-01-04 | International Business Machines Corporation | Revenue prediction for a sales pipeline using optimized weights |
| CN107590569A (en) * | 2017-09-25 | 2018-01-16 | 山东浪潮云服务信息科技有限公司 | A kind of data predication method and device |
-
2018
- 2018-02-09 CN CN201810136485.5A patent/CN108230049A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184678A (en) * | 2015-09-18 | 2015-12-23 | 齐齐哈尔大学 | Method for constructing photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms |
| US20180005253A1 (en) * | 2016-06-30 | 2018-01-04 | International Business Machines Corporation | Revenue prediction for a sales pipeline using optimized weights |
| CN106408341A (en) * | 2016-09-21 | 2017-02-15 | 北京小米移动软件有限公司 | Goods sales volume prediction method and device, and electronic equipment |
| CN106971249A (en) * | 2017-05-05 | 2017-07-21 | 北京挖玖电子商务有限公司 | A kind of Method for Sales Forecast and replenishing method |
| CN107220764A (en) * | 2017-05-25 | 2017-09-29 | 北京中电普华信息技术有限公司 | A kind of electricity sales amount Forecasting Methodology compensated based on preamble analysis and factor and device |
| CN107515898A (en) * | 2017-07-22 | 2017-12-26 | 复旦大学 | Based on data diversity and the multifarious tire enterprise sales forecasting method of task |
| CN107590569A (en) * | 2017-09-25 | 2018-01-16 | 山东浪潮云服务信息科技有限公司 | A kind of data predication method and device |
Non-Patent Citations (1)
| Title |
|---|
| 李俊 等: "基于组合预测的商品销售量预测方法", 《统计与决策》 * |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110766184A (en) * | 2018-07-25 | 2020-02-07 | 北京京东尚科信息技术有限公司 | Order quantity prediction method and device |
| CN110956480A (en) * | 2018-09-26 | 2020-04-03 | 北京嘀嘀无限科技发展有限公司 | User structure estimation method and device and server |
| CN109636013A (en) * | 2018-11-27 | 2019-04-16 | 拉扎斯网络科技（上海）有限公司 | Dispense generation method, device, electronic equipment and the storage medium of range |
| CN109636013B (en) * | 2018-11-27 | 2021-06-29 | 拉扎斯网络科技（上海）有限公司 | Distribution range generation method and device, electronic equipment and storage medium |
| CN111260101B (en) * | 2018-11-30 | 2022-07-08 | 北京嘀嘀无限科技发展有限公司 | Information processing method and device |
| CN111260424A (en) * | 2018-11-30 | 2020-06-09 | 北京嘀嘀无限科技发展有限公司 | Information processing method and device |
| CN111260101A (en) * | 2018-11-30 | 2020-06-09 | 北京嘀嘀无限科技发展有限公司 | Information processing method and device |
| CN111292105A (en) * | 2018-12-06 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Service demand determination method and device |
| CN111292105B (en) * | 2018-12-06 | 2023-12-08 | 北京嘀嘀无限科技发展有限公司 | Service demand determining method and device |
| CN111292106A (en) * | 2018-12-06 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Method and device for determining business demand influence factors |
| WO2020124977A1 (en) * | 2018-12-19 | 2020-06-25 | 平安科技（深圳）有限公司 | Method and apparatus for processing production data, computer device, and storage medium |
| CN109816442A (en) * | 2019-01-16 | 2019-05-28 | 四川驹马科技有限公司 | A kind of various dimensions freight charges prediction technique and its system based on feature tag |
| CN111612197A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Security event order detection method and device and electronic equipment |
| WO2020186380A1 (en) * | 2019-03-15 | 2020-09-24 | State Street Corporation | Techniques to forecast future orders using deep learning |
| CN110109800A (en) * | 2019-04-10 | 2019-08-09 | 网宿科技股份有限公司 | A kind of management method and device of server cluster system |
| CN110084438A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110188926A (en) * | 2019-05-10 | 2019-08-30 | 重庆天蓬网络有限公司 | A kind of order information forecasting system and method |
| CN112561551A (en) * | 2019-09-26 | 2021-03-26 | 富士通株式会社 | Method, apparatus, and storage medium for optimizing object prediction |
| CN112583610B (en) * | 2019-09-27 | 2023-04-11 | 顺丰科技有限公司 | System state prediction method, system state prediction device, server and storage medium |
| CN112583610A (en) * | 2019-09-27 | 2021-03-30 | 顺丰科技有限公司 | System state prediction method, system state prediction device, server and storage medium |
| CN112749821A (en) * | 2019-10-29 | 2021-05-04 | 顺丰科技有限公司 | Express delivery quantity prediction method and device, computer equipment and storage medium |
| CN112749821B (en) * | 2019-10-29 | 2024-03-05 | 顺丰科技有限公司 | Express delivery quantity prediction method, device, computer equipment and storage medium |
| CN111046909A (en) * | 2019-11-05 | 2020-04-21 | 新奥数能科技有限公司 | Load prediction method and device |
| CN112990624A (en) * | 2019-12-13 | 2021-06-18 | 顺丰科技有限公司 | Task allocation method, device, equipment and storage medium |
| CN111125951A (en) * | 2019-12-16 | 2020-05-08 | 新奥数能科技有限公司 | Optimization method and device of evaporator scaling prediction model |
| CN111125951B (en) * | 2019-12-16 | 2023-11-03 | 新奥数能科技有限公司 | Optimization method and device for evaporator scaling prediction model |
| CN111242747A (en) * | 2020-01-21 | 2020-06-05 | 北京爱数智慧科技有限公司 | Order quotation method and device |
| CN113657916B (en) * | 2020-05-12 | 2024-01-26 | 北京市天元网络技术股份有限公司 | Method and device for obtaining budget of buried pipe |
| CN113657916A (en) * | 2020-05-12 | 2021-11-16 | 北京市天元网络技术股份有限公司 | Method and device for acquiring budget of buried pipe |
| CN111768093A (en) * | 2020-06-23 | 2020-10-13 | 中国工商银行股份有限公司 | Network point configuration method and device based on time sequence |
| CN113781156A (en) * | 2021-05-13 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Malicious order recognition method, malicious order model training method, malicious order recognition equipment and malicious order model training storage medium |
| CN113326983A (en) * | 2021-05-28 | 2021-08-31 | 重庆能源大数据中心有限公司 | Natural gas consumption prediction system and method |
| CN113393279A (en) * | 2021-07-08 | 2021-09-14 | 北京沃东天骏信息技术有限公司 | Order quantity estimation method and system |
| CN116467645A (en) * | 2023-06-20 | 2023-07-21 | 中通服建设有限公司 | Pollution source collection monitoring system |
| CN116467645B (en) * | 2023-06-20 | 2023-09-19 | 中通服建设有限公司 | Pollution source collection monitoring system |
| CN116882598A (en) * | 2023-09-08 | 2023-10-13 | 四川丝路易购科技有限公司 | Import and export goods trade order management method and system |
| CN116882598B (en) * | 2023-09-08 | 2023-12-29 | 四川丝路易购科技有限公司 | Import and export goods trade order management method and system |
| CN117436709A (en) * | 2023-12-20 | 2024-01-23 | 四川宽窄智慧物流有限责任公司 | Cross-region order data overall warning method |
| CN117436709B (en) * | 2023-12-20 | 2024-03-19 | 四川宽窄智慧物流有限责任公司 | Cross-region order data overall warning method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108230049A (en) | The Forecasting Methodology and system of order | |
| CN105117602B (en) | A kind of metering device running status method for early warning | |
| CN106533750A (en) | System and method for predicting non-steady application user concurrency in cloud environment | |
| CN109919356B (en) | BP neural network-based interval water demand prediction method | |
| CN107016469A (en) | Methods of electric load forecasting | |
| CN109389238B (en) | Ridge regression-based short-term load prediction method and device | |
| CN106980910B (en) | Medium-and-long-term power load measuring and calculating system and method | |
| CN110826855A (en) | Method and system for testing network access performance of intelligent power distribution room state monitoring sensor | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| CN104636874A (en) | Method and equipment for detecting business exception | |
| CN112686481A (en) | Runoff forecasting method and processor | |
| Majumder et al. | Real time reliability monitoring of hydro‐power plant by combined cognitive decision‐making technique | |
| CN113449257A (en) | Power distribution network line loss prediction method, control device, and storage medium | |
| CN115545333A (en) | Method for predicting load curve of multi-load daily-type power distribution network | |
| CN116451556A (en) | Construction method of concrete dam deformation observed quantity statistical model | |
| CN112149976B (en) | Power grid accurate investment project decision method based on artificial intelligence | |
| CN114254833A (en) | Reservoir water level prediction and scheduling method based on multiple linear regression and meteorological data | |
| CN116960962A (en) | Mid-long term area load prediction method for cross-area data fusion | |
| CN111967684A (en) | Metering asset active distribution method based on big data analysis | |
| CN106408119A (en) | Large power grid load prediction method based on comprehensive evaluation of analog-meteorological subareas | |
| CN110110944A (en) | One kind predicting wind electricity digestion demand quantity algorithm based on modified wavelet neural network | |
| CN109993374A (en) | Car loading prediction technique and device | |
| CN105117605B (en) | A kind of device and method of case prediction | |
| CN109190830B (en) | Energy demand prediction method based on empirical decomposition and combined prediction | |
| CN106447091A (en) | Regional meteorological condition similarity-based large power network load prediction method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180629 |