CN107316093A - The method and device of a kind of rolling forecast - Google Patents
The method and device of a kind of rolling forecast Download PDFInfo
- Publication number
- CN107316093A CN107316093A CN201610266574.2A CN201610266574A CN107316093A CN 107316093 A CN107316093 A CN 107316093A CN 201610266574 A CN201610266574 A CN 201610266574A CN 107316093 A CN107316093 A CN 107316093A
- Authority
- CN
- China
- Prior art keywords
- features
- combinations
- sequence
- rolling forecast
- target signature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Abstract
The embodiment of the invention discloses a kind of method of rolling forecast and device, the realization of wherein method includes：Determine the lag period of the hysteresis quality feature of each sequence in reference sequences；Included in the reference sequences and external sequence and/or its own sequence are included at least two sequences, the reference sequences；According to the lag period of the hysteresis quality feature of each sequence in the reference sequences, the hysteresis quality feature to each sequence in the reference sequences is reordered, and the hysteresis quality feature that the same period is belonged to after rearrangement is combined and obtains combinations of features to be selected；Combinations of features is selected to carry out rolling forecast to the checking collection of known results from the combinations of features to be selected, the combinations of features that selection rolling forecast result conforms to a predetermined condition is combined as target signature；Rolling forecast is carried out to its own sequence using target signature combination.Influence of its own sequence/external sequence to rolling forecast result is considered, rolling forecast result is more accurately and stably.
Description
Technical field
The present invention relates to field of computer technology, the method and device of more particularly to a kind of rolling forecast.
Background technology
Often there are some indexs for weighing itself management state in enterprise, such as：Prime operating revenue, profit etc..These indexs with
Mutually data in order before and after time cumulation formation and having in time, these data belong to sequence data, these data when
Between on have before and after interdependent order referred to as time series.For the traffic-operating period of enterprise, people are merely not only to be concerned about currently to refer to
Mark, can be also thoughtful of the future these index possible values, and good strategy preparation is done in advance, evade following possible risk or Instructing manufacture
The decision-making of operation.
Under normal circumstances, people can also want to know about future except in the case of the period These parameters that are thoughtful of the future
The situation that this refers to target value in longer time, such as：Monthly income value, the Revenue that business administration people is thoughtful of the future 1 month, also can
Want to know about following 12 months monthly income value.Therefore the rolling forecast of time series many phases forward is widely used in enterprise
In.
The method studied in educational circles is predicted usually using the rule of its own sequence.For a certain index A itself sequence
Row, refer to the sequence of the A itself historical data composition.Comparatively, external sequence is then going through for other indexs outside the A
The sequence of history data composition.Generally its own sequence can be referred to as target sequence, on the basis of target sequence, other sequences are referred to as
External sequence.
With attention of the enterprise for internal data, more and more external sequences are collected and used the rolling of target sequence
There is certain lag correlation in the data of prediction, these external sequences and target sequence；Shipment data meeting before such as 3 months
The income in this month is had influence on, is referred to as the lag period within this 3 months；Therefore suitable lag period external sequence is selected to improve prediction
Accuracy.
The implementation of rolling forecast is carried out the following is connected applications external sequence and its own sequence, is specifically included：
First, the lag period is determined：
Correlation of the reference sequences (comprising target sequence and external sequence) between target sequence is calculated, selection is related
Property highest lag period t, determines whether correlation notable by correlation test, if correlation significantly if be defined as having hysteresis quality
Influence and the lag period is t.Wherein, for the lag period of target sequence itself, auto-correlation function can be used, when obtaining different
The autocorrelation value of phase, by correlation test confirm correlation it is whether notable, if correlation significantly if determine that corresponding phase
For the lag period, otherwise in the absence of the lag period, i.e.,：Lag period is 0.
2nd, model is set up：
Suitable model is set up according to the lag period between reference sequences and target sequence, for example：Multiple linear regression mould
Type.
3rd, rolling forecast:
The feature for inputting above-mentioned target sequence carries out rolling forecast, when required surface missing, can first predict
The characteristic value of external sequence, then inputs this feature value and carries out rolling forecast.
Rolling forecast is carried out using above scheme, the relatively low stability of accuracy is relatively low.
The content of the invention
The embodiments of the invention provide a kind of method of rolling forecast and device, the accuracy for improving rolling forecast
And stability.
On the one hand the embodiments of the invention provide a kind of method of rolling forecast, including：
Determine the lag period of the hysteresis quality feature of each sequence in reference sequences；At least two sequences are included in the reference sequences
External sequence and/or its own sequence are included in row, the reference sequences；
According to the lag period of the hysteresis quality feature of each sequence in the reference sequences, to each sequence in the reference sequences
Hysteresis quality feature is reordered, and the hysteresis quality feature that the same period is belonged to after rearrangement is combined and obtains feature group to be selected
Close；
Select combinations of features to carry out rolling forecast to the checking collection of known results from the combinations of features to be selected, choose rolling
The dynamic combinations of features conformed to a predetermined condition that predicts the outcome is combined as target signature；
Rolling forecast is carried out to its own sequence using target signature combination.
Reference sequences can only have external sequence, can also there was only its own sequence, preferably simultaneously using external sequence and
Its own sequence.Checking collection is the sequence of knowledge of result, is predicted which is known that to it using combinations of features to be selected
Preferably, other are then relatively poor for combinations of features prediction effect；Here it is preferable in order to select prediction effect to conform to a predetermined condition
Combinations of features.
It is described to select combinations of features from the combinations of features to be selected in a kind of possible implementation, including：
In the way of traversal combinations of features is selected from the combinations of features to be selected；Or, in a random way from described
Combinations of features is selected in combinations of features to be selected；Or, selected with traversal or random manner from the combinations of features to be selected
Hysteresis quality feature and at least one its own sequence comprising at least one external sequence in combinations of features, and the combinations of features of selection
Hysteresis quality feature.
Three of the above selection combinations of features scheme in, wherein the first amount of calculation is larger, but advantage be it is more complete
Face, therefore in computing capability compared with strong or preferred implementation scheme can be used as in the case that computing resource is more；Second of calculating
Amount is relatively small, can improve computational efficiency；The third advantage is while considering external sequence also its own sequence to rolling
The dynamic influence predicted the outcome.
In a kind of possible implementation, the target signature combination includes at least two features conformed to a predetermined condition
Combination；Before rolling forecast is carried out to its own sequence using target signature combination, methods described also includes：
Selected characteristic combination obtains objective cross to be determined from the combinations of features conformed to a predetermined condition；
Rolling forecast is carried out to the checking collection of the known results using the objective cross to be determined, it is predetermined meeting
The objective cross to be determined is combined as the target signature during stop condition.
The present embodiment has been further combined combinations of features, and this contributes to the Stability and veracity for further improving prediction.
Alternatively, it is also possible to carry out certain screening to combinations of features.
In a kind of possible implementation, the predetermined stop condition includes：
New selection combinations of features is added to the objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, the number of times for having lifting to be chosen in the case of predicting the outcome to same combinations of features reaches predetermined number of times.
In three of the above stop condition, the first is optimal for target to predict the outcome, and second of control is the simplest, smaller
Pre-determined number can prevent over-fitting；The third control is also relatively simple, when being repeatedly drawn into same combinations of features
When stop selection, it may be determined that other combinations of features can not all improve prediction effect；Here the specific number of times of " multiple " can be pre-
First set, such as：2nd, 5 or other values.
In a kind of possible implementation, at least two combinations of features are included in the target signature combination, it is described to make
Carrying out rolling forecast to its own sequence with target signature combination includes：
Each combinations of features in being combined using the target signature carries out rolling forecast to its own sequence respectively, then
Calculate the weighted average of the result of rolling forecast.
The embodiments of the invention provide a kind of rolling forecast device in terms of two, including：
Lag period determining unit, the lag period of the hysteresis quality feature for determining each sequence in reference sequences；The reference
Included in sequence and external sequence and/or its own sequence are included at least two sequences, the reference sequences；
Rearrangement units, for the lag period according to the hysteresis quality feature of each sequence in the reference sequences, to the reference
The hysteresis quality feature of each sequence is reordered in sequence；
Assembled unit, feature group to be selected is obtained for the hysteresis quality feature for belonging to the same period after rearrangement to be combined
Close；
Feature selection unit, for selected from the combinations of features to be selected checking of the combinations of features to known results collect into
Row rolling forecast, the combinations of features that selection rolling forecast result conforms to a predetermined condition is combined as target signature；
Rolling forecast unit, for carrying out rolling forecast to its own sequence using target signature combination.
In a kind of possible implementation, the feature selection unit, in the way of traversal from the spy to be selected
Levy in combination and select combinations of features；Or, select combinations of features from the combinations of features to be selected in a random way；Or,
Selected with traversal or random manner from the combinations of features to be selected in combinations of features, and the combinations of features of selection comprising extremely
The hysteresis quality feature of a few external sequence and the hysteresis quality feature of at least one its own sequence.
In a kind of possible implementation, the target signature combination includes at least two features conformed to a predetermined condition
Combination；Described device also includes：
Selecting unit is combined, for being combined in the rolling forecast unit using the target signature to its own sequence
Carry out before rolling forecast, selected characteristic combination obtains target group to be determined from the combinations of features conformed to a predetermined condition
Close；
The rolling forecast unit, be also used for checking of the objective cross to be determined to the known results collect into
Row rolling forecast；
The feature selection unit, for carrying out meeting predetermined stop during rolling forecast in the rolling forecast unit
Only the objective cross to be determined is combined as the target signature during condition.
In a kind of possible implementation, the predetermined stop condition includes：
New selection combinations of features is added to the objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, new selection combinations of features is added to the objective cross to be determined and predicted the outcome the feelings that lifting predicts the outcome
Under condition, the number of times chosen to same combinations of features reaches predetermined number of times.
In a kind of possible implementation, at least two combinations of features are included in the target signature combination；
The rolling forecast unit, for using the target signature combine in each combinations of features respectively to it is described itself
Sequence carries out rolling forecast, then calculates the weighted average of the result of rolling forecast.
As can be seen from the above technical solutions, the embodiment of the present invention has advantages below：At least two sequences have been used to make
For reference, comprising a variety of hysteresis quality features, different hysteresis quality features its lag period of correspondence, by resetting and combinations of features
It is determined that, consider influence of its own sequence/external sequence to rolling forecast result, rolling forecast result is more accurate and steady
It is fixed.
Brief description of the drawings
Technical scheme in order to illustrate the embodiments of the present invention more clearly, makes required in being described below to embodiment
Accompanying drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this
For the those of ordinary skill in field, without having to pay creative labor, it can also be obtained according to these accompanying drawings
His accompanying drawing.
Fig. 1 is system structure diagram of the embodiment of the present invention；
Fig. 2 is integrated prediction modular structure schematic diagram of the embodiment of the present invention；
Fig. 3 is present invention method schematic flow sheet；
Fig. 4 is apparatus structure schematic diagram of the embodiment of the present invention；
Fig. 5 is apparatus structure schematic diagram of the embodiment of the present invention；
Fig. 6 is apparatus structure schematic diagram of the embodiment of the present invention；
Fig. 7 is server architecture schematic diagram of the embodiment of the present invention.
Embodiment
In order that the object, technical solutions and advantages of the present invention are clearer, below in conjunction with accompanying drawing the present invention is made into
One step it is described in detail, it is clear that described embodiment is only embodiment of the invention a part of, rather than whole implementation
Example.Based on the embodiment in the present invention, what those of ordinary skill in the art were obtained under the premise of creative work is not made
All other embodiment, belongs to the scope of protection of the invention.
In addition, using the rolling forecast scheme in background technology, being carried out in application external sequence to target sequence many forward
During phase rolling forecast, the problem of missing of following many phase surfaces can be run into, for example establish the shipment data before 3 months with
When the relational model of monthly income, when needing rolling forecast 4th month, of that month delivering amount may not be determined and can not answered
With for two problems above-mentioned, presently, there are some technologies can solve, but there are still certain defect.
The problem of embodiment of the present invention will be solved is how to utilize its own sequence feature and outside hysteresis quality sequence signature pair
The joint effect predicted in target sequence；Furthermore it is also possible to further avoid the influence of following surface missing, realize to
The rolling forecast of preceding many phases.
The problem of for m phase rolling forecasts forward, the embodiment of the present invention can set the minimum lag phase respectively from 1 to the m phases,
Hysteresis quality feature to external sequence and its own sequence is calculated, and is re-started combinations of features, is obtained forecast model, by preceding
The integrated multiple built-up patterns of filtering model Integrated Strategy, at the same consider its own sequence hysteresis quality feature and external sequence it is delayed
Property combinations of features for predicting the influence of target sequence, then different predicting the outcome for minimum lag phase (1~m) are rolled forward
Dynamic prediction is integrated, the final rolling forecast for realizing the m phases forward.The technical scheme of the embodiment of the present invention is mainly comprising following several sides
Face：
First, hysteresis quality feature set is generated, and feature set is combined；Generate the feature group of its own sequence and external sequence
Close.
2nd, filtering model Integrated Strategy before proposing, so as to realize that the multiple its own sequence features of consideration are combined with surface
Influence；
3rd, pass through the rolling forecast of different minimum lag phases, it is to avoid the problem of following surface is lacked, thus realize to
The rolling forecast of preceding many phases.
Based on described above, refer to shown in Fig. 1, including：5 modules, i.e.,：Retarding characteristic generation module, combinations of features
Generation module, forecast model module, integrated prediction module and rolling forecast integration module.
1st, retarding characteristic generation module：
The minimum lag phase k that retarding characteristic generation module is mainly determined according to early stage, by its own sequence of input and outside
Sequence, generation hysteresis quality feature is alternatively gathered, and the retarding characteristic generation module is to be directed to faced hysteresis quality feature, will be inputted
Its own sequence and external sequence pre-processed, the data of its own sequence and external sequence are converted into required for subsequent module
Data.
2nd, combinations of features generation module：
The main hysteresis quality feature in retarding characteristic generation module of combinations of features generation module is alternatively gathered, carry out group
Close, generation feature group intersection (set of combinations of features) is predicted analysis, so as to realize while examining in units of combinations of features
Surface and unique characteristics built-up pattern are considered to the joint effect that predicts the outcome.
3rd, forecast model module：
Forecast model module is that combinations of features is input among forecast model module, is exported and predicted according to corresponding model
As a result, conventional model can be linear regression, time series models etc..
4th, integrated prediction module:
Integrated prediction module is mainly the prediction effect for considering how to be concentrated in checking according to combinations of features model, and selection is closed
Suitable combinations of features model is used for following prediction, so that final prediction effect is more stablized.
5th, rolling forecast integration module：
Rolling forecast module is mainly the prediction output for being summarised in different minimum lag phases k, and minimum lag phase k is from 1 to mesh
Mark m, can export 1 phase predicting the outcome to the m phases forward forward respectively, integrated these predict the outcome can just export it is final to
Preceding many phases predict the outcome.Be 1~12 in the k values shown in Fig. 1, minimum lag phase specific value in different its own sequences and
Difference is had in the application example of external sequence, the citing should not be construed as the restriction to k spans.
In subsequent embodiment, the side of implementing of five nucleus modules more than in the embodiment of the present invention will be described in
Case, five nucleus modules of the above include：Retarding characteristic generation module, combinations of features generation module, integrated prediction module is rolled pre-
Survey integration module.
First, retarding characteristic generation module：
Retarding characteristic generation module is that the original sequence data of input is converted into retarding characteristic alternatively to collect.Key step has
3 step below:
1st, minimum lag phase k and maximum delayed issue q is determined；
2nd, the lag period is generated respectively for k, k+1 for each sequence ..., k+q-1 hysteresis quality feature；
3rd, the lag period feature of all sequences is reordered, generates alternative features collection.
It is overall process such as table 1~2 below, and shown in formula (1):
Table 1
After input delay feature k and q, it is transformed to shown in table 2：
Table 2
Hysteresis quality feature rearranges sequence, such as formula (1)：
Wherein
In being illustrated more than, single hysteresis quality feature generating mode is as follows:Original series (can be that external sequence can also be
Its own sequence) beTarget sequence isThe characteristic sequence of a delayed phase with
The corresponded manner of target sequence is as follows：The characteristic sequence of a so delayed phase is:
As shown in 3~table of subscript 5：
Table 3
Table 4
Table 5
Hysteresis quality feature delayed 1 phase of the wherein first row of table 4 obtains table 5.
2nd, combinations of features generation module：
Combinations of features generation module is reconfigured hysteresis quality feature set generated above, obtains combinations of features, i.e.,
Selection obtains combinations of features from the items of formula (1), is handled in units of combinations of features, so as to realize while considering outside
Sequence and influence of its own sequence to finally predicting.In general, 2 to 5 hysteresis quality features can be selected as a feature
Combination, the strategy that selection hysteresis quality feature is combined mainly has following three kinds：
1st, ergodic strategy：I.e. for the individual hysteresis quality features of q* (p+1) above, n hysteresis quality feature of selection of traversal
For a combination.This tactful advantage is that the combination considered is more comprehensive, when without the concern for calculating speed
It is contemplated that using the strategy；
N is uncertain in ergodic strategy, and value minimum can be 1, and maximum can be the sum of hysteresis quality feature
Amount.
2nd, randomness strategy：The feature group that random (taking-up is put back to or do not put back to) M size is n is alternatively concentrated from feature
Close.This tactful advantage is can to select less combinations of features number, improves computational efficiency.
Randomness strategy is that randomness strategy defines that the hysteresis quality of selection is special compared to being distinguished for ergodic strategy
Number n is levied, and defines the number M of combinations of features.
3rd, external sequence and its own sequence combined strategy:It is n's that traversal is alternatively concentrated from feature or size is randomly selected
Combinations of features, wherein there is n1>0 external sequence (i.e. x), has n2>0 its own sequence is (i.e. y)；This tactful advantage is institute
The combination of consideration necessarily contains external sequence and its own sequence feature simultaneously.
External sequence is with its own sequence combined strategy compared to first two strategy, and difference is to limit must in combinations of features
The hysteresis quality feature of external sequence and internal sequence must be included simultaneously.
3rd, forecast model module：
Combinations of features is input among forecast model module, predicted the outcome according to the output of corresponding model.
4th, integrated prediction module：
Integrated prediction module is mainly the output result of integrated above multiple built-up patterns (combinations of features i.e. hereinbefore),
Predictablity rate is improved, while making prediction more stablize.It is mainly in view of two points：
The otherness of built-up pattern each model in the prediction effect and built-up pattern storehouse that checking is concentrated.
Based on above-mentioned two point, filtering forward model selection Integrated Strategy is used for integrated before the embodiment of the present invention is proposed.It is main
There are following two aspects, as shown in Figure 2：
1st, based on the built-up pattern obtained, such as M obtained built-up pattern is trained in randomness strategy, to checking
Collection carries out rolling forecast, chooses the preferably preceding T built-up pattern of rolling forecast effect, generation alternative model storehouse.
2nd, select built-up pattern into Integrated Model Base Systems from alternative model storehouse, the built-up pattern of selection needs to meet integrated
The prediction effect of checking collection is increased in model library, is set up until stop condition and then stops continuing to select.Therefore, integrated mould
Type storehouse can include multiple built-up patterns.
Stop condition can have the following two kinds：
1., at most choose q times；Stop selection after selecting q times.Less q values can prevent over-fitting.
2. selection, is stopped when being repeatedly drawn into same built-up pattern, it may be determined that other built-up patterns can not all be carried
Highly integrated prediction effect.Here the specific number of times of " multiple " can be preset, such as：2nd, 5 or other values.
5th, rolling forecast integration module：
Based on above four modules, the Integrated Model Base Systems obtained can use Integrated Model Base Systems to enter target sequence
Row rolling forecast, it is specific as follows：
Rolling forecast integration module is mainly by corresponding different minimum lag phases k of each hysteresis quality feature in Integrated Model Base Systems
The progress that predicts the outcome collect it is integrated, so as to realize the rolling forecast of many phases (m phases) forward.
Assuming that current have been observed that time point for t, that is, yt, z1, t, z2, t... zQ* (p+1), tPreceding all presence, pin
For each minimum lag phase k, it is output as the rolling forecast of k phases forward, and target and what the different minimum lag phases determine
Relation between the hysteresis quality feature of sample, such as：The minimum lag phase is 1, then the relation that built-up pattern is set up is current target
ytWith upper phase zI, t-jRelation between (j >=1), then predicting next phase target yt+1When, the feature used is：
zI, t-j+1, because t-j+1≤t, zI, t-j+1So certainly exist.
Likewise, when the minimum lag phase is m, the target y of h (h≤m) phasest+hIt is used to be characterized as zI, t-j+h, its
Middle j >=m, so t-j+h≤t, is present.As shown in table 6 below and table 7:
Table 6
Predicting the outcome when table 6 show minimum lag phase K=1The predicted value in next period can be exported
Table 7
With upper table 7 for minimum lag phase k=m, the predicted value in following m period can be exported
From process above it can be seen that the features that use of smaller minimum lag phase k closer to current period spy
Levy, that is, k smaller, used information is closer to newest information.It is pre- by the different minimum lag phases in this module
Survey result progress is integrated, can combine the information of different lag periods, realize the rolling forecast of m phases forward.Its structure is as follows:
Namely for minimum lag phase k from 1 to m, then for the predicted value at t+i momentHave m-i+1 in advance
Measured value, then final rolling forecast exactly by it is multiple predict the outcome to be weighted collect, i.e.,Specifically
It is as follows：
Delayed (k=1) forecast model, outputObtain predicting the outcome for：w1；
Delayed (k=2) forecast model, outputObtain predicting the outcome for：w2；
Delayed (k=m) forecast model, outputObtain predicting the outcome for：wm；
Rolling forecast result is exportedWherein
The embodiment of the present invention, for external sequence carry out time series forward many phase rolling forecasts the problem of, the present invention is real
Applying example can be while considers the method that its own sequence feature is predicted with external sequence feature there is provided one, and application integration is pre-
While surveying determination suitable hysteresis quality feature, the Stability and veracity of prediction is improved.It is more and more richer with surface
It is rich, it will to there are more and more external datas to be applied to traditional time series forecasting, in a foreseeable future, how to use outside
It can be many enterprise's concerns that feature, which improves predictablity rate,.Here propose a system mechanism to apply outside sequence
Row feature is predicted.
In addition, different delayed by setting different minimum lag phase k, minimum lag phase k to be applied respectively from 1 to m
Property feature information, the final integrated rolling forecast so as to realize many phases forward of rolling forecast that uses, it is ensured that prediction following many phases
When it is used be characterized in exist, it is to avoid using surface predicted value, increase prediction uncertainty.
The method of the embodiment of the present invention can apply to finance and economics income forecast, and can equally be well applied to other has external sequence
The prediction of time series problem, such as raw material month demand.
Introduction based on previous embodiment, the embodiments of the invention provide a kind of method of rolling forecast, as shown in figure 3,
Including：
301：Determine the lag period of the hysteresis quality feature of each sequence in reference sequences；At least two are included in above-mentioned reference sequences
External sequence and/or its own sequence are included in individual sequence, above-mentioned reference sequences；
In the present embodiment, reference sequences can only have external sequence, can also there was only its own sequence, preferably make simultaneously
With external sequence and its own sequence.
302：According to the lag period of the hysteresis quality feature of each sequence in above-mentioned reference sequences, to each sequence in above-mentioned reference sequences
The hysteresis quality feature of row is reordered, and the hysteresis quality feature that the same period is belonged to after rearrangement is combined and obtains feature to be selected
Combination；
Different external sequences and its own sequence, has the different lag periods；Therefore different sequences can be calculated,
It refer to shown in table 3~5.Combinations of features may be referred to the items of formula (1).
303：Combinations of features is selected to carry out rolling forecast, choosing to the checking collection of known results from above-mentioned combinations of features to be selected
The combinations of features for taking rolling forecast result to conform to a predetermined condition is combined as target signature；
Checking collection is the sequence of knowledge of result, it is predicted using combinations of features to be selected it is known which is special
Levy combined prediction effect preferably, other are then relatively poor；Here it is preferable in order to select prediction effect to conform to a predetermined condition
Combinations of features, as to how selection, the embodiment of the present invention does not make uniqueness restriction, such as：The optimal combinations of features of selection, selection
Preferably preceding T combinations of features etc., combination of other algorithms etc. can also be carried out after selecting in addition.Target signature is combined
It is that used combination is finally predicted to its own sequence, its own sequence should be as a rule target sequence here, i.e.,：Need
Carry out the sequence of rolling forecast.
304：Rolling forecast is carried out to above-mentioned its own sequence using the combination of above-mentioned target signature.
After being obtained comprising the combinations of features of hysteresis quality feature, how to carry out rolling forecast and may be referred to existing rolling
Algorithm is realized in prediction, and the embodiment of the present invention does not make uniqueness restriction to this.
In the embodiment of the present invention, use at least two sequences as reference, comprising a variety of hysteresis quality features, different is stagnant
Afterwards property feature its lag period of correspondence, by rearrangement and the determination of combinations of features, its own sequence/external sequence pair is considered
The influence of rolling forecast result, rolling forecast result is more accurately and stably.
The embodiment of the present invention additionally provides the implementation for how selecting combinations of features, specific as follows：It is above-mentioned to be treated from above-mentioned
Select and combinations of features is selected in combinations of features, including：
In the way of traversal combinations of features is selected from above-mentioned combinations of features to be selected；Or, in a random way from above-mentioned
Combinations of features is selected in combinations of features to be selected；Or, selected with traversal or random manner from above-mentioned combinations of features to be selected
Hysteresis quality feature and at least one its own sequence comprising at least one external sequence in combinations of features, and the combinations of features of selection
Hysteresis quality feature.
Three of the above selection combinations of features scheme in, wherein the first amount of calculation is larger, but advantage be it is more complete
Face, therefore in computing capability compared with strong or preferred implementation scheme can be used as in the case that computing resource is more；Second of calculating
Amount is relatively small, can improve computational efficiency；The third advantage is while considering external sequence also its own sequence to rolling
The dynamic influence predicted the outcome.
Further, because qualified combinations of features may be a lot, and it may not be certain to be required for, therefore the present invention
Embodiment provides the scheme for simplifying combinations of features, specific as follows：Above-mentioned target signature combination meets predetermined comprising at least two
The combinations of features of condition；Before rolling forecast is carried out to above-mentioned its own sequence using the combination of above-mentioned target signature, the above method
Also include：
Selected characteristic combination obtains objective cross to be determined from the above-mentioned combinations of features conformed to a predetermined condition；
Rolling forecast is carried out to the checking collection of above-mentioned known results using above-mentioned objective cross to be determined, it is predetermined meeting
Above-mentioned objective cross to be determined is combined as above-mentioned target signature during stop condition.
The present embodiment has been further combined combinations of features, and this contributes to the Stability and veracity for further improving prediction.
Alternatively, it is also possible to carry out certain screening to combinations of features.
Above stop condition can carry out any setting, this implementation according to the requirement for improving forecasting accuracy and stability
Example gives wherein three kinds optional implementations for reference, specific as follows：Above-mentioned predetermined stop condition includes：
New selection combinations of features is added to above-mentioned objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, the number of times for having lifting to be chosen in the case of predicting the outcome to same combinations of features reaches predetermined number of times.
In three of the above stop condition, the first is optimal for target to predict the outcome, and second of control is the simplest, smaller
Pre-determined number can prevent over-fitting；The third control is also relatively simple, when being repeatedly drawn into same combinations of features
When stop selection, it may be determined that other combinations of features can not all improve prediction effect；Here the specific number of times of " multiple " can be pre-
First set, such as：2nd, 5 or other values.
Due to can at least include two hysteresis quality features in target signature combination, therefore the embodiment of the present invention additionally provides rolling
The dynamic numerical procedure predicted the outcome, it is specific as follows：At least two combinations of features are included in above-mentioned target signature combination, it is above-mentioned to use
Above-mentioned target signature combination carries out rolling forecast to above-mentioned its own sequence to be included：
Each combinations of features in being combined using above-mentioned target signature carries out rolling forecast to above-mentioned its own sequence respectively, then
Calculate the weighted average of the result of rolling forecast.
In the present embodiment, contained in target signature combination and hysteresis quality feature is included in combinations of features, combinations of features, it is right
May be referred to existing rolling forecast algorithm in the rolling forecast implementation of each combinations of features, the embodiment of the present invention to this not
Make uniqueness restriction；Which kind of weights is used for different combinations of features, can be determined by training algorithm, the present invention is real
Apply example and also do not make uniqueness restriction to this, weighted average calculation mode may be referred in previous embodiment on rolling forecast in addition
Integration module is illustrated.
The embodiment of the present invention additionally provides a kind of rolling forecast device, as shown in figure 4, including：
Lag period determining unit 401, the lag period of the hysteresis quality feature for determining each sequence in reference sequences；Above-mentioned ginseng
Examine to include in sequence and external sequence and/or its own sequence are included at least two sequences, above-mentioned reference sequences；
Rearrangement units 402, for the lag period according to the hysteresis quality feature of each sequence in above-mentioned reference sequences, to above-mentioned ginseng
The hysteresis quality feature for examining each sequence in sequence is reordered；
Assembled unit 403, feature to be selected is obtained for the hysteresis quality feature for belonging to the same period after rearrangement to be combined
Combination；
Feature selection unit 404, for selecting checking of the combinations of features to known results from above-mentioned combinations of features to be selected
Collection carries out rolling forecast, and the combinations of features that selection rolling forecast result conforms to a predetermined condition is combined as target signature；
Rolling forecast unit 405, for carrying out rolling forecast to above-mentioned its own sequence using the combination of above-mentioned target signature.
In the present embodiment, reference sequences can only have external sequence, can also there was only its own sequence, preferably make simultaneously
With external sequence and its own sequence.
Different external sequences and its own sequence, has the different lag periods；Therefore different sequences can be calculated,
It refer to shown in table 3~5.Combinations of features may be referred to the items of formula (1).
Checking collection is the sequence of knowledge of result, it is predicted using combinations of features to be selected it is known which is special
Levy combined prediction effect preferably, other are then relatively poor；Here it is preferable in order to select prediction effect to conform to a predetermined condition
Combinations of features, as to how selection, the embodiment of the present invention does not make uniqueness restriction.
After being obtained comprising the combinations of features of hysteresis quality feature, how to carry out rolling forecast and may be referred to existing rolling
Algorithm is realized in prediction, and the embodiment of the present invention does not make uniqueness restriction to this.
In the embodiment of the present invention, use at least two sequences as reference, comprising a variety of hysteresis quality features, different is stagnant
Afterwards property feature its lag period of correspondence, by rearrangement and the determination of combinations of features, its own sequence/external sequence pair is considered
The influence of rolling forecast result, rolling forecast result is more accurately and stably.
The embodiment of the present invention additionally provides the implementation for how selecting combinations of features, specific as follows：Features described above is selected
Unit 404, for selecting combinations of features from above-mentioned combinations of features to be selected in the way of traversal；Or, in a random way from
Combinations of features is selected in above-mentioned combinations of features to be selected；Or, with traversal or random manner from above-mentioned combinations of features to be selected
Select the hysteresis quality feature comprising at least one external sequence and at least one itself in combinations of features, and the combinations of features of selection
The hysteresis quality feature of sequence.
Three of the above selection combinations of features scheme in, wherein the first amount of calculation is larger, but advantage be it is more complete
Face, therefore in computing capability compared with strong or preferred implementation scheme can be used as in the case that computing resource is more；Second of calculating
Amount is relatively small, can improve computational efficiency；The third advantage is while considering external sequence also its own sequence to rolling
The dynamic influence predicted the outcome.
Further, because qualified combinations of features may be a lot, and it may not be certain to be required for, therefore the present invention
Embodiment provides the scheme for simplifying combinations of features, specific as follows：As shown in figure 5, above-mentioned target signature combination includes at least two
The individual combinations of features conformed to a predetermined condition；Said apparatus also includes：
Combine selecting unit 501, for above-mentioned rolling forecast unit 405 using above-mentioned target signature combine to it is above-mentioned from
Body sequence is carried out before rolling forecast, and selected characteristic combination obtains mesh to be determined from the above-mentioned combinations of features conformed to a predetermined condition
Mark combination；
Above-mentioned rolling forecast unit 405, is also used for checking of the above-mentioned objective cross to be determined to above-mentioned known results
Collection carries out rolling forecast；
Features described above selecting unit 404, for carrying out meeting pre- during rolling forecast in above-mentioned rolling forecast unit 405
Above-mentioned objective cross to be determined is combined as above-mentioned target signature during fixed stop condition.
The present embodiment has been further combined combinations of features, and this contributes to the Stability and veracity for further improving prediction.
Alternatively, it is also possible to carry out certain screening to combinations of features.
Above stop condition can carry out any setting, this implementation according to the requirement for improving forecasting accuracy and stability
Example gives wherein three kinds optional implementations for reference, specific as follows：Above-mentioned predetermined stop condition includes：
New selection combinations of features is added to above-mentioned objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, new selection combinations of features is added to above-mentioned objective cross to be determined and predicted the outcome the feelings that lifting predicts the outcome
Under condition, the number of times chosen to same combinations of features reaches predetermined number of times.
In three of the above stop condition, the first is optimal for target to predict the outcome, and second of control is the simplest, smaller
Pre-determined number can prevent over-fitting；The third control is also relatively simple, when being repeatedly drawn into same combinations of features
When stop selection, it may be determined that other combinations of features can not all improve prediction effect；Here the specific number of times of " multiple " can be pre-
First set, such as：2nd, 5 or other values.
Due to can at least include two hysteresis quality features in target signature combination, therefore the embodiment of the present invention additionally provides rolling
The dynamic numerical procedure predicted the outcome, it is specific as follows：At least two combinations of features are included in above-mentioned target signature combination；
Above-mentioned rolling forecast unit 405, for using each combinations of features in the combination of above-mentioned target signature respectively to above-mentioned
Its own sequence carries out rolling forecast, then calculates the weighted average of the result of rolling forecast.
In the present embodiment, contained in target signature combination and hysteresis quality feature is included in combinations of features, combinations of features, it is right
May be referred to existing rolling forecast algorithm in the rolling forecast implementation of each combinations of features, the embodiment of the present invention to this not
Make uniqueness restriction；Which kind of weights is used for different combinations of features, can be determined by training algorithm, the present invention is real
Apply example and also do not make uniqueness restriction to this, weighted average calculation mode may be referred in previous embodiment on rolling forecast in addition
Integration module is illustrated.
Well-behaved inventive embodiments additionally provide another electronic equipment, can be used for rolling forecast, as shown in fig. 6, including：
Processor 601, memory 602 can also include input equipment 603 and output equipment 604；Wherein memory 602 can be used for
Processor 601 performs the caching required for data processing, and the data and output also received for storage input equipment 603 are set
Standby 604 data that will be sent；
Wherein, processor 601, the lag period of the hysteresis quality feature for determining each sequence in reference sequences；Above-mentioned reference
Included in sequence and external sequence and/or its own sequence are included at least two sequences, above-mentioned reference sequences；According to above-mentioned reference sequence
The lag period of the hysteresis quality feature of each sequence in row, the hysteresis quality feature to each sequence in above-mentioned reference sequences reorders,
The hysteresis quality feature for belonging to the same period after rearrangement is combined and obtains combinations of features to be selected；From above-mentioned combinations of features to be selected
Select combinations of features to carry out rolling forecast to the checking collection of known results, choose the feature that rolling forecast result conforms to a predetermined condition
Combination is combined as target signature；Rolling forecast is carried out to above-mentioned its own sequence using the combination of above-mentioned target signature.
In the present embodiment, reference sequences can only have external sequence, can also there was only its own sequence, preferably make simultaneously
With external sequence and its own sequence.
Different external sequences and its own sequence, has the different lag periods；Therefore different sequences can be calculated,
It refer to shown in table 3~5.Combinations of features may be referred to the items of formula (1).
Checking collection is the sequence of knowledge of result, it is predicted using combinations of features to be selected it is known which is special
Levy combined prediction effect preferably, other are then relatively poor；Here it is preferable in order to select prediction effect to conform to a predetermined condition
Combinations of features, as to how selection, the embodiment of the present invention does not make uniqueness restriction.
After being obtained comprising the combinations of features of hysteresis quality feature, how to carry out rolling forecast and may be referred to existing rolling
Algorithm is realized in prediction, and the embodiment of the present invention does not make uniqueness restriction to this.
In the embodiment of the present invention, use at least two sequences as reference, comprising a variety of hysteresis quality features, different is stagnant
Afterwards property feature its lag period of correspondence, by rearrangement and the determination of combinations of features, its own sequence/external sequence pair is considered
The influence of rolling forecast result, rolling forecast result is more accurately and stably.
The embodiment of the present invention additionally provides the implementation for how selecting combinations of features, specific as follows：Above-mentioned processor
601, for selecting combinations of features from above-mentioned combinations of features to be selected, including：
For selecting combinations of features from above-mentioned combinations of features to be selected in the way of traversal；Or, in a random way from
Combinations of features is selected in above-mentioned combinations of features to be selected；Or, with traversal or random manner from above-mentioned combinations of features to be selected
Select the hysteresis quality feature comprising at least one external sequence and at least one itself in combinations of features, and the combinations of features of selection
The hysteresis quality feature of sequence.
Three of the above selection combinations of features scheme in, wherein the first amount of calculation is larger, but advantage be it is more complete
Face, therefore in computing capability compared with strong or preferred implementation scheme can be used as in the case that computing resource is more；Second of calculating
Amount is relatively small, can improve computational efficiency；The third advantage is while considering external sequence also its own sequence to rolling
The dynamic influence predicted the outcome.
Further, because qualified combinations of features may be a lot, and it may not be certain to be required for, therefore the present invention
Embodiment provides the scheme for simplifying combinations of features, specific as follows：Above-mentioned target signature combination meets predetermined comprising at least two
The combinations of features of condition；Above-mentioned processor 601, is additionally operable to roll above-mentioned its own sequence using the combination of above-mentioned target signature
Before dynamic prediction, selected characteristic combination obtains objective cross to be determined from the above-mentioned combinations of features conformed to a predetermined condition；
Rolling forecast is carried out to the checking collection of above-mentioned known results using above-mentioned objective cross to be determined, it is predetermined meeting
Above-mentioned objective cross to be determined is combined as above-mentioned target signature during stop condition.
The present embodiment has been further combined combinations of features, and this contributes to the Stability and veracity for further improving prediction.
Alternatively, it is also possible to carry out certain screening to combinations of features.
Above stop condition can carry out any setting, this implementation according to the requirement for improving forecasting accuracy and stability
Example gives wherein three kinds optional implementations for reference, specific as follows：Above-mentioned predetermined stop condition includes：
New selection combinations of features is added to above-mentioned objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, the number of times for having lifting to be chosen in the case of predicting the outcome to same combinations of features reaches predetermined number of times.
In three of the above stop condition, the first is optimal for target to predict the outcome, and second of control is the simplest, smaller
Pre-determined number can prevent over-fitting；The third control is also relatively simple, when being repeatedly drawn into same combinations of features
When stop selection, it may be determined that other combinations of features can not all improve prediction effect；Here the specific number of times of " multiple " can be pre-
First set, such as：2nd, 5 or other values.
Due to can at least include two hysteresis quality features in target signature combination, therefore the embodiment of the present invention additionally provides rolling
The dynamic numerical procedure predicted the outcome, it is specific as follows：At least two combinations of features, above-mentioned processing are included in above-mentioned target signature combination
Device 601, includes for carrying out rolling forecast to above-mentioned its own sequence using the combination of above-mentioned target signature：
Each combinations of features in for being combined using above-mentioned target signature carries out rolling forecast to above-mentioned its own sequence respectively,
Then the weighted average of the result of rolling forecast is calculated.
In the present embodiment, contained in target signature combination and hysteresis quality feature is included in combinations of features, combinations of features, it is right
May be referred to existing rolling forecast algorithm in the rolling forecast implementation of each combinations of features, the embodiment of the present invention to this not
Make uniqueness restriction；Which kind of weights is used for different combinations of features, can be determined by training algorithm, the present invention is real
Apply example and also do not make uniqueness restriction to this, weighted average calculation mode may be referred in previous embodiment on rolling forecast in addition
Integration module is illustrated.
The embodiment of the present invention additionally provides a kind of server, and the server can apply to rolling forecast, be as shown in Figure 7
A kind of server architecture schematic diagram provided in an embodiment of the present invention, the server 700 can produce ratio because of configuration or performance difference
Larger difference, can include one or more central processing units (central processing units, CPU) 722
(for example, one or more processors) and memory 732, one or more storage application programs 742 or data 744
Storage medium 730 (such as one or more mass memory units).Wherein, memory 732 and storage medium 730 can be with
It is of short duration storage or persistently storage.Be stored in storage medium 730 program can include one or more modules (diagram does not have
Mark), each module can include operating the series of instructions in server.Further, central processing unit 722 can be with
It is set to communicate with storage medium 730, the series of instructions operation in storage medium 730 is performed on server 700.
Server 700 can also include one or more power supplys 726, one or more wired or wireless networks
Interface 750, one or more input/output interfaces 758, and/or, one or more operating systems 741, for example
Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM etc..
The rolling forecast in previous embodiment can be performed in above-described embodiment based on the server architecture shown in the Fig. 7.
It is another it should be noted that, the rolling forecast implementation of the embodiment of the present invention can also be based on terminal device, however it is not limited to server.
It is worth noting that, in above-mentioned rolling forecast device embodiment, included unit is simply patrolled according to function
Collect what is divided, but be not limited to above-mentioned division, as long as corresponding function can be realized；In addition, each function list
The specific name of member is also only to facilitate mutually distinguish, the protection domain being not intended to limit the invention.
In addition, one of ordinary skill in the art will appreciate that realizing all or part of step in above-mentioned each method embodiment
It can be by program to instruct the hardware of correlation to complete, corresponding program can be stored in a kind of computer-readable recording medium
In, storage medium mentioned above can be read-only storage, disk or CD etc..
The present invention preferably embodiment is these are only, but protection scope of the present invention is not limited thereto, it is any
Those familiar with the art the change that can readily occur in or replaces in the technical scope that the embodiment of the present invention is disclosed
Change, should all be included within the scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of claim
Enclose and be defined.
Claims (10)
1. a kind of method of rolling forecast, it is characterised in that including：
Determine the lag period of the hysteresis quality feature of each sequence in reference sequences；At least two sequences are included in the reference sequences,
External sequence and/or its own sequence are included in the reference sequences；
According to the lag period of the hysteresis quality feature of each sequence in the reference sequences, in the reference sequences each sequence it is delayed
Property feature is reordered, and the hysteresis quality feature that the same period is belonged to after rearrangement is combined and obtains combinations of features to be selected；
Select combinations of features to carry out rolling forecast to the checking collection of known results from the combinations of features to be selected, choose and roll in advance
The combinations of features that survey result conforms to a predetermined condition is combined as target signature；
Rolling forecast is carried out to its own sequence using target signature combination.
2. method according to claim 1, it is characterised in that described to select combinations of features from the combinations of features to be selected,
Including：
In the way of traversal combinations of features is selected from the combinations of features to be selected；Or, in a random way from described to be selected
Combinations of features is selected in combinations of features；Or, feature is selected from the combinations of features to be selected with traversal or random manner
Combination, and selection combinations of features in comprising at least one external sequence hysteresis quality feature and at least one its own sequence it is stagnant
Property feature afterwards.
3. method according to claim 1 or claim 2, it is characterised in that the target signature combination meets pre- comprising at least two
The combinations of features of fixed condition；Before rolling forecast is carried out to its own sequence using target signature combination, the side
Method also includes：
Selected characteristic combination obtains objective cross to be determined from the combinations of features conformed to a predetermined condition；
Rolling forecast is carried out to the checking collection of the known results using the objective cross to be determined, is meeting predetermined stopping
The objective cross to be determined is combined as the target signature during condition.
4. method according to claim 3, it is characterised in that the predetermined stop condition includes：
New selection combinations of features is added to the objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, the number of times for having lifting to be chosen in the case of predicting the outcome to same combinations of features reaches predetermined number of times.
5. according to Claims 1-4 any one methods described, it is characterised in that comprising at least in the target signature combination
Two combinations of features, it is described that its own sequence progress rolling forecast is included using target signature combination：
Each combinations of features in being combined using the target signature carries out rolling forecast to its own sequence respectively, then calculates
The weighted average of the result of rolling forecast.
6. a kind of rolling forecast device, it is characterised in that including：
Lag period determining unit, the lag period of the hysteresis quality feature for determining each sequence in reference sequences；The reference sequences
In include at least two sequences, external sequence and/or its own sequence are included in the reference sequences；
Rearrangement units, for the lag period according to the hysteresis quality feature of each sequence in the reference sequences, to the reference sequences
In the hysteresis quality feature of each sequence reordered；
Assembled unit, combinations of features to be selected is obtained for the hysteresis quality feature for belonging to the same period after rearrangement to be combined；
Feature selection unit, for selecting combinations of features to roll the checking collection of known results from the combinations of features to be selected
Dynamic prediction, the combinations of features that selection rolling forecast result conforms to a predetermined condition is combined as target signature；
Rolling forecast unit, for carrying out rolling forecast to its own sequence using target signature combination.
7. device according to claim 6, it is characterised in that
The feature selection unit, for selecting combinations of features from the combinations of features to be selected in the way of traversal；Or, with
Random manner selects combinations of features from the combinations of features to be selected；Or, treated with traversal or random manner from described
Select and combinations of features selected in combinations of features, and selection combinations of features in comprising at least one external sequence hysteresis quality feature and
The hysteresis quality feature of at least one its own sequence.
8. according to the described device of claim 6 or 7, it is characterised in that the target signature combination meets pre- comprising at least two
The combinations of features of fixed condition；Described device also includes：
Selecting unit is combined, its own sequence is carried out for being combined in the rolling forecast unit using the target signature
Before rolling forecast, selected characteristic combination obtains objective cross to be determined from the combinations of features conformed to a predetermined condition；
The rolling forecast unit, is also used for the objective cross to be determined and the checking collection of the known results is rolled
Dynamic prediction；
The feature selection unit, for carrying out meeting predetermined stopping bar during rolling forecast in the rolling forecast unit
The objective cross to be determined is combined as the target signature during part.
9. device according to claim 8, it is characterised in that the predetermined stop condition includes：
New selection combinations of features is added to the objective cross rolling forecast result to be determined and no longer lifted；
Or, the number of times of selection combinations of features reaches pre-determined number；
Or, new selection combinations of features is added to the objective cross to be determined and predicted the outcome the situation that lifting predicts the outcome
Under, the number of times chosen to same combinations of features reaches predetermined number of times.
10. according to claim 6 to 9 any one described device, it is characterised in that comprising at least in the target signature combination
Two combinations of features；
The rolling forecast unit, for using each combinations of features in target signature combination respectively to its own sequence
Rolling forecast is carried out, the weighted average of the result of rolling forecast is then calculated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610266574.2A CN107316093B (en) | 2016-04-26 | 2016-04-26 | Rolling prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610266574.2A CN107316093B (en) | 2016-04-26 | 2016-04-26 | Rolling prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107316093A true CN107316093A (en) | 2017-11-03 |
| CN107316093B CN107316093B (en) | 2021-01-05 |
Family
ID=60184499
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610266574.2A Active CN107316093B (en) | 2016-04-26 | 2016-04-26 | Rolling prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107316093B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1795463A (en) * | 2003-05-22 | 2006-06-28 | 珀欣投资有限责任公司 | Customer revenue prediction method and system |
| CN101551884A (en) * | 2009-05-08 | 2009-10-07 | 华北电力大学 | A fast CVR electric load forecast method for large samples |
| CN102214338A (en) * | 2010-04-06 | 2011-10-12 | 上海驭策信息技术有限公司 | Sales forecasting system and method |
| CN102346964A (en) * | 2010-08-05 | 2012-02-08 | 王学鹰 | Real-time jam prediction and intelligent management system for road traffic network area |
| CN103279804A (en) * | 2013-04-29 | 2013-09-04 | 清华大学 | Super short-period wind power prediction method |
| CN105117810A (en) * | 2015-09-24 | 2015-12-02 | 国网福建省电力有限公司泉州供电公司 | Residential electricity consumption mid-term load prediction method under multistep electricity price mechanism |
-
2016
- 2016-04-26 CN CN201610266574.2A patent/CN107316093B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1795463A (en) * | 2003-05-22 | 2006-06-28 | 珀欣投资有限责任公司 | Customer revenue prediction method and system |
| CN101551884A (en) * | 2009-05-08 | 2009-10-07 | 华北电力大学 | A fast CVR electric load forecast method for large samples |
| CN102214338A (en) * | 2010-04-06 | 2011-10-12 | 上海驭策信息技术有限公司 | Sales forecasting system and method |
| CN102346964A (en) * | 2010-08-05 | 2012-02-08 | 王学鹰 | Real-time jam prediction and intelligent management system for road traffic network area |
| CN103279804A (en) * | 2013-04-29 | 2013-09-04 | 清华大学 | Super short-period wind power prediction method |
| CN105117810A (en) * | 2015-09-24 | 2015-12-02 | 国网福建省电力有限公司泉州供电公司 | Residential electricity consumption mid-term load prediction method under multistep electricity price mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107316093B (en) | 2021-01-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490335A (en) | A kind of method and device calculating participant's contribution rate | |
| CN103607459B (en) | The dynamic resource monitoring of a kind of cloud computing platform IaaS layer and dispatching method | |
| KR20210082210A (en) | Creating an Integrated Circuit Floor Plan Using Neural Networks | |
| CN106102079B (en) | Based on the C-RAN carrier wave emigration resource requirement prediction technique for improving PSO | |
| CN103853786B (en) | The optimization method and system of database parameter | |
| CN105703927B (en) | A kind of resource allocation methods, the network equipment and network system | |
| CN108280552A (en) | Methods of electric load forecasting and system, storage medium based on deep learning | |
| CN102930155B (en) | Obtain the method and device of the early-warning parameters of electricity needs | |
| Galstyan et al. | Resource allocation games with changing resource capacities | |
| CN103677960B (en) | Game resetting method for virtual machines capable of controlling energy consumption | |
| CN110118566B (en) | Robot operation path generation method and device | |
| CN113222403B (en) | Big data-based power regulation method and device, storage medium and electronic equipment | |
| CN114301935B (en) | Reputation-based internet of things edge cloud collaborative federal learning node selection method | |
| CN107133690A (en) | A kind of lake water systems connects engineering proposal preference ordering method | |
| CN106790529A (en) | The dispatching method of computing resource, control centre and scheduling system | |
| CN109993205A (en) | Time Series Forecasting Methods, device, readable storage medium storing program for executing and electronic equipment | |
| Moradi et al. | Defending smart electrical power grids against cyberattacks with deep Q-learning | |
| CN104993966B (en) | A kind of electric integrated service network method for predicting | |
| CN110191015A (en) | Cloud service performance intelligent Forecasting and device based on CPI index | |
| CN105743720B (en) | A kind of link-quality-evaluating method and its device | |
| CN107958344A (en) | A kind of power distribution network Development Strategy Analysis method based on AHP and SWOT | |
| CN107316093A (en) | The method and device of a kind of rolling forecast | |
| CN106469346B (en) | A kind of risk control method and equipment based on region | |
| Wang et al. | A study of situation awareness-based resource management scheme in cloud environment | |
| Yan et al. | Service caching for meteorological emergency decision-making in cloud-edge computing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |