CN109583625A - One kind pulling part amount prediction technique, system, equipment and storage medium - Google Patents
One kind pulling part amount prediction technique, system, equipment and storage medium Download PDFInfo
- Publication number
- CN109583625A CN109583625A CN201811223374.4A CN201811223374A CN109583625A CN 109583625 A CN109583625 A CN 109583625A CN 201811223374 A CN201811223374 A CN 201811223374A CN 109583625 A CN109583625 A CN 109583625A
- Authority
- CN
- China
- Prior art keywords
- parameter
- data
- historical data
- model
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to one kind to pull part amount prediction technique, system, equipment and storage medium, described to pull part amount prediction technique, comprising: obtains historical data, is classified and obtained the initial parameter of different attribute to historical data；Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；It selects to determine that parameter establishes prediction model, does not need to meet data stationarity or treated data steady, auto-correlation, white noise etc. compared to more existing temporal model, solve the constraint that temporal model requires condition.
Description
Technical field
The present invention relates to pull part amount electric powder prediction more particularly to one kind pull part amount prediction technique, system, equipment and
Storage medium.
Background technique
Currently, the time series data prediction of industry is chiefly used in day degree, it is pre- to be suitable for wave time for the longer prediction such as monthly
The prediction technique for surveying (such as: half an hour, hour, shift) is less, and relatively conventional method has traditional temporal model, broad sense can
Add model.
But due to the movable particularity of logistic industry, wave time prediction is more complex, for example that there are multimodals is more for wave time data
The phenomenon that paddy, and influenced vulnerable to other extraneous events, the prediction for flexibly switching simultaneously high-accuracy can not be carried out by conventional model,
It is unable to satisfy demand of the loglstics enterprise to the wave time stability forecast of part amount.
When prewave time part amount predicts there is following difficult point:
(1) the short period time granularity and granularity predicted are uncertain.Current temporal model is chiefly used in day degree, monthly
Etc. longer prediction, department pattern can also be used for the wave time prediction of hour dimension；But it in logistic industry wave time prediction, has
The prediction of other complicated periods, such as the shift part amount prediction of logistic industry, the prediction of half an hour, 20 minutes pre-
What survey etc., the prediction short period time granularity of these situations, and the prediction for being not belonging to common time granularity need to be predicted
Time granularity exists uncertain.
(2) it defers outside.For wave time prediction, such as the prediction of half an hour, there are 48 and a half hours within one day, then extrapolating
30 days are 48*30=1440 time span of forecasts of extrapolation, and extrapolation issue is more, exists to model accuracy and challenges.
(3) multicycle.Traditional time series models (such as ARIMA and Holter-Winters) are using difference or put down
The case where sliding method removes periodicity, but it is only applicable to few period.Analysis finds that there are classes for the wave time part amount of loglstics enterprise
The period of secondary, all, the moon, season, year etc., period are complicated.
(4) exogenous shock.Under practical business scene, part amount data will receive the influence such as weather of some external events, political affairs
Plan, activity etc. are difficult to that these influence factors, such as price rebate activity, typhoon weather is added in temporal model.
(5) festivals or holidays influence.When festivals or holidays mutually conflict with cycle effect or multiple festivals or holidays have intersection, just make
Whole prediction becomes complicated, and the festivals or holidays mode and coverage that national Mainland, Taiwan and Hong Kong is different, these factors have part amount
It influences.
(6) the more difficult satisfaction of assumed condition.General model has certain requirement to time series data, for example ARIMA requires sequence
Column are steady.
(7) pattern switching.Current temporal model is predicted as same mode, but actual scene to any moment or date
In, it may appear that different modes, such as weekend and working day, festivals or holidays and non-festivals or holidays etc.；New model is needed, it can be to not
Flexible model selection is carried out with mode.
By upper, the characteristics of logistics activity has its own and industry scene, existing time sequence forecasting method can not be captured accurately
Its changing rule and periodicity need a kind of suitable for logistic industry, flexible wave time time sequence forecasting method.
Summary of the invention
In order to solve the above-mentioned technical problem, the purpose of the present invention is to provide one kind to pull part amount prediction technique, system, sets
Standby and storage medium.
According to an aspect of the invention, there is provided one kind pulls part amount prediction technique, comprising:
Historical data is obtained, is classified and obtained the initial parameter of different attribute to historical data；
Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；
Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；
It selects to determine that parameter establishes prediction model.
Further, classify to historical data, comprising:
Historical data is divided into training data, test data and corresponding training result, test result, and sometimes
Between sequence concept historical data test data when being classified time after the time of training data.
Further, the attribute of the parameter comprises at least one of the following: Temporal Order, festivals or holidays attribute and period belong to
Property.
Further, the festivals or holidays attribute comprises at least one of the following: the affiliated festivals or holidays property of the historical data,
Number of days after number of days, section, respectively corresponds parameter a before the order of festivals or holidays, section1、a2、a3、a4；
And/or
The Temporal Order comprises at least one of the following: the date locating for the historical data, month, this month ten days number, when
Month all numbers, when week number of weeks, respectively correspond parameter b1、b2、b3、 b4、b5。
Further, according to the initial parameter of different attribute, cross validation is carried out to the historical data, comprising:
Determine pulling the part time, pulling part address for training data；
According to part address is pulled, choose training data festivals or holidays attribute and corresponding initial parameter；
According to the part time is pulled, choose training data Temporal Order and corresponding initial parameter；
Initial parameter is modified by training result, obtains corrected parameter；
Cross validation, the determination parameter of Selection Model are carried out to the corrected parameter by test data, test result.
Further, according to the initial parameter of different attribute, cross validation is carried out to the historical data, further includes:
Configure the duration of predetermined period；
Period locating for historical data is divided into several predetermined period, the wave in each predetermined period is numbered；
The cyclic attributes and corresponding initial parameter of training data are obtained according to number.
Further, prediction model prediction wave time when a length of following at least one moon, week, day, hour, half small
When, 20 minutes and the uncertain time granularity of time span.
According to another aspect of the present invention, it provides one kind and pulls part amount forecasting system, comprising:
Data acquisition module is configured to obtain historical data, is classified to historical data and obtain different attribute
Initial parameter；
Model building module is configured to:
Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；
Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；
It selects to determine that parameter establishes prediction model.
Further, the data acquisition module includes acquisition unit and taxon；
Acquisition unit is configured to acquisition historical data；
Taxon is configured to for historical data to be divided into training data, test data and corresponding training knot
Fruit, test result, and the time of when having time sequence concept historical data is classified test data is in training data
Time after.
Further, the attribute of the parameter comprises at least one of the following: Temporal Order, festivals or holidays attribute and period belong to
Property.
Further, the festivals or holidays attribute comprises at least one of the following: the affiliated festivals or holidays property of the historical data,
Number of days after number of days, section, respectively corresponds parameter a before the order of festivals or holidays, section1、a2、a3、a4；
And/or
The Temporal Order comprises at least one of the following: date locating for the historical data, month, ten days number, all number, star
Issue respectively corresponds parameter b1、b2、b3、b4、b5。
Further, the model building module includes:
Data identification unit is configured to determine pulling the part time, pulling part address for training data；
Festivals or holidays parameter selection unit is configured to choose according to pulling part address the festivals or holidays attribute of training data and right
The initial parameter answered；
Time sequence parameter selection unit is configured to choose according to pulling the part time Temporal Order of training data and corresponding
Initial parameter.
Parameters revision unit is configured to training result and is modified to initial parameter, obtains corrected parameter；
Parameter determination unit is configured to test data, test result intersect to the corrected parameter and be tested
Card, the determination parameter of Selection Model.
Further, the model building module further include:
Period setting unit is configured to the duration of configuration predetermined period；
Data number unit is configured to period locating for historical data being divided into several predetermined period, to each prediction
Wave in period is numbered；
Cycle parameter selection unit is configured to obtain cyclic attributes of training data and corresponding initial according to number
Parameter.
Further, prediction model prediction wave time when a length of following at least one moon, week, day, hour, half small
When, 20 minutes and the uncertain time granularity of time span.
According to another aspect of the present invention, a kind of equipment is provided, including
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of
Processor executes as above described in any item methods.
According to another aspect of the present invention, a kind of computer-readable storage medium for being stored with computer program is provided
Matter realizes as above described in any item methods when the program is executed by processor.
Compared with prior art, the invention has the following advantages:
1, the present invention pulls part amount prediction technique and obtains historical data, selects xgboost model, increases festivals or holidays attribute ginseng
Several and cyclic attributes parameter compares more existing temporal model, does not need to meet that initial data is steady or treated that data are flat
Surely；By the design to festivals or holidays attribute, the precision of prediction that model pulls part amount is improved；
There is apparent mechanical periodicity due to pulling part amount, cyclic attributes parameter can reduce dependence of the model to Temporal Order,
Influence of the Temporal Order to prediction result when can prevent from deferring more outside increases prediction data accuracy rate；
2, the present invention pulls part amount forecasting system according to the initial parameter of different attribute, intersects to the historical data
Verifying, obtains the determination parameter of model, can carry out Accurate Prediction to part amount is pulled.
3, the exemplary equipment of the present invention pulls part amount prediction technique by processor execution, can be suitable for logistic industry.
4, the exemplary readable storage medium storing program for executing of the present invention stores the part amount of pulling realized when being executed by processor and predicts
Method, convenient for pulling using and promoting for part amount forecasting system.
Detailed description of the invention
Fig. 1 is flow chart of the present invention.
Fig. 2 is present system block diagram.
Fig. 3 pulls part amount in prediction with existing prediction model for the present invention in embodiment 1 and actually pulls part amount averagely relatively accidentally
The comparison of difference, new represent prediction error of the invention, and prophet represents the prediction error of existing model.
Specific embodiment
In order to be better understood by technical solution of the present invention, combined with specific embodiments below, Figure of description is to the present invention
It is described further.
Embodiment 1:
The present embodiment provides one kind to pull part amount forecasting system, including data acquisition module, model building module；
Data acquisition module is configured to obtain historical data, is classified to historical data and obtain different attribute
The attribute of initial parameter, the parameter comprises at least one of the following: Temporal Order, festivals or holidays attribute and cyclic attributes.
Specifically, the data acquisition module includes acquisition unit and taxon；
Acquisition unit is configured to acquisition historical data；
Taxon is configured to for historical data to be divided into training data, test data and corresponding training knot
Fruit, test result, and the time of when having time sequence concept historical data is classified test data is in training data
Time after.
Model building module is configured to:
Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；
Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；
It selects to determine that parameter establishes prediction model.
Specifically, the model building module includes:
Data identification unit is configured to determine pulling the part time, pulling part address for training data；
Festivals or holidays parameter selection unit is configured to choose according to pulling part address the festivals or holidays attribute of training data and right
The initial parameter answered；Pulling part address is Mainland, Taiwan and Hong Kong, and the festivals or holidays attribute comprises at least one of the following: the historical data
Number of days after number of days, section, respectively corresponds parameter a before affiliated festivals or holidays property, the order of festivals or holidays, section1、a2、a3、a4；
Time sequence parameter selection unit is configured to choose according to pulling the part time Temporal Order of training data and corresponding
Initial parameter；The Temporal Order comprises at least one of the following: date locating for the historical data, month, ten days number are (in up and down
Ten days), of that month all numbers, when the number of weeks (such as Monday) in week, respectively correspond parameter b1、b2、b3、b4、b5。
Period setting unit is configured to the duration of configuration predetermined period；
Data number unit is configured to period locating for historical data being divided into several predetermined period, to each prediction
Wave in period is numbered；
Cycle parameter selection unit is configured to obtain cyclic attributes of training data and corresponding initial according to number
Parameter.
Parameters revision unit is configured to training result and is modified to initial parameter, obtains corrected parameter；
Parameter determination unit is configured to test data, test result intersect to the corrected parameter and be tested
Card, the determination parameter of Selection Model.
Select to determine that parameter establishes prediction model, the prediction model prediction wave time when a length of following at least one moon,
Week, day, hour, half an hour, 20 minutes and the uncertain time granularity of time span (such as shift).
Pull the corresponding prediction technique of part amount forecasting system, comprising:
S1: obtain historical data, the present embodiment historical data by certain site pull part record pull part record for, example is such as
Under:
| Date | Half an hour section | Quantity |
| 2016-07-01 | 08:00-08:30 | X1 |
| 2016-07-01 | 08:30-09:00 | X2 |
| 2016-07-01 | 09:00-09:30 | X3 |
| 2016-07-01 | 09:30-10:00 | X4 |
| 2016-07-01 | 10:00-10:30 | X5 |
| 2016-07-01 | 10:30-11:00 | X6 |
| 2016-07-01 | 11:00-11:30 | X7 |
| 2016-07-01 | 11:30-12:00 | X8 |
| 2016-07-01 | 12:00-12:30 | X9 |
S1-1: historical data is divided into training data, test data and corresponding training result, test result, is obtained
Take different attribute initial parameter (initial parameter can be according to sorted historical data establish parameter pond, be existing skill
Art, details are not described herein again)；And having time sequence concept historical data when being classified the time of the test data instructing
After the time for practicing data；The attribute of the parameter comprises at least one of the following: Temporal Order, festivals or holidays attribute and period belong to
Property.
The festivals or holidays attribute comprises at least one of the following: the affiliated festivals or holidays property of the historical data, time of festivals or holidays
Number of days after number of days, section, respectively corresponds parameter a before sequence, section1、a2、a3、 a4。
The Temporal Order comprises at least one of the following: date locating for the historical data, month, ten days number, all number, star
Issue respectively corresponds parameter b1、b2、b3、b4、b5。
S2: selection xgboost model, prediction model prediction wave time when a length of following at least one moon, week, day,
Hour, half an hour, 20 minutes and the uncertain time granularity of time span；
S2-1: pulling the part time, pulling part address for training data is determined；
S2-2: according to part address is pulled, choose training data festivals or holidays attribute and corresponding initial parameter；
S2-3: according to the part time is pulled, choose training data Temporal Order and corresponding initial parameter；
S2-4: configuring the duration of predetermined period, and the present embodiment sets two predetermined period according to festivals or holidays, non-festivals or holidays
Duration.
Situation 1 (non-festivals or holidays): it using week as predetermined period, obtains:
1) period of target wave time, and target wave is numbered in predetermined period, if target wave time is for Monday
08:00-08:30,08:30-09 are then numbered: 0-1-1,0-1-2；
2) last week is the same as week while segment amount；
3) week before last is the same as week while segment amount；
Using week as predetermined period, default historical time section is numbered as unit of half an hour, for example target wave traces back
It is 1-1-1,1-1-2 that 08:00-08:30,08:30-09:00 of a cycle Monday is numbered respectively, and target wave traces back the
It is 2-1-1,2-1-2 that 08:00-08:30,08:30-09:00 of two Mondays in period is numbered respectively, by target wave time and prediction
The wave secondary association of reference numeral in period obtains the cyclic attributes and corresponding initial parameter of training data.
Situation 2 (festivals or holidays): it using year as predetermined period, obtains:
1) festivals or holidays attribute (data under festivals or holidays mode can add), not to different regions (such as Mainland, Taiwan and Hong Kong) setting
Same variable:
2) festivals or holidays name parameter；
2. festivals or holidays property is positive influence or negative influence h_efffect to part amount；
3. the order h_order of festivals or holidays belonging to the same day, such as the 1st day, the 2nd day, the 3rd day；
4. number of days before saving；Number of days after section；
Default historical time section is numbered as unit of half an hour, for example, vacation first day 08:00-08:30,08:
30-09:00 number respectively be 1-1- 1., 1-1- 2., save preceding first day 08:00-08:30,08:30-09:00 and number respectively and is
R1-1- 1., R1-1- 2., after section first day 08:00-08:30,08:30-09:00 number respectively be S1-1- 1., S1-1- 2.；
Wave secondary association by target wave time with reference numeral in predetermined period, obtain training data cyclic attributes and it is corresponding initially
Parameter；
S2-5: by the initial parameter input xgboost model of training data, different attribute and to being modified, according to instruction
Practice result to be modified initial parameter, obtains multiple parameter combinations；
Parameter combination includes but can be not limited to following:
The depth capacity of max_dept tree；
The amount of n_estimators construction tree；
Min_child_weight minimum leaf node sample weights；
Learning_rate learning rate；
Subsample accounts for the ratio of entire sample set for the subsample of training pattern；
Colsample_bytree is when establishing tree to the ratio of feature stochastical sampling.
S3: cross validation is carried out to the corrected parameter by test data, test result, by prediction result and test
As a result it compares, the superiority and inferiority of each parameter combination is measured according to error (such as absolute error quadratic sum), so that selection is best
Parameter combination is to determine parameter.
For the present embodiment by taking situation 1 as an example, predetermined period is to carry out weekly once, in conjunction with various parameters, selection past two
The data of the moon (8 weeks) carry out 8 folding cross validations, and prediction result is compared with test data respectively, prediction result is filtered out and surveys
The smallest parameter combination of data error is tried, optimal model parameter is obtained,
S4: pulling part amount using xgboost model prediction target wave time (half an hour), specific to predict error such as Fig. 3 institute
Show, influence of the Temporal Order to prediction result when can prevent from deferring more outside, increases prediction data accuracy rate.
The above is only related become can be added according to different business scene to each parametric prediction model according to this city embodiment
Weather variable weather, policy, social activities (such as double 11, world cup etc.) etc. can be added in amount, the prediction of part amount.
The present embodiment provides a kind of equipment, the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of
Processor executes method as described in any one of the above embodiments.
A kind of computer readable storage medium being stored with computer program is realized such as when the program is executed by processor
Method described in any of the above embodiments.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Art technology
Personnel should be appreciated that invention scope involved in the application, however it is not limited to skill made of the specific combination of above-mentioned technical characteristic
Art scheme, while should also cover in the case where not departing from the inventive concept, by above-mentioned technical characteristic or its equivalent feature into
Row any combination and the other technical solutions formed.Such as features described above and (but being not limited to) disclosed herein have class
Like function.
Claims (16)
1. one kind pulls part amount prediction technique, characterized in that include:
Historical data is obtained, is classified and obtained the initial parameter of different attribute to historical data；
Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；
Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；
It selects to determine that parameter establishes prediction model.
2. according to claim 1 pull part amount prediction technique, characterized in that classify to historical data, comprising:
Historical data is divided into training data, test data and corresponding training result, test result, and having time sequence
The time of concept historical data test data when being classified is after the time of training data.
3. according to claim 1 pull part amount prediction technique, characterized in that the attribute of the parameter includes following at least one
Kind: Temporal Order, festivals or holidays attribute and cyclic attributes.
4. according to claim 3 pull part amount prediction technique, characterized in that
The festivals or holidays attribute comprises at least one of the following: the affiliated festivals or holidays property of the historical data, the order of festivals or holidays, section
Number of days after preceding number of days, section respectively corresponds parameter a1、a2、a3、a4；
And/or
The Temporal Order comprises at least one of the following: the date locating for the historical data, month, this month ten days number, of that month
All numbers, when week number of weeks, respectively correspond parameter b1、b2、b3、b4、b5。
5. according to claim 3 pull part amount prediction technique, characterized in that according to the initial parameter of different attribute, to institute
It states historical data and carries out cross validation, comprising:
Determine pulling the part time, pulling part address for training data；
According to part address is pulled, choose training data festivals or holidays attribute and corresponding initial parameter；
According to the part time is pulled, choose training data Temporal Order and corresponding initial parameter；
Initial parameter is modified by training result, obtains corrected parameter；
Cross validation, the determination parameter of Selection Model are carried out to the corrected parameter by test data, test result.
6. according to claim 5 pull part amount prediction technique, characterized in that according to the initial parameter of different attribute, to institute
It states historical data and carries out cross validation, further includes:
Configure the duration of predetermined period；
Period locating for historical data is divided into several predetermined period, the wave in each predetermined period is numbered；
The cyclic attributes and corresponding initial parameter of training data are obtained according to number.
7. according to claim 1 pull part amount prediction technique, characterized in that prediction model prediction wave time when it is a length of
The following at least one moon, week, day, hour, half an hour, 20 minutes and the uncertain time granularity of time span.
8. one kind pulls part amount forecasting system, characterized in that include:
Data acquisition module is configured to obtain historical data, is classified to historical data and obtain the initial of different attribute
Parameter；
Model building module is configured to:
Xgboost model is selected, the initial parameter of different attribute is inputted into xgboost model and to being modified；
Cross validation, the determination parameter of Selection Model are carried out to revised initial parameter；
It selects to determine that parameter establishes prediction model.
9. according to claim 8 pull part amount forecasting system, characterized in that the data acquisition module includes acquisition unit
And taxon；
Acquisition unit is configured to acquisition historical data；
Taxon is configured to for historical data to be divided into training data, test data and corresponding training result, test
As a result, and having time sequence concept historical data test data when being classified time the time of training data it
Afterwards.
10. according to claim 8 pull part amount forecasting system, characterized in that the attribute of the parameter include it is following at least
It is a kind of: Temporal Order, festivals or holidays attribute and cyclic attributes.
11. according to claim 10 pull part amount forecasting system, characterized in that the festivals or holidays attribute include it is following at least
A kind of: number of days after number of days, section before the affiliated festivals or holidays property of the historical data, the order of festivals or holidays, section respectively corresponds parameter
a1、a2、a3、a4；
And/or
The Temporal Order comprises at least one of the following: the date locating for the historical data, month, this month ten days number, of that month
All numbers, when week number of weeks, respectively correspond parameter b1、b2、b3、b4、b5。
12. according to claim 10 pull part amount forecasting system, characterized in that the model building module includes:
Data identification unit is configured to determine pulling the part time, pulling part address for training data；
Festivals or holidays parameter selection unit is configured to choose according to pulling part address the festivals or holidays attribute of training data and corresponding
Initial parameter；
Time sequence parameter selection unit is configured to choose according to pulling the part time Temporal Order of training data and corresponding initial
Parameter；
Parameters revision unit is configured to training result and is modified to initial parameter, obtains corrected parameter；
Parameter determination unit is configured to test data, test result and carries out cross validation to the corrected parameter, chooses
The determination parameter of model.
13. according to claim 12 pull part amount forecasting system, characterized in that the model building module further include:
Period setting unit is configured to the duration of configuration predetermined period；
Data number unit is configured to period locating for historical data being divided into several predetermined period, to each predetermined period
Interior wave is numbered；
Cycle parameter selection unit is configured to obtain the cyclic attributes and corresponding initial parameter of training data according to number.
14. according to claim 8 pull part amount forecasting system, characterized in that the duration of the prediction model prediction wave time
For following at least one moon, week, day, hour, half an hour, 20 minutes and the uncertain time granularity of time span.
15. a kind of equipment, characterized in that the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors
Execute method as claimed in any one of claims 1-9 wherein.
16. a kind of computer readable storage medium for being stored with computer program, characterized in that when the program is executed by processor
Realize method as claimed in any one of claims 1-9 wherein.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811223374.4A CN109583625A (en) | 2018-10-19 | 2018-10-19 | One kind pulling part amount prediction technique, system, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811223374.4A CN109583625A (en) | 2018-10-19 | 2018-10-19 | One kind pulling part amount prediction technique, system, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109583625A true CN109583625A (en) | 2019-04-05 |
Family
ID=65920243
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811223374.4A Pending CN109583625A (en) | 2018-10-19 | 2018-10-19 | One kind pulling part amount prediction technique, system, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109583625A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110333556A (en) * | 2019-06-03 | 2019-10-15 | 深圳中兴网信科技有限公司 | Air Quality Forecast method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN111898786A (en) * | 2019-05-05 | 2020-11-06 | 顺丰科技有限公司 | Electronic lock demand prediction method, system, equipment and storage medium |
| CN112183832A (en) * | 2020-09-17 | 2021-01-05 | 上海东普信息科技有限公司 | Express pickup quantity prediction method, device, equipment and storage medium |
| CN112183873A (en) * | 2020-10-10 | 2021-01-05 | 苏州创旅天下信息技术有限公司 | Traffic prediction method, system, device and storage medium |
| CN112631882A (en) * | 2020-12-03 | 2021-04-09 | 四川新网银行股份有限公司 | Capacity estimation method combined with online service index characteristics |
| CN112801327A (en) * | 2019-11-14 | 2021-05-14 | 顺丰科技有限公司 | Method, device, equipment and storage medium for predicting logistics flow and modeling thereof |
| CN112884189A (en) * | 2019-11-29 | 2021-06-01 | 顺丰科技有限公司 | Order quantity prediction model training method, device and equipment |
| CN112966849A (en) * | 2019-12-13 | 2021-06-15 | 顺丰科技有限公司 | Method, device and equipment for establishing component prediction model |
| CN112990526A (en) * | 2019-12-16 | 2021-06-18 | 顺丰科技有限公司 | Method and device for predicting logistics arrival quantity and storage medium |
| CN113191537A (en) * | 2021-04-15 | 2021-07-30 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for predicting express package data |
| CN112966849B (en) * | 2019-12-13 | 2024-06-07 | 顺丰科技有限公司 | Method, device and equipment for establishing part quantity prediction model |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104899405A (en) * | 2014-03-04 | 2015-09-09 | 携程计算机技术（上海）有限公司 | Data prediction method and system and alarming method and system |
| CN107590567A (en) * | 2017-09-13 | 2018-01-16 | 南京航空航天大学 | A kind of Recognition with Recurrent Neural Network short-term load forecasting method based on comentropy cluster and notice mechanism |
| CN107895283A (en) * | 2017-11-07 | 2018-04-10 | 重庆邮电大学 | A kind of businessman's volume of the flow of passengers big data Forecasting Methodology based on Time Series |
| CN108375808A (en) * | 2018-03-12 | 2018-08-07 | 南京恩瑞特实业有限公司 | Dense fog forecasting procedures of the NRIET based on machine learning |
| CN108647373A (en) * | 2018-03-21 | 2018-10-12 | 浙江大学 | A kind of industrial process flexible measurement method based on xgboost models |
-
2018
- 2018-10-19 CN CN201811223374.4A patent/CN109583625A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104899405A (en) * | 2014-03-04 | 2015-09-09 | 携程计算机技术（上海）有限公司 | Data prediction method and system and alarming method and system |
| CN107590567A (en) * | 2017-09-13 | 2018-01-16 | 南京航空航天大学 | A kind of Recognition with Recurrent Neural Network short-term load forecasting method based on comentropy cluster and notice mechanism |
| CN107895283A (en) * | 2017-11-07 | 2018-04-10 | 重庆邮电大学 | A kind of businessman's volume of the flow of passengers big data Forecasting Methodology based on Time Series |
| CN108375808A (en) * | 2018-03-12 | 2018-08-07 | 南京恩瑞特实业有限公司 | Dense fog forecasting procedures of the NRIET based on machine learning |
| CN108647373A (en) * | 2018-03-21 | 2018-10-12 | 浙江大学 | A kind of industrial process flexible measurement method based on xgboost models |
Non-Patent Citations (2)
| Title |
|---|
| 刘振亚，等编著: "《金融数据挖掘》", 30 September 2016 * |
| 鄢仁祥，等编著: "《蛋白质结构生物信息学》", 31 May 2017 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898786A (en) * | 2019-05-05 | 2020-11-06 | 顺丰科技有限公司 | Electronic lock demand prediction method, system, equipment and storage medium |
| CN110333556A (en) * | 2019-06-03 | 2019-10-15 | 深圳中兴网信科技有限公司 | Air Quality Forecast method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN112801327A (en) * | 2019-11-14 | 2021-05-14 | 顺丰科技有限公司 | Method, device, equipment and storage medium for predicting logistics flow and modeling thereof |
| CN112884189A (en) * | 2019-11-29 | 2021-06-01 | 顺丰科技有限公司 | Order quantity prediction model training method, device and equipment |
| CN112966849B (en) * | 2019-12-13 | 2024-06-07 | 顺丰科技有限公司 | Method, device and equipment for establishing part quantity prediction model |
| CN112966849A (en) * | 2019-12-13 | 2021-06-15 | 顺丰科技有限公司 | Method, device and equipment for establishing component prediction model |
| CN112990526A (en) * | 2019-12-16 | 2021-06-18 | 顺丰科技有限公司 | Method and device for predicting logistics arrival quantity and storage medium |
| CN112183832A (en) * | 2020-09-17 | 2021-01-05 | 上海东普信息科技有限公司 | Express pickup quantity prediction method, device, equipment and storage medium |
| CN112183873A (en) * | 2020-10-10 | 2021-01-05 | 苏州创旅天下信息技术有限公司 | Traffic prediction method, system, device and storage medium |
| CN112183873B (en) * | 2020-10-10 | 2023-09-12 | 苏州创旅天下信息技术有限公司 | Traffic prediction method, system, device and storage medium |
| CN112631882B (en) * | 2020-12-03 | 2023-06-06 | 四川新网银行股份有限公司 | Capacity estimation method combining on-line business index characteristics |
| CN112631882A (en) * | 2020-12-03 | 2021-04-09 | 四川新网银行股份有限公司 | Capacity estimation method combined with online service index characteristics |
| CN113191537A (en) * | 2021-04-15 | 2021-07-30 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for predicting express package data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583625A (en) | One kind pulling part amount prediction technique, system, equipment and storage medium | |
| US10754688B2 (en) | Systems and methods of a production environment tool | |
| Aastveit et al. | Nowcasting GDP in real time: A density combination approach | |
| Goh et al. | Modeling and forecasting tourism demand for arrivals with stochastic nonstationary seasonality and intervention | |
| Van Slyke | Monte Carlo methods and the PERT problem | |
| Rozenfeld et al. | The area and population of cities: New insights from a different perspective on cities | |
| CN108170765A (en) | Recommend method based on the poverty-stricken mountains in school behavioral data multidimensional analysis | |
| CN108345958A (en) | A kind of order goes out to eat time prediction model construction, prediction technique, model and device | |
| CN107239846B (en) | Parking lot berth prediction processing method and device | |
| US20180365608A1 (en) | Quantitive time estimation systems and methods of project management systems | |
| Lin et al. | The bullwhip effect in hybrid supply chain | |
| CN105809264B (en) | Power load prediction method and device | |
| Jebeile et al. | Multi-model ensembles in climate science: Mathematical structures and expert judgements | |
| Du et al. | Cost performance as a stochastic process: EAC projection by Markov Chain simulation | |
| Liu et al. | Adaptive optimisation methods in system-level bridge management | |
| Ionescu et al. | Data analysis of delays in airline networks | |
| Peschiera et al. | A novel solution approach with ML-based pseudo-cuts for the Flight and Maintenance Planning problem | |
| Dash et al. | A model for planning the product portfolio and launch timings under resource constraints | |
| CN112001786A (en) | Client credit card limit configuration method and device based on knowledge graph | |
| US11867054B2 (en) | Systems and methods for estimating well parameters and drilling wells | |
| Khazai et al. | Integrated risk modelling toolkit and database for earthquake risk assessment | |
| CN108615117B (en) | Shipping bearing capacity evaluation method and device and storage medium | |
| Thébaud et al. | Evaluation of management strategies in Ningaloo Marine Park, Western Australia | |
| Hwang | A dynamic model for valuing customers: a case study | |
| AU2016202814A1 (en) | Systems and methods for managing cpu usage during qualitatively assessment of task data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190405 |
|
| RJ01 | Rejection of invention patent application after publication |