CN109767031A - Model classifiers method for building up, device, computer equipment and storage medium - Google Patents
Model classifiers method for building up, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109767031A CN109767031A CN201811555208.4A CN201811555208A CN109767031A CN 109767031 A CN109767031 A CN 109767031A CN 201811555208 A CN201811555208 A CN 201811555208A CN 109767031 A CN109767031 A CN 109767031A
- Authority
- CN
- China
- Prior art keywords
- model
- production data
- prediction
- historical production
- single model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000004519 manufacturing process Methods 0.000 claims abstract description 197
- 238000012549 training Methods 0.000 claims abstract description 60
- 230000011218 segmentation Effects 0.000 claims description 42
- 230000007423 decrease Effects 0.000 claims description 40
- 238000004590 computer program Methods 0.000 claims description 31
- 235000013399 edible fruits Nutrition 0.000 claims description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000001965 increasing effect Effects 0.000 description 9
- 238000010276 construction Methods 0.000 description 8
- 239000000284 extract Substances 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013499 data model Methods 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application involves artificial intelligence field, especially a kind of model classifiers method for building up, device, computer equipment and storage medium.The described method includes: obtaining historical production data；Training predicted time is obtained, and the trained predicted time and the historical production data are input in prediction single model and obtain single model prediction result, prediction single model is that preparatory training obtains；Feature is selected according to the single model prediction result and historical production data building object module；Feature and the prediction single model is selected to be trained to obtain model classifiers the object module.It can be improved the accuracy of model prediction using this method.
Description
Technical field
This application involves field of artificial intelligence, set more particularly to a kind of model classifiers method, apparatus, computer
Standby and storage medium.
Background technique
In financial industry, operating income refers to enterprise in production and operating activities, because sale product or utility service take
The every income obtained, it is related to the survival and development of enterprise, there is important meaning to enterprise operation, therefore, Accurate Prediction enterprise
Industry business revenue is the important content of investment analysis.
However, current for the prediction of operating income is obtained according to prediction model, general prediction model is only
It is trained by target value of predicted value, such prediction model is easy to ignore certain specific factors or causes extremely to predict
Value, therefore model accuracy is inadequate.
Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of model that can be improved model prediction accuracy point
Class device method for building up, device, computer equipment and storage medium.
A kind of model classifiers method for building up, which comprises
Obtain historical production data；
Training predicted time is obtained, and the trained predicted time and the historical production data are input to prediction list
Single model prediction result is obtained in model, the prediction single model is that preparatory training obtains；
Feature is selected according to the single model prediction result and historical production data building object module；
Feature and the prediction single model is selected to be trained to obtain model classifiers the object module.
It is described in one of the embodiments, to select feature and the prediction single model to instruct the object module
Get model classifiers, comprising:
Legitimate reading is extracted from the historical production data；
The difference for calculating the legitimate reading and the single model prediction result obtains the smallest prediction single mode of the difference
Type is as optimum prediction single model；
The optimum prediction single model, object module selection feature are trained to obtain model classifiers.
It is described to first model selection feature, the second model selection feature and institute in one of the embodiments,
Prediction single model is stated to be trained to obtain model classifiers, comprising:
Legitimate reading is extracted from the historical production data；
It is single to obtain the prediction according to the ratio for the ratio for calculating the single model prediction result and the legitimate reading
The weight of model；
Weight, the object module selection feature of the prediction single model are trained to obtain model classifiers.
It is described in one of the embodiments, to be constructed according to the single model prediction result and the historical production data
Object module selects feature, comprising:
The first model, which is constructed, according to the single model prediction result and the historical production data selects feature；
The second model, which is constructed, according to the historical production data selects feature.
It is described in one of the embodiments, that the trained predicted time and the historical production data are input to institute
It states in prediction single model and obtains single model prediction result, comprising:
Obtain the trained predicted time corresponding eigenperiod；
Predicted value corresponding with the eigenperiod is calculated by the prediction single model；
It is described that first model selection feature, packet are constructed according to the single model prediction result and the historical production data
It includes:
True value corresponding with the eigenperiod is extracted from the historical production data；
The first model selection feature is calculated according to the predicted value and the true value.
It is described in one of the embodiments, that the second model selection feature is constructed according to the historical production data, comprising:
Predetermined period length and section are obtained, the historical production data is divided according to the predetermined period length
Section；
The historical production data of corresponding segmentation in each section is obtained, and acquired historical production data is arranged
Sequence, and the sequence valve of the historical production data after tag sort；
Calculate the deviation of the sequence valve of the historical production data of the corresponding segmentation in each section；
The average value for calculating the deviation obtains periodical relative strength index as the second model and selects feature.
It is described in one of the embodiments, that the second model selection feature is constructed according to the historical production data, comprising:
Predetermined period length is obtained, the historical production data is segmented according to the predetermined period length；
Increase and decrease amplitude is calculated according to the historical production data of the historical production data of each segmentation and a upper segmentation；
The increase and decrease amplitude mark value of the historical production data of each segmentation is obtained according to the increase and decrease amplitude；
The average value for calculating the increase and decrease amplitude mark value obtains tendency relative strength index as the second model and selects feature.
A kind of model classifiers establish device, and described device includes:
Historical data obtains module, for obtaining historical production data；
Single model prediction result obtains module, for obtaining trained predicted time, and by the trained predicted time and
The historical production data is input in prediction single model and obtains single model prediction result, and the prediction single model is preparatory training
It obtains；
Fisrt feature constructs module, for constructing mesh according to the single model prediction result and the historical production data
It marks model and selects feature；
Training module, for selecting feature and the prediction single model to be trained to obtain model the object module
Classifier.
A kind of computer equipment, including memory and processor, the memory are stored with computer program, the processing
The step of device realizes any of the above-described the method when executing the computer program.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
The step of method described in any of the above embodiments is realized when row.
Above-mentioned model classifiers method for building up, device, computer equipment and storage medium, establish model classifiers when
It waits, has fully considered the single model prediction result and historical production data of a multiple single models, and predict to tie according to single model
Fruit and historical production data construct to have obtained the first model selection feature, are constructed to obtain the choosing of the second model according to historical production data
Feature is selected, thus select feature and the second model selection feature to be trained to have obtained model classifiers according to the first model, this
The feature of the characteristics of sample model classifiers have fully considered each model and historical production data, it is pre- so as to improve model
The accuracy of survey.
Detailed description of the invention
Fig. 1 is the application scenario diagram of model classifiers method for building up in one embodiment；
Fig. 2 is the flow diagram of model classifiers method for building up in one embodiment；
Fig. 3 is the structural block diagram that model classifiers establish device in one embodiment；
Fig. 4 is the internal structure chart of computer equipment in one embodiment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
Model classifiers method for building up provided by the present application, can be applied in application environment as shown in Figure 1.Wherein,
Terminal 102 is communicated by network with server 104.Terminal 102 can get historical production data from server 104,
So as to be trained to obtain multiple prediction single models according to the creation data, then according to multiple prediction single models and history
Creation data is trained to obtain model classifiers, so as to obtain predicting most accurately to predict list according to model classifiers
The prediction result of the either comprehensive multiple prediction single models of model improves prediction result to obtain most accurate prediction result
Accuracy.Specifically, terminal 102 is available to training predicted time, and will training predicted time and historical production data
It is input in prediction single model and obtains single model prediction result, so as to produce number according to single model prediction result and history
According to the first model selection feature is obtained, the prediction that first model selection feature has fully taken into account each prediction single model is missed
Difference, and in order to fully take into account the feature of historical production data, terminal 102 can also obtain second according only to historical production data
Model selects feature, so that terminal 102 can select the first model feature, the second model selection feature and prediction single model
It is trained to obtain model classifiers, so as to obtain predicting according to model classifiers most accurately predicting single model either
The prediction result of comprehensive multiple prediction single models improves the accuracy of prediction result to obtain most accurate prediction result.Its
In, terminal 102 can be, but not limited to be various personal computers, laptop, smart phone, tablet computer and it is portable can
Wearable device, server 104 can be realized with the server cluster of the either multiple server compositions of independent server.
Specifically, multiple prediction single models have been pre-generated in the application, it can be to life by multiple prediction single model
It produces data to be predicted, such as business revenue data is predicted；But since the single model prediction result of prediction single model may
There are deviation, it is accurate for leading to the single model prediction result that can not finally determine which prediction single model, therefore in order to keep away
The occurrence of exempting from this kind, single model prediction result of the application previously according to historical production data and each prediction single model
Multiple object module selection features are generated, object module selection feature may include for measuring each single model prediction knot
The feature of the accuracy of fruit and the feature of historical production data self attributes.To which the application can be according to multiple target mould
Type selects feature and prediction single model to be trained to obtain the optimum prediction of creation data as a result, for example can be according to target
Model selects feature to select to obtain optimum prediction single model, or selects feature to obtain each prediction according to object module
The weight of single model, so as to obtain creation data according to the single model prediction result of the weight and each prediction single model
Prediction result.
In one embodiment, as shown in Fig. 2, providing a kind of model classifiers method for building up, it is applied in this way
It is illustrated for terminal in Fig. 1, comprising the following steps:
S202: historical production data is obtained.
Specifically, historical production data is enterprise's generated data in previous transaction or production and living, such as
It can be history business revenue data of each enterprise etc..Terminal can be gone through to the corresponding financial server transmission of each enterprise first
History business revenue data acquisition instruction, so that the finance service of each enterprise is incited somebody to action after receiving history business revenue data acquisition instruction
History business revenue data are sent to corresponding terminal, and terminal deposits history business revenue data after receiving history business revenue data
Storage, such as be stored in the database or server of a safety.In addition, the financial server of each enterprise can also periodically to
Terminal submits the history business revenue data of each enterprise nearest a period of time, to guarantee to be stored in the history business revenue data of terminal side
Timeliness.
And optionally, after terminal gets historical production data, which can also be located in advance
It manages to improve the accuracy of model training.For example, terminal can with suppressing exception data either to historical production data at
Reason obtains the historical production data of model training needs.Wherein suppressing exception data may include that terminal judges historical production data
In it is incomplete with the presence or absence of data, be either empty data for 0, if it is present deleting those abnormal datas.Terminal is to history
The historical production data that creation data is handled to obtain model training needs can be to historical production data progress difference, example
Such as after suppressing exception data, terminal may determine that obtained historical production data master worker needs to carry out difference, such as work as institute
When obtained historical production data is the historical production data of half a year, then terminal can be subtracted by the historical production data of the half a year
The historical production data of the first quarter is removed, so as to obtain the historical production data of the second quarter.
S204: training predicted time is obtained, and training predicted time and historical production data are input to prediction single mode
Single model prediction result is obtained in type, prediction single model is that preparatory training obtains.
Specifically, prediction single model may include multiple classifications, for example, time series class model, trend fitting class model,
Timing is fitted comprehensive class model and multiple-factor class model etc..Wherein time series class model can be arima model, be to utilize
The model that time series data itself is established, historical production data include mainly wherein two fields, and one is time field, separately
Outer one is business revenue data field, so as to establish arima model according to two fields；Trend fitting class model can be with
It is that discrete point is constructed according to known business revenue data for polyfit model, i.e. the time is x-axis, and history business revenue data are y-axis,
Then constructor is fitted according to discrete point as trend fitting model, i.e., obtained fitting function is that trend is quasi-
Molding type；Timing, which is fitted comprehensive class model, can be prophet model, be continuously to be increased according to business revenue data sectional to generate
Segmented model, wherein business revenue data can be segmented according to different increasing degrees, then obtain each segmentation
All combination of function are obtained prediction single model by corresponding function；Multi-sector model can be xgboost model, be root
The prediction single model constructed according to the thought of classifier classifies history business revenue data according to time and business revenue data,
Then sorted history business revenue data are learnt to obtain prediction single model.
Specifically, training predicted time its be also historical time, i.e. past time, when training pattern,
It needs that historical production data is divided into sample data and verification data according to training predicted time, wherein sample data is for pre-
It surveys, whether verification data are for adjusting prediction model or being used to examine prediction result accurate.
Terminal can be using the corresponding historical production data of training predicted time as verification data；And it will training predicted time
Historical production data before is as sample data；The sample data and training predicted time are input to prediction single model by terminal
In each available single model prediction result for being exported of prediction single model.Such as training predicted time is third in 2018
Season, then the sample data is input to prediction as sample data by the available historical production data to before 2018
It can be obtained in single model and training predicted time corresponding single model prediction result, such as available arima model
Prediction result, the prediction result of polyfit model, the prediction result of prophet model and the prediction result of xgboost model
Deng.
S206: feature is selected according to single model prediction result and historical production data building object module.
Specifically, object module selection feature may include the spy for measuring the accuracy of each single model prediction result
The feature of sign and historical production data self attributes, for example, object module selection feature may include that the selection of the first model is special
Sign and the second model select feature, wherein the selection of the first model is characterized in having fully taken into account single model prediction result and history
The error between legitimate reading in creation data, the selection of the second model is characterized in for identifying historical production data essence
Feature.
Specifically, the first model selection is characterized in having fully taken into account in single model prediction result and historical production data
Error between legitimate reading, so as to consider the forecasting accuracy of each prediction single model.Wherein single model is predicted
It as a result is prediction result, and historical production data is then the characterization of legitimate reading, therefore can pass through prediction result and true knot
Fruit selects feature to construct the first model, such as can choose the training corresponding with historical production data of single model prediction result in advance
The ratio for surveying the legitimate reading of time selects feature as the first model.Optionally, first model selection feature can be each
Predicted that the prediction in the ring of single model period more year-on-year than the history of the prediction error in a upper period and/or each prediction single model missed
Difference.
Specifically, the selection of the second model is characterized in the feature for identifying historical production data essence, the selection of the second model
Feature can be to be calculated according to the historical production data before training predicted time, may include periodical power refers to
Mark and/or tendency relative strength index.Wherein periodical relative strength index is intended to indicate that periodic in historical production data, is become
Gesture relative strength index is intended to indicate that the tendency of historical production data, such as increases either decline etc..
S208: feature and prediction single model is selected to be trained to obtain model classifiers object module.
Specifically, the training process of model classifiers can be will prediction single model as Y value, then by constructed the
One model selects feature and the second model to select feature as X value, is learnt to obtain to the corresponding relationship between Y value and X value
Model classifiers.Terminal determines training predicted time first, then corresponding with training predicted time according to single model predicted value
Historical production data determines the corresponding Y value of training predicted time, so as to the corresponding relationship established between Y value and X value, by institute
Some Y values and X value, which are marked on, obtains multiple discrete points in coordinate system, then to multiple discrete points be fitted so as to
Obtain model classifiers.After obtaining model classifiers, terminal can be with the input prediction time, so that terminal can be pre- according to this
The survey time gets corresponding historical production data, and predicted time and historical production data are then input to model classifiers and obtained
To the corresponding prediction creation data of predicted time.
Above-mentioned model classifiers method for building up has fully considered a multiple single models when establishing model classifiers
Single model prediction result and historical production data, and construct to obtain according to single model prediction result and historical production data
First model selects feature, is constructed to obtain the second model selection feature according to historical production data, to be selected according to the first model
It selects feature and the second model selection feature is trained to have obtained model classifiers, such model classifiers have fully considered each
The feature of the characteristics of model and historical production data, so as to improve the accuracy of model prediction.
Feature and prediction single model is selected to be trained to obtain model point object module in one of the embodiments,
Class device may include: that legitimate reading is extracted from historical production data；Calculate the difference of legitimate reading and single model prediction result
Value obtains the smallest prediction single model of difference as optimum prediction single model；It is special to optimum prediction single model, object module selection
Sign is trained to obtain model classifiers.
Feature and prediction single model is selected to be trained to obtain model point object module in one of the embodiments,
Class device may include: that legitimate reading is extracted from historical production data；Calculate the ratio of single model prediction result and legitimate reading
Value obtains the weight of prediction single model according to ratio；Weight, the object module selection feature of prediction single model are trained
To model classifiers.
Specifically model classifiers can be divided into two kinds, and different according to the difference of Y value, wherein Y value can predict list
Model, the weight for being also possible to each prediction single model is then to pass through mould when wherein Y value is a prediction single model
Type classifier is chosen to optimum prediction single model, and Y value, which is the weight of each prediction single model, to be obtained by model classifiers
The weight of each prediction single model is got, to fully take into account the prediction result of each prediction single model.
Specifically, when Y value is a prediction single model, terminal can extract legitimate reading from historical production data,
I.e. with the corresponding legitimate reading of training predicted time, then legitimate reading is compared with single model prediction result, obtain and
The immediate single model prediction result of legitimate reading, and using the corresponding prediction single model of the single model prediction result as Y value, so
Y value and corresponding X value are learnt afterwards, such as Y value and X value is fitted to obtain each corresponding weight of X value, from
And it can establish model selector.
When Y value is the weight of each prediction single model, terminal can extract true knot from historical production data
Legitimate reading, is then compared with single model prediction result, passes through by fruit, i.e., legitimate reading corresponding with training predicted time
The ratio of each single model prediction result and legitimate reading it is available each prediction single model as optimal single model can
Energy property, i.e. weight, and optionally, terminal can be normalized using a prediction single model as a possibility that optimal single model
The corresponding weight of above-mentioned each prediction single model can be obtained, so that terminal learns those weights and corresponding X value
It can obtain model classifiers.In this way when inputted in model selector the first model selection feature and the second model selection feature
When, model classifiers output is each weight for predicting single model, such as the first prediction single model weight is 0.25, and second is pre-
Surveying single model weight is 0.15, and third predicts that single model weight is 0.2, and the 4th single model weight is 0.4.To which terminal can root
Prediction result to the end is obtained according to the single model prediction result and weight calculation of those prediction single models, i.e. predicted time is corresponding
Business revenue data.
In above-described embodiment, Y value can predict single model, be also possible to the weight of each prediction single model；When Y value is
When one prediction single model, then it can be and choose an optimal single model；When Y value be each prediction single model weight, then
The prediction result for fully taking into account each prediction single model, can guarantee through the obtained prediction result of model classifiers
Accuracy.
Training predicted time and historical production data are input in prediction single model in one of the embodiments, and obtained
It may include: to obtain training predicted time corresponding eigenperiod to single model prediction result；By predicting that single model calculates
To predicted value corresponding with eigenperiod；Processor execute realized when computer program according to single model prediction result and
Historical production data constructs the first model and selects feature, may include: corresponding with eigenperiod from historical production data extraction
True value；The first model selection feature is calculated according to predicted value and true value.Ring wherein be can be eigenperiod than upper one
Period or on year-on-year basis a upper period, when be ring than a upper period when, the selection of the first model is characterized in ring than predicting error, when being year-on-year
When a upper period, the selection of the first model was characterized in a year-on-year upper period.
The second model is constructed according to historical production data in one of the embodiments, and selects feature, may include: to obtain
Predetermined period length and section are segmented historical production data according to predetermined period length；It obtains corresponding in each section
Segmentation historical production data, and acquired historical production data is ranked up, and the history production after tag sort
The sequence valve of data；Calculate the deviation of the sequence valve of the historical production data of the corresponding segmentation in each section；Calculate deviation
Average value obtain periodical relative strength index as the second model select feature.
The second model is constructed according to historical production data in one of the embodiments, and selects feature, may include: to obtain
Predetermined period length is segmented historical production data according to predetermined period length；Number is produced according to the history of each segmentation
Increase and decrease amplitude is calculated according to the historical production data with a upper segmentation；The history production of each segmentation is obtained according to increase and decrease amplitude
The increase and decrease amplitude mark value of data；The average value for calculating increase and decrease amplitude mark value obtains tendency relative strength index as the second model
Select feature.
Specifically, the first model selection feature and the second model selection feature are related in above-mentioned model classifiers, and is
Convenience, hereinafter by two features and referred to as model selection feature is illustrated.I.e. model selection feature may include ring
Than the prediction error of upper period forecasting result, the prediction error of year-on-year upper period forecasting result, periodical relative strength index, become
Gesture relative strength index.
Its middle ring can refer to an either year in a season than the period in the prediction error of upper period forecasting result
Part, ring is for for each prediction single model and training predicted time than the prediction error of upper period forecasting result
, prediction error of the corresponding ring of each prediction single model than upper period forecasting result.Terminal is available to instruction
Practice predicted time, then got ring than a upper period, and by predicting that it is more corresponding than a upper period that ring is calculated in single model
First predicted value, and the first true value more corresponding than a upper period with ring, terminal and then basis were extracted from historical production data
The ratio of first predicted value and the first true value obtains prediction error of the ring than upper period forecasting result, i.e., the first above-mentioned mould
Type selects feature.
Specifically, a trained predicted time of single model is predicted with one of them to be illustrated, it is assumed that training prediction
Time is fourth quarter in 2018, then terminal gets prediction single model of the third season in 2018 corresponding first first and predicts
Value, then terminal extracts the first true value of the prediction single model of the third season in 2018 from historical production data, by the
Prediction error of the ratio of one predicted value and the first true value as ring than upper period forecasting result.
Wherein the period can refer to an either year in a season in the prediction error of upper period forecasting result on year-on-year basis
Part, the prediction error of year-on-year upper period forecasting result is for for each prediction single model and training predicted time
, the prediction error of the corresponding year-on-year upper period forecasting result of each prediction single model.Terminal is available to instruction
Practice predicted time, then gets a year-on-year upper period, and by predicting that it is corresponding that a year-on-year upper period was calculated in single model
Second predicted value, and the second true value corresponding with a upper period on year-on-year basis, terminal and then basis were extracted from historical production data
The ratio of second predicted value and the second true value obtains the prediction error of year-on-year upper period forecasting result, i.e., the first above-mentioned mould
Type selects feature.
Specifically, a trained predicted time of single model is predicted with one of them to be illustrated, it is assumed that training prediction
Time is fourth quarter in 2018, then terminal gets prediction single model of fourth quarter in 2017 corresponding second first and predicts
Value, then terminal extracts the second true value of the prediction single model of fourth quarter in 2017 from historical production data, by the
Prediction error of the ratio of two predicted values and the second true value as year-on-year upper period forecasting result.
Wherein, periodical relative strength index is for characterizing the periodic feature in historical production data.Terminal is obtaining
To after historical production data, predetermined period length and section are got, such as predetermined period length is season, section can be
Year, i.e. it may include multiple periods in a section.Terminal can be segmented historical production data according to cycle length, so
The historical production data in each section in corresponding segmentation is got afterwards and is ranked up, and the history after sequence is produced
Data are marked to obtain sequence valve, then calculate the inclined of the sequence valve of the historical production data of the corresponding segments in each section
Difference, the average value for calculating deviation obtain periodical relative strength index, i.e., the second above-mentioned model selects feature.
Specifically, using cycle length as season, siding-to-siding block length be the time for be illustrated, it is assumed that historical production data packet
2015,2016 and 2017 business revenue data are included, therefore terminal first carries out those business revenue data according to cycle length
Segmentation, is divided into the first quarter in 2015, the second quarter, the third season and fourth quarter, the first quarter in 2016, the second quarter,
The third season and fourth quarter and the first quarter in 2017, the second quarter, the third season and fourth quarter, then to each
Corresponding segmentation is ranked up in section, and the sequence valve of the segmentation after tag sort, specifically may refer to shown in the following table 1:
1 ranking results of table
| Sequence valve | The first quarter | The second quarter | The third season | Fourth quarter |
| 2015 | 1 | 2 | 3 | 2 |
| 2016 | 2 | 1 | 2 | 3 |
| 2017 | 3 | 3 | 1 | 1 |
As shown above, 2015 are first interval, and 2016 are second interval, and 2017 are 3rd interval, wherein often
The first quarter, the first quarter in 2016 and the first quarter in 2017 of corresponding segments i.e. 2015 year in one section, eventually
The historical production data that end can be segmented this 3 is ranked up, i.e., is ranked up to history business revenue data, and after tag sort
The sequence valve of each segmentation, such as maximum value are 1, are secondly 2, successively label goes down.In this way by the correspondence in each section point
Section all sequences are completed.
After the completion of sequence, terminal gets the sequence valve in each section, and deviation is calculated according to the sequence valve
Value, such as the sequence valve of above-mentioned corresponding first interval in 2015 includes 1,2,3,2, terminal calculates 1,2,3,2 deviation, example
1,2,3,2 standard deviation can be such as calculated, similarly, terminal can also calculate the standard deviation in other sections, finally by all marks
The average value of quasi- difference is as periodical relative strength index.
Wherein, tendency relative strength index is the tendency feature for characterizing historical production data.Terminal is being got
After historical production data, predetermined period length is got, such as predetermined period length is season.Terminal can be according to cycle length
Historical production data is segmented, then according to the historical production data of the historical production data of current fragment and a upper segmentation
Increase and decrease amplitude is calculated, and the historical production data of each segmentation can be marked according to the increase and decrease amplitude and be increased and decreased
Amplitude mark value, the average value that terminal finally calculates increase and decrease amplitude mark value obtain tendency relative strength index, i.e., and above-mentioned second
Model selects feature.
Specifically, it is illustrated so that cycle length is season as an example, it is assumed that historical production data includes 2015,2016
With business revenue data in 2017, therefore those business revenue data were segmented by terminal according to cycle length first, are divided into 2015
The first quarter, the second quarter, the third season and fourth quarter, the first quarter in 2016, the second quarter, the third season and the fourth season
Then increasing is calculated for each segmentation in degree and the first quarter, the second quarter, the third season and fourth quarter in 2017
Amount of decrease angle value, such as the slope of line of the business revenue data and business revenue data of the second quarter in 2015 of the first quarter in 2015 can
Using the increase and decrease range value as the second quarter in 2015, similarly, other segmentations, i.e., the increase and decrease in other periods can also be calculated
Range value.After increase and decrease range value has been calculated, increase and decrease amplitude mark value is calculated according to increase and decrease range value in terminal, such as works as increasing
When amount of decrease angle value is greater than 0, then it is+1 that it, which increases and decreases amplitude mark value, when increasing and decreasing range value less than 0, then increases and decreases amplitude mark value
It is -1, when the amplitude of increase and decrease is equal to 0, then increasing and decreasing amplitude mark value is 0.Terminal is getting all increase and decrease amplitude mark values
Afterwards, the average value of label amplitude mark value can be calculated, those average values are tendency relative strength index, i.e., the second above-mentioned mould
Type selects feature.
In above-described embodiment, ring can be calculated than a upper period according to historical production data and each prediction single model
The prediction error of prediction result, on year-on-year basis the prediction error of upper period forecasting result, periodical relative strength index, tendency power refer to
Mark, to carry out the training of model classifiers again, ensure that the accuracy of the prediction of model classifiers.
Specifically, to carry out in detail the training process of model trainer below in conjunction with above-mentioned model selection feature
Illustrate:
With above-mentioned ring than upper period forecasting result prediction error, on year-on-year basis upper period forecasting result prediction error,
Periodical relative strength index, tendency relative strength index select feature as model；The business revenue for being -2017 years 2015 with historical data
Data instance, such as when obtained history production business revenue is the history business revenue data of half a year, then terminal can be by this partly
The history business revenue data in year subtract the history business revenue data of the first quarter, so as to obtain the history business revenue number of the second quarter
According to terminal is so as to according to the above-mentioned periodicity relative strength index A of history business revenue data calculating and tendency power after standardization
Then index B calculates prediction of the corresponding ring of each prediction single model than upper period forecasting result according to training predicted time
The prediction error of error, on year-on-year basis upper period forecasting result, it is assumed that there are 4 prediction single models, then corresponding there are 4 rings than upper
Prediction error M, N, P, Q of prediction error C, D, E, F of one period forecasting result and 4 year-on-year upper period forecasting results,
Therefore model classifiers Y=a1*A+b1*B+c1*C+d1*D+e1*E+f1*F+m1*M+n1*N+p1*P+q1*Q is established, wherein
A1, b1, c1, d1, e1, f1, m1, n1, p1 and q1 are the weight that each model selects feature, available by fitting training
The value of a1, b1, c1, d1, e1, f1, m1, n1, p1 and q1, it is true in the value of a1, b1, c1, d1, e1, f1, m1, n1, p1 and q1
After fixed, then available model classifiers, and after obtaining model classifiers, when user has input predicted time, terminal can be with
The corresponding history business revenue data of the predicted time are got, predicted time and history business revenue data are then input to category of model
Prediction business revenue data corresponding with predicted time can be obtained in device.
In above-described embodiment, when establishing model classifiers, fully consider that the single model of a multiple single models is pre-
Result and historical production data are surveyed, and is constructed to have obtained the choosing of the first model according to single model prediction result and historical production data
Feature is selected, constructs to obtain the second model selection feature according to historical production data, to select feature and the according to the first model
The characteristics of two models selection feature is trained to have obtained model classifiers, and such model classifiers have fully considered each model
And the feature of historical production data, so as to improve the accuracy of model prediction.
It should be understood that although each step in the flow chart of Fig. 2 is successively shown according to the instruction of arrow, this
A little steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly state otherwise herein, these steps
It executes there is no the limitation of stringent sequence, these steps can execute in other order.Moreover, at least part in Fig. 2
Step may include that perhaps these sub-steps of multiple stages or stage are executed in synchronization to multiple sub-steps
It completes, but can execute at different times, the execution sequence in these sub-steps or stage, which is also not necessarily, successively to be carried out,
But it can be executed in turn or alternately at least part of the sub-step or stage of other steps or other steps.
In one embodiment, as shown in figure 3, providing a kind of model classifiers establishes device, comprising: historical data obtains
Modulus block 100, single model prediction result obtain module 200, feature construction module 300 and training module 400, in which:
Historical data obtains module 100, for obtaining historical production data.
Single model prediction result obtains module 200, for obtaining trained predicted time, and by training predicted time and goes through
History creation data, which is input in prediction single model, obtains single model prediction result, and prediction single model is that preparatory training obtains.
Feature construction module 300, for according to single model prediction result and historical production data building object module choosing
Select feature.
Training module 400, for being carried out to the first model selection feature, the second model selection feature and prediction single model
Training obtains model classifiers.
The second training module 400 may include: in one of the embodiments,
First extraction unit, for extracting legitimate reading from historical production data.
It is the smallest pre- to obtain difference for calculating the difference of legitimate reading and single model prediction result for first determination unit
Single model is surveyed as optimum prediction single model.
First training unit, for being trained to obtain model point to optimum prediction single model, object module selection feature
Class device.
Training module 400 may include: in one of the embodiments,
Second extraction unit, for extracting legitimate reading from historical production data.
Second determination unit is predicted for calculating the ratio of single model prediction result and legitimate reading according to ratio
The weight of single model.
Second training unit is trained to obtain model for weight, the object module selection feature to prediction single model
Classifier.
Feature construction module 300 may include: in one of the embodiments,
Fisrt feature construction unit, for constructing the choosing of the first model according to single model prediction result and historical production data
Select feature；
Second feature construction unit selects feature for constructing the second model according to historical production data.
Single model prediction result acquisition module 200 may include: in one of the embodiments,
Period 1 acquiring unit, for obtaining trained predicted time corresponding eigenperiod.
First predictor calculation unit, for by predicting that predicted value corresponding with eigenperiod is calculated in single model.
Fisrt feature construction unit may include:
True value acquiring unit, for extracting true value corresponding with eigenperiod from historical production data；
Computing unit, for the first model selection feature to be calculated according to the first predicted value and true value.
Second feature construction unit may include: in one of the embodiments,
First segmenting unit produces number to history according to predetermined period length for obtaining predetermined period length and section
According to being segmented.
First marking unit, for obtaining the historical production data of corresponding segmentation in each section, and to acquired
Historical production data is ranked up, and the sequence valve of the historical production data after tag sort.
Deviation computing unit, the deviation of the sequence valve of the historical production data for calculating the corresponding segmentation in each section
Value.
Periodical relative strength index computing unit, the average value for calculating deviation obtain periodical relative strength index as
Two models select feature.
Second feature construction unit may include: in one of the embodiments,
Second segmenting unit carries out historical production data according to predetermined period length for obtaining predetermined period length
Segmentation.
Second marking unit, based on according to the historical production data of each segmentation and the historical production data of a upper segmentation
Calculation obtains increase and decrease amplitude.
Increase and decrease range value computing unit, the increase and decrease width of the historical production data for obtaining each segmentation according to increase and decrease amplitude
Scale designation value.
Tendency relative strength index computing unit, the average value for calculating increase and decrease amplitude mark value obtain tendency power and refer to
It is denoted as selecting feature for the second model.
The specific restriction for establishing device about model classifiers may refer to above for model classifiers method for building up
Restriction, details are not described herein.Above-mentioned model classifiers establish the modules in device can be fully or partially through software, hard
Part and combinations thereof is realized.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment,
It can also be stored in a software form in the memory in computer equipment, execute the above modules in order to which processor calls
Corresponding operation.
In one embodiment, a kind of computer equipment is provided, which can be terminal, internal structure
Figure can be as shown in Figure 4.The computer equipment includes processor, the memory, network interface, display connected by system bus
Screen and input unit.Wherein, the processor of the computer equipment is for providing calculating and control ability.The computer equipment is deposited
Reservoir includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system and computer journey
Sequence.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The network interface of machine equipment is used to communicate with external terminal by network connection.When the computer program is executed by processor with
Realize a kind of model classifiers method for building up.The display screen of the computer equipment can be liquid crystal display or electric ink is aobvious
Display screen, the input unit of the computer equipment can be the touch layer covered on display screen, be also possible to computer equipment shell
Key, trace ball or the Trackpad of upper setting can also be external keyboard, Trackpad or mouse etc..
It will be understood by those skilled in the art that structure shown in Fig. 4, only part relevant to application scheme is tied
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment
It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored with
Computer program, the processor perform the steps of acquisition historical production data when executing computer program；Obtain training prediction
Time, and training predicted time and historical production data are input in prediction single model and obtain single model prediction result, in advance
Surveying single model is that preparatory training obtains；It is special according to single model prediction result and historical production data building object module selection
Sign；Feature and prediction single model is selected to be trained to obtain model classifiers object module.
In one embodiment, processor execute realized when computer program feature and pre- is selected to object module
It surveys single model to be trained to obtain model classifiers, may include: to extract legitimate reading from historical production data；It calculates true
As a result with the difference of single model prediction result, the smallest prediction single model of difference is obtained as optimum prediction single model；To optimal
Prediction single model, object module selection feature are trained to obtain model classifiers.
In one embodiment, processor execute realized when computer program feature and pre- is selected to object module
It surveys single model to be trained to obtain model classifiers, may include: to extract legitimate reading from historical production data；Calculate single mode
The ratio of type prediction result and legitimate reading obtains the weight of prediction single model according to ratio；Weight, mesh to prediction single model
Mark model selection feature is trained to obtain model classifiers.
That is realized when processor execution computer program in one of the embodiments, predicts to tie according to the single model
Fruit and historical production data building object module select feature, may include: raw according to single model prediction result and history
It produces data and constructs the first model selection feature；The second model, which is constructed, according to historical production data selects feature.
In one embodiment, that is realized when processor execution computer program will training predicted time and history life
Production data are input in prediction single model and obtain single model prediction result, may include: to obtain the corresponding spy of training predicted time
Levy the period；By predicting that predicted value corresponding with eigenperiod is calculated in single model；Processor executes computer program when institute
That realizes constructs the first model selection feature according to single model prediction result and historical production data, may include: from history
Creation data extracts true value corresponding with eigenperiod；It is special that the selection of the first model is calculated according to predicted value and true value
Sign.
In one embodiment, that is realized when processor execution computer program constructs second according to historical production data
Model selects feature, may include: to obtain predetermined period length and section, according to predetermined period length to historical production data into
Row segmentation；The historical production data of corresponding segmentation in each section is obtained, and acquired historical production data is arranged
Sequence, and the sequence valve of the historical production data after tag sort；Calculate the historical production data of the corresponding segmentation in each section
The deviation of sequence valve；The average value for calculating deviation obtains periodical relative strength index as the second model and selects feature.
In one embodiment, that is realized when processor execution computer program constructs second according to historical production data
Model selects feature, may include: to obtain predetermined period length, is divided according to predetermined period length historical production data
Section；Increase and decrease amplitude is calculated according to the historical production data of the historical production data of each segmentation and a upper segmentation；According to increasing
Amount of decrease degree obtains the increase and decrease amplitude mark value of the historical production data of each segmentation；The average value for calculating increase and decrease amplitude mark value obtains
Feature is selected as the second model to tendency relative strength index.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
Machine program performs the steps of acquisition historical production data when being executed by processor；Training predicted time is obtained, and will be trained pre-
It surveys the time and historical production data is input in prediction single model and obtains single model prediction result, prediction single model is to instruct in advance
It gets；Feature is selected according to single model prediction result and historical production data building object module；Object module is selected
It selects feature and prediction single model is trained to obtain model classifiers.
In one embodiment, realized when computer program is executed by processor to object module select feature and
Prediction single model is trained to obtain model classifiers, may include: that legitimate reading is extracted from historical production data；It calculates true
The difference of real result and single model prediction result obtains the smallest prediction single model of difference as optimum prediction single model；To most
Excellent prediction single model, object module selection feature are trained to obtain model classifiers.
In one embodiment, realized when computer program is executed by processor to object module select feature and
Prediction single model is trained to obtain model classifiers, may include: that legitimate reading is extracted from historical production data；It calculates single
The ratio of model prediction result and legitimate reading obtains the weight of prediction single model according to ratio；To prediction single model weight,
Object module selection feature is trained to obtain model classifiers.
That is realized when computer program is executed by processor in one of the embodiments, predicts according to the single model
As a result and historical production data building object module selects feature, may include: according to single model prediction result and history
Creation data constructs the first model and selects feature；The second model, which is constructed, according to historical production data selects feature.
In one embodiment, that is realized when computer program is executed by processor will training predicted time and history
Creation data is input in prediction single model and obtains single model prediction result, may include: to obtain to train predicted time corresponding
Eigenperiod；By predicting that predicted value corresponding with eigenperiod is calculated in single model；When processor executes computer program
That is realized constructs the first model selection feature according to single model prediction result and historical production data, may include: from going through
History creation data extracts true value corresponding with eigenperiod；It is special that the selection of the first model is calculated according to predicted value and true value
Sign.
In one embodiment, realized when computer program is executed by processor according to historical production data building the
Two models select feature, may include: to obtain predetermined period length and section, according to predetermined period length to historical production data
It is segmented；The historical production data of corresponding segmentation in each section is obtained, and acquired historical production data is carried out
Sequence, and the sequence valve of the historical production data after tag sort；Calculate the historical production data of the corresponding segmentation in each section
Sequence valve deviation；The average value for calculating deviation obtains periodical relative strength index as the second model and selects feature.
In one embodiment, realized when computer program is executed by processor according to historical production data building the
Two models select feature, may include: to obtain predetermined period length, are divided according to predetermined period length historical production data
Section；Increase and decrease amplitude is calculated according to the historical production data of the historical production data of each segmentation and a upper segmentation；According to increasing
Amount of decrease degree obtains the increase and decrease amplitude mark value of the historical production data of each segmentation；The average value for calculating increase and decrease amplitude mark value obtains
Feature is selected as the second model to tendency relative strength index.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of model classifiers method for building up, which comprises
Obtain historical production data；
Training predicted time is obtained, and the trained predicted time and the historical production data are input to prediction single model
In obtain single model prediction result, the prediction single model is that preparatory training obtains；
Feature is selected according to the single model prediction result and historical production data building object module；
Feature and the prediction single model is selected to be trained to obtain model classifiers the object module.
2. the method according to claim 1, wherein described to object module selection feature and described pre-
Single model is surveyed to be trained to obtain model classifiers, comprising:
Legitimate reading is extracted from the historical production data；
The difference for calculating the legitimate reading and the single model prediction result obtains the smallest prediction single model of the difference and makees
For optimum prediction single model；
The optimum prediction single model, object module selection feature are trained to obtain model classifiers.
3. the method according to claim 1, wherein described select feature, the second model to first model
Selection feature and the prediction single model are trained to obtain model classifiers, comprising:
Legitimate reading is extracted from the historical production data；
The ratio for calculating the single model prediction result and the legitimate reading obtains the prediction single model according to the ratio
Weight；
Weight, the object module selection feature of the prediction single model are trained to obtain model classifiers.
4. according to claim 1 to method described in 3 any one, which is characterized in that described to predict to tie according to the single model
Fruit and historical production data building object module select feature, comprising:
The first model, which is constructed, according to the single model prediction result and the historical production data selects feature；
The second model, which is constructed, according to the historical production data selects feature.
5. according to the method described in claim 4, it is characterized in that, described that the trained predicted time and the history is raw
Production data, which are input in the prediction single model, obtains single model prediction result, comprising:
Obtain the trained predicted time corresponding eigenperiod；
Predicted value corresponding with the eigenperiod is calculated by the prediction single model；
It is described that first model selection feature is constructed according to the single model prediction result and the historical production data, comprising:
True value corresponding with the eigenperiod is extracted from the historical production data；
The first model selection feature is calculated according to the predicted value and the true value.
6. according to the method described in claim 4, it is characterized in that, described construct the second model according to the historical production data
Select feature, comprising:
Predetermined period length and section are obtained, the historical production data is segmented according to the predetermined period length；
The historical production data of corresponding segmentation in each section is obtained, and acquired historical production data is ranked up,
And the sequence valve of the historical production data after tag sort；
Calculate the deviation of the sequence valve of the historical production data of the corresponding segmentation in each section；
The average value for calculating the deviation obtains periodical relative strength index as the second model and selects feature.
7. according to the method described in claim 4, it is characterized in that, described construct the second model according to the historical production data
Select feature, comprising:
Predetermined period length is obtained, the historical production data is segmented according to the predetermined period length；
Increase and decrease amplitude is calculated according to the historical production data of the historical production data of each segmentation and a upper segmentation；
The increase and decrease amplitude mark value of the historical production data of each segmentation is obtained according to the increase and decrease amplitude；
The average value for calculating the increase and decrease amplitude mark value obtains tendency relative strength index as the second model and selects feature.
8. a kind of model classifiers establish device, which is characterized in that described device includes:
Historical data obtains module, for obtaining historical production data；
Single model prediction result obtains module, for obtaining trained predicted time, and by the trained predicted time and described
Historical production data, which is input in prediction single model, obtains single model prediction result, and the prediction single model is that preparatory training obtains
's；
Fisrt feature constructs module, for constructing target mould according to the single model prediction result and the historical production data
Type selects feature；
Training module, for selecting feature and the prediction single model to be trained to obtain category of model the object module
Device.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 7 the method when executing the computer program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 7 is realized when being executed by processor.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811555208.4A CN109767031A (en) | 2018-12-19 | 2018-12-19 | Model classifiers method for building up, device, computer equipment and storage medium |
| PCT/CN2019/093146 WO2020124977A1 (en) | 2018-12-19 | 2019-06-27 | Method and apparatus for processing production data, computer device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811555208.4A CN109767031A (en) | 2018-12-19 | 2018-12-19 | Model classifiers method for building up, device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109767031A true CN109767031A (en) | 2019-05-17 |
Family
ID=66450250
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811555208.4A Pending CN109767031A (en) | 2018-12-19 | 2018-12-19 | Model classifiers method for building up, device, computer equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109767031A (en) |
| WO (1) | WO2020124977A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110188490A (en) * | 2019-06-03 | 2019-08-30 | 珠海格力电器股份有限公司 | Method and device for improving data simulation efficiency, storage medium and electronic device |
| WO2020124977A1 (en) * | 2018-12-19 | 2020-06-25 | 平安科技（深圳）有限公司 | Method and apparatus for processing production data, computer device, and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103136638A (en) * | 2011-11-22 | 2013-06-05 | 上海宝钢钢材贸易有限公司 | Adaptive adjustment method of steel production |
| CN102968670B (en) * | 2012-10-23 | 2016-08-17 | 北京京东世纪贸易有限公司 | The method and apparatus of prediction data |
| CN108230049A (en) * | 2018-02-09 | 2018-06-29 | 新智数字科技有限公司 | The Forecasting Methodology and system of order |
| CN109767031A (en) * | 2018-12-19 | 2019-05-17 | 平安科技（深圳）有限公司 | Model classifiers method for building up, device, computer equipment and storage medium |
-
2018
- 2018-12-19 CN CN201811555208.4A patent/CN109767031A/en active Pending
-
2019
- 2019-06-27 WO PCT/CN2019/093146 patent/WO2020124977A1/en active Application Filing
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020124977A1 (en) * | 2018-12-19 | 2020-06-25 | 平安科技（深圳）有限公司 | Method and apparatus for processing production data, computer device, and storage medium |
| CN110188490A (en) * | 2019-06-03 | 2019-08-30 | 珠海格力电器股份有限公司 | Method and device for improving data simulation efficiency, storage medium and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020124977A1 (en) | 2020-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109255506B (en) | Internet financial user loan overdue prediction method based on big data | |
| CN110390465A (en) | Air control analysis and processing method, device and the computer equipment of business datum | |
| CN109360105A (en) | Product risks method for early warning, device, computer equipment and storage medium | |
| Flores et al. | Using artificial intelligence techniques to improve the prediction of copper recovery by leaching | |
| CN109583468A (en) | Training sample acquisition methods, sample predictions method and corresponding intrument | |
| CN104572449A (en) | Automatic test method based on case library | |
| Weintraub et al. | A priori and a posteriori aggregation procedures to reduce model size in MIP mine planning models | |
| CN109615280A (en) | Employee's data processing method, device, computer equipment and storage medium | |
| CN109583682A (en) | Recognition methods, device and the computer equipment of business finance fraud risk | |
| CN111738504A (en) | Enterprise financial index fund amount prediction method and device, equipment and storage medium | |
| US20210090101A1 (en) | Systems and methods for business analytics model scoring and selection | |
| CN109801151A (en) | Financial fraud risk monitoring and control method, apparatus, computer equipment and storage medium | |
| CN110310012A (en) | Data analysing method, device, equipment and computer readable storage medium | |
| CN109767031A (en) | Model classifiers method for building up, device, computer equipment and storage medium | |
| CN110909975B (en) | Scientific research platform benefit evaluation method and device | |
| CN115130887A (en) | Reservoir dam environmental impact evaluation method and device, electronic equipment and storage medium | |
| CN108764553B (en) | User scale prediction method and device and computer equipment | |
| Zhou et al. | A shale gas production prediction model based on masked convolutional neural network | |
| CN111325344A (en) | Method and apparatus for evaluating model interpretation tools | |
| CN109829745A (en) | Business revenue data predication method, device, computer equipment and storage medium | |
| CN109767263A (en) | Business revenue data predication method, device, computer equipment and storage medium | |
| CN106709621A (en) | Technical trend prediction method and system | |
| CN113689020A (en) | Service information prediction method, device, computer equipment and storage medium | |
| CN109697528A (en) | Business revenue data predication method, device, computer equipment and storage medium | |
| CN109783876A (en) | Time series models method for building up, device, computer equipment and storage medium |
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: 20190517 |
|
| RJ01 | Rejection of invention patent application after publication |