CN110503447A - For determining the method and device of Sales Volume of Commodity predicted value - Google Patents
For determining the method and device of Sales Volume of Commodity predicted value Download PDFInfo
- Publication number
- CN110503447A CN110503447A CN201810466256.XA CN201810466256A CN110503447A CN 110503447 A CN110503447 A CN 110503447A CN 201810466256 A CN201810466256 A CN 201810466256A CN 110503447 A CN110503447 A CN 110503447A
- Authority
- CN
- China
- Prior art keywords
- sales
- data
- sample set
- machine learning
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application provides a kind of methods for determining Sales Volume of Commodity predicted value, comprising: obtains the article sales data in predetermined amount of time, the article sales data includes at least sales date and sales volume；Based on acquired article sales data, training sample set is generated；One or more machine learning models are trained using training sample set generated；And Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training.Using this method, the accuracy of merchandise sales prediction can be improved.
Description
Technical field
The application is usually directed to field of computer technology, more particularly, to one kind for determining Sales Volume of Commodity predicted value
Method and device.
Background technique
In merchandise sales enterprise, Method for Sales Forecast is the important link in enterprise marketing decision and strategic management, for example, pin
Selling prediction can be applied to commodity Replenishment Policy, commodity inventory control, commodity production Project Planning etc..Since Sales Volume of Commodity is looked forward to
Many factors in industry internal environment and external environment influence, and the accuracy of Method for Sales Forecast is unsatisfactory, to how to improve quotient
Product Method for Sales Forecast accuracy becomes merchandise sales field and expects to solve the problems, such as.
Summary of the invention
In view of the above problems, this application provides a kind of for determining the method and device of Sales Volume of Commodity predicted value.It utilizes
This method and device, by using the training sample set generated based on article sales data come to one or more machine learning moulds
Type is trained；And Sales Volume of Commodity predicted value, Ke Yiti are obtained based on one or more machine learning models after training
The accuracy of high merchandise sales prediction.
According to the one aspect of the application, a kind of method for determining Sales Volume of Commodity predicted value is provided, comprising: obtain
Article sales data in predetermined amount of time, the article sales data include at least sales date and sales volume；Based on being obtained
The article sales data taken generates training sample set；Using training sample set generated come to one or more machine learning
Model is trained；And Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training.
Optionally, in an example of above-mentioned aspect, using training sample set generated come to one or more machines
Device learning model, which is trained, to be realized based on bagging algorithm, and based on one or more machine learning after training
Model is come to obtain Sales Volume of Commodity predicted value may include: to obtain each machine in one or more of machine learning models
Candidate commodity Method for Sales Forecast value under learning model；And calculate being averaged for acquired each candidate commodity Method for Sales Forecast value
Value, as Sales Volume of Commodity predicted value.
Optionally, in an example of above-mentioned aspect, using training sample set generated come to one or more machines
Device learning model, which is trained, to be realized based on boosting algorithm, and based on one or more machine learning after training
Model is come to obtain Sales Volume of Commodity predicted value may include: to obtain each machine in one or more of machine learning models
Candidate commodity Method for Sales Forecast value and corresponding prediction weighted value under learning model；And to acquired each candidate commodity
Method for Sales Forecast value carries out weight adduction, to obtain Sales Volume of Commodity predicted value.
Optionally, in an example of above-mentioned aspect, the machine learning model may include at least one in following
Kind: Random Forest model, extreme gradient decline model, quantile Random Forest model and supporting vector machine model.
Optionally, in an example of above-mentioned aspect, training sample set is generated based on acquired article sales data
It may include: that acquired article sales data is converted into model available feature data；And spy can be used based on the model
It levies data and generates training sample set.
Optionally, in an example of above-mentioned aspect, based on acquired article sales data, training sample set is generated
It can also include: Feature Dimension Reduction processing to be carried out to the model available feature data, and be based on the model available feature number
It may include: to generate instruction based on by the Feature Dimension Reduction treated model available feature data according to training sample set is generated
Practice sample set.
Optionally, in an example of above-mentioned aspect, the Feature Dimension Reduction processing may include: available from the model
The characteristic information unrelated with sales volume is removed in characteristic；It is more than predetermined that the degree of correlation is removed from the model available feature data
The characteristic information of threshold value；And/or characteristic information elimination is carried out to the model available feature data using random forests algorithm.
Optionally, in an example of above-mentioned aspect, the method can also include: to acquired merchandise sales number
According to progress data prediction.
Optionally, in an example of above-mentioned aspect, the method can also include: to use training sample generated
Collection is adjusted the model parameter of one or more of machine learning models.
Optionally, in an example of above-mentioned aspect, the model parameter adjustment is using first heuristic algorithm and/or shellfish
Leaf this optimisation technique is realized.
Optionally, in an example of above-mentioned aspect, the article sales data can also include it is following at least
It is a kind of: selling price；Sales promotion information；Macroeconomic data information；And Weather information.
It according to the another aspect of the application, provides a kind of for determining the device of Sales Volume of Commodity predicted value, comprising: sale
Data capture unit, for obtaining the article sales data in predetermined amount of time, the article sales data includes at least sale
Date and sales volume；Sample set generation unit, for generating training sample set based on acquired article sales data；Training
Unit, for being trained using training sample set generated to one or more machine learning models；And prediction is single
Member, for obtaining Sales Volume of Commodity predicted value based on one or more machine learning models after training.
Optionally, in an example of above-mentioned aspect, the training unit is based on bagging algorithm or boosting is calculated
Method trains one or more of machine learning models.
Optionally, in an example of above-mentioned aspect, the sample set generation unit may include: characteristic conversion
Module, for acquired article sales data to be converted to model available feature data；And sample set generation module, it is used for
Training sample set is generated based on the model available feature data.
Optionally, in an example of above-mentioned aspect, the sample set generation unit can also include: at Feature Dimension Reduction
Module is managed, for being used for model available feature data progress Feature Dimension Reduction processing and the sample set generation module:
Based on by the Feature Dimension Reduction treated model available feature data, training sample set is generated.
Optionally, in an example of above-mentioned aspect, described device can also include: model parameter adjustment unit, use
The model parameter of one or more of machine learning models is adjusted in using training sample set generated.
According to the another aspect of the application, a kind of calculating equipment is provided, comprising: one or more processors, Yi Jiyu
The memory of one or more of processor couplings, the memory store instruction, when described instruction is by one or more
When a processor executes, so that one or more of processors execute as described above for determining Sales Volume of Commodity predicted value
Method.
According to the another aspect of the application, a kind of non-transitory machinable medium is provided, being stored with can hold
Row instruction, described instruction execute the machine as described above for determining the side of Sales Volume of Commodity predicted value
Method.
Detailed description of the invention
By referring to following attached drawing, may be implemented to further understand the nature and advantages of present disclosure.In
In attached drawing, similar assembly or feature can have identical appended drawing reference.
Fig. 1 shows the flow chart of the method for determining Sales Volume of Commodity predicted value according to the application；
Fig. 2 shows an examples of the information list according to included by the article sales data of the application；
Fig. 3 shows an exemplary flow chart of the generating process of training sample set；
Fig. 4 shows the instantiation procedure that machine learning model training is carried out based on bagging algorithm according to the application
Flow chart；
Fig. 5, which is shown, is carrying out the Sales Volume of Commodity predicted value under machine learning model training based on bagging algorithm
The flow chart of the instantiation procedure of acquisition；
Fig. 6 shows the instantiation procedure that machine learning model training is carried out based on boosting algorithm according to the application
Flow chart；
Fig. 7, which is shown, is carrying out the Sales Volume of Commodity predicted value under machine learning model training based on boosting algorithm
The flow chart of the instantiation procedure of acquisition；
Fig. 8 shows the block diagram of the device for determining merchandise sales predicted value according to the application；
Fig. 9 shows the block diagram of an implementation example of the sample set generation unit in Fig. 8；With
Figure 10 shows the block diagram of the calculating equipment for determining merchandise sales predicted value according to the application.
Specific embodiment
Theme described herein is discussed referring now to example embodiment.It should be understood that discussing these embodiments only
It is in order to enable those skilled in the art can better understand that being not to claim to realize theme described herein
Protection scope, applicability or the exemplary limitation illustrated in book.It can be in the protection scope for not departing from present disclosure
In the case of, the function and arrangement of the element discussed are changed.Each example can according to need, omit, substitute or
Add various processes or component.For example, described method can be executed according to described order in a different order, with
And each step can be added, omits or combine.In addition, feature described in relatively some examples is in other examples
It can be combined.
As used in this article, term " includes " and its modification indicate open term, are meant that " including but not limited to ".
Term "based" indicates " being based at least partially on ".Term " one embodiment " and " embodiment " expression " at least one implementation
Example ".Term " another embodiment " expression " at least one other embodiment ".Term " first ", " second " etc. may refer to not
Same or identical object.Here may include other definition, either specific or implicit.Unless bright in context
It really indicates, otherwise the definition of a term is consistent throughout the specification.
Fig. 1 shows the flow chart of the method for determining Sales Volume of Commodity predicted value according to the application.
As shown in Figure 1, obtaining the article sales data in predetermined amount of time, the article sales data is extremely in block S110
It less include sales date and corresponding sales volume.The acquisition of article sales data can be acquired using any suitable data
Equipment/device is realized.Here, the value of predetermined amount of time can be preset by user, for example, 2 years.It is described predetermined
It is different and different that the setting value of period can be the type based on commodity.
Optionally, in the example of the application, article sales data can also include selling price.
Optionally, in the example of the application, article sales data can also include sales promotion information.The promotion letter
Breath is the information for describing commercial promotions feature.For example, the sales promotion information may include: promotion period information；Promote class
Type；And/or the information such as promotion dynamics.For example, the promotion period information can be from this Monday to this Friday, promotional form can
To be to promote (alternatively, buying two send a promotion) at a reduced price, promotion dynamics is 7 foldings.
Optionally, in another example of the application, article sales data can also include macroeconomic data information.Institute
Stating macroeconomic data information is the information for the macroeconomic data during merchandise sales.For example, the macroeconomy number
It is believed that breath is such as may include national GDP value, oil price, stock market.
Optionally, in another example of the application, article sales data can also include Weather information.The weather letter
Breath for example may include weather pattern, temperature, humidity and/or rainfall.
Fig. 2 shows the schematic diagrames according to the information list of the article sales data of the application.As shown in Figure 2, in institute
In the information of illustration, sales volume date and sales volume are essential information, and sales promotion information, macroeconomic data information and Weather information
It is optional information.
After getting Sales Volume of Commodity data, training sample is generated based on acquired Sales Volume of Commodity data in block S120
Collection.Optionally, in the example of the application, acquired Sales Volume of Commodity data are also based on, generate test sample collection.
Here, training sample set refers to that sample set and test sample collection for training machine learning model refer to and is used to pair
The sample set that the machine learning model trained is tested.For example, according to one embodiment of the application, training sample set and
Test sample collection can be in the sample obtained from based on acquired Sales Volume of Commodity data and extract and obtain in proportion, than
Such as, training sample set occupies 90% sample and the sample of test sample collection occupancy remaining 10%；Or training sample set accounts for
Remaining 20% sample is occupied with 60% sample and test sample collection.The sample and test sample collection that training sample is concentrated
In sample cannot repeat.
Optionally, according to one embodiment of the application, being in the format of Sales Volume of Commodity data cannot be by machine learning mould
In the case where the data format that type uses, it is also necessary to carry out Feature Conversion processing to Sales Volume of Commodity data.Fig. 3 shows trained sample
One exemplary flow chart of the generating process of this collection.
As shown in figure 3, firstly, acquired article sales data is converted to model available feature data in block S121.
Here, model available feature data refer to the identifiable data of machine learning model, are usually to have model specification structure
Structural data.Model available feature data are usually the structural data with multiple dimensions.For example, if the quotient got
Product sales data is that " sales volume on Tuesday of first month second week is 100, and the temperature on the same day is 35 degrees Celsius, selling price
It is 100 yuan, which has advertising campaign, and promotion amplitude is eight foldings ", then the model available feature data after converting are (moon=1, week
=2, day=2, sales volume=100, temperature=35 degree Celsius, selling price=100 yuan, advertising campaign=1, promotion amplitude=
0.8)。
Then, in block S123, the model available feature data are based on, generate training sample set.
After as above generating training sample set, in block S130, using training sample set generated come to one or more
Machine learning model is trained.According to one embodiment of the application, the machine learning model may include in following
It is at least one: Random Forest model, extreme gradient decline model, quantile Random Forest model and supporting vector machine model.
Then, in block S140, Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training.
In the example of the application, using training sample set generated come to one or more machine learning models
Being trained can be based on the realization of bagging algorithm.Fig. 4 show according to the application based on bagging algorithm come into
The flow chart of the instantiation procedure of row machine learning model training.
As shown in figure 4, firstly, being trained specimen sample for each machine learning model in block S131.Show at one
In example, it can use and can put back to the mode of grab sample to former training set to be trained specimen sample.For example, it is assumed that depositing
In n machine learning model, each machine learning model needs k sample.K sampling so is carried out to former training set, is obtained
For the first training set of the first machine learning model, sample is then put back into former concentration.Then, former training set is carried out again
K sampling, to obtain the second training set for the second machine learning module.As above operation n times are repeated, to be directed to
N training set of the n machine learning model.
Then, in block S133, corresponding machine learning model is trained using obtained n training set.
Fig. 5, which is shown, is carrying out the Sales Volume of Commodity predicted value under machine learning model training based on bagging algorithm
The flow chart of the instantiation procedure of acquisition.
As shown in figure 5, in block S141, based on each machine learning module by the training of training process shown in Fig. 4,
Obtain corresponding candidate commodity Method for Sales Forecast value.Then, in block S143, each candidate commodity Method for Sales Forecast value obtained is calculated
Average value, as Sales Volume of Commodity predicted value.
In the example of the application, using training sample set generated come to one or more machine learning models
Being trained can be based on the realization of boosting algorithm.According to one embodiment of the application, Boosting algorithm can be with
Including XGBoost, Gradient Boosting etc..Fig. 6 is shown according to the application based on boosting algorithm come the machine of progress
The flow chart of the instantiation procedure of device learning model training.
As shown in fig. 6, all assigning same weight for all samples in block S131 '.Then, it executes from block S133 '
To the multiple circulate operation of S137 '.
Specifically, in block S133 ', current training sample training "current" model is used.In block S135 ', it is determined whether be directed to
All machine learning models are completed aforesaid operations or the precision of machine learning model reaches specified precision.Such as gynophore
The precision for being completed aforesaid operations or machine learning model to all machine learning models reaches specified precision, then holds
The operation of block S140 of the row into Fig. 1.Otherwise proceed to block S137 '.In block S137 ', adjusted based on the training result of model
The weight of training sample, for example, the weight of the biggish sample of test error is improved.In one embodiment, can every time by
Weight improves fixed setting value.Equally, using the training sample after being adjusted as current training sample, and to next machine
Learning model then proceedes to block S133 ' as current machine learning model.
Finally by the predicted value of learner according to a ratio (this ratio is determined by the global error level of learner)
The predicted value for adding up to the end.
Fig. 7, which is shown, is carrying out the Sales Volume of Commodity predicted value under machine learning model training based on boosting algorithm
The flow chart of the instantiation procedure of acquisition.
As shown in fig. 7, first in block S141 ', each engineering in one or more machine learning models is being stated in acquisition
Practise the candidate commodity Method for Sales Forecast value under model and corresponding prediction weighted value.Here, prediction weighted value is based on engineering
The global error of model prediction is practised to determine.Then, in block S143 ', to acquired each candidate commodity Method for Sales Forecast value
Weight adduction is carried out, to obtain Sales Volume of Commodity predicted value.
It optionally, in addition, can also include to the model before block S123 in one embodiment of the application
Available feature data carry out dimension-reduction treatment.According to one embodiment of the application, the Feature Dimension Reduction processing may include: from mould
The characteristic information unrelated with sales volume is removed in type available feature data；It is more than pre- that the degree of correlation is removed from model available feature data
Determine the characteristic information of threshold value；And/or characteristic information elimination is carried out to model available feature information using random forests algorithm.
When the dimension-reduction treatment is to remove the characteristic information unrelated with sales volume from model available feature data, for example,
Can be used single argument feature selection approach, remove in multiple features of model available feature data with the obvious not phase of sales volume feature
The feature of pass.Alternatively, can be commented for each feature in model available feature data using statistical checks such as Chi-square Tests
The relationship between this feature and sales volume feature is estimated to calculate score, then removes the feature that score is lower than predetermined value, to drop
The dimension of the characteristic information of low model available feature data.
It is that the characteristic information that the degree of correlation is more than predetermined threshold is removed from model available feature data in the dimension-reduction treatment
When, it can use the linearly related degree that Pearson's coefficient comes between each feature in computation model available feature data, then
One characteristic information of selection (for example, it may be random selection) from the characteristic information that linearly related degree is greater than predetermined threshold, together
When delete other feature information, to reduce the dimension of the characteristic information of model available feature data.
It is that characteristic information elimination is carried out to model available feature information using random forests algorithm in the dimension-reduction treatment
When, it can use random forests algorithm to calculate feature importance index, be then based on the calculated feature importance of institute and refer to
Mark is to be screened, to reduce the dimension of the characteristic information of model available feature data.In the case where correlated characteristic is more,
It can also be right using principal component analysis (principal components analysis, PCA) and the technology of deep learning
Characteristic information carries out reasonable dimensionality reduction.
It optionally, in addition, can be with before acquired article sales data is converted to model available feature data
Including carrying out data prediction to acquired article sales data.According to one embodiment of the application, the data are located in advance
Reason may include that abnormal data screens out and/or missing values are filled up.
In this application, during article sales data obtains, if an article sales data obtains acquired in point
Significant changes (such as mistake occurs for the average article sales data during article sales data is obtained relative to the article sales data
It is big or too small) when, it is believed that it is abnormal data that the article sales data, which obtains the acquired article sales data of point,.For example, can be with
It will be greater than the merchandise sales number of the prearranged multiple (for example, 3 times) of the average article sales data during article sales data obtains
According to being determined as abnormal data.Alternatively, can be by the predetermined of the average article sales data during being less than article sales data acquisition
The article sales data of percentage (for example, 10%) is determined as abnormal data.In such case, the data prediction be can wrap
It includes abnormal data to screen out, for screening out such abnormal data from acquired article sales data.
In this application, occur missing values in article sales data and refer to that some information in article sales data does not count
According to for example, being directed to every article sales data, the data for needing to acquire include three fields: selling time, selling price and pin
Sell quantity.However, when acquiring sales data, it is possible that the value of the certain field in three fields is sky.In this feelings
Under condition, the data prediction may include filling up to the missing values.For example, it is assumed that the collected historic sales data of institute
As follows: first historic sales data is " 2017-1-1,2,3.5 yuan ", Article 2 historic sales data be " 2017-1-2,4
Part, 3.4 yuan ", Article 3 historic sales data is " 2017-1-4,3.7 yuan " and Article 4 historic sales data is
" 2017-1-5,3,3.5 yuan "；The then numerical value missing of the sales volume in Article 3 historic sales data.In this case, it needs
Missing values are carried out to the value of the sales volume in Article 3 historic sales data to fill up.A variety of sides can be used by filling up missing values
Formula, it can be common that replaced using the mean value of non-missing values under the field, such as in above-mentioned example, the sales volume of 2017-1-4 can be with
It is replaced with (2+4+3)/3=3.
It optionally, in addition, can also include: using institute according to the method for determining merchandise sales predicted value of the application
The training sample set of generation is adjusted the model parameter of one or more of machine learning models.According to the application's
One embodiment, model parameter adjustment, which can be using first heuristic algorithm and/or Bayes's optimisation technique, to be realized.Institute
Stating first heuristic algorithm may include genetic algorithm.Optionally, in addition, it is also based on the article sales data newly obtained, to institute
The model parameter for stating one or more machine learning models is adjusted.For example, can be based on the article sales data newly obtained
New training sample set is obtained, new training sample set is then based on and carrys out mould to one or more of machine learning models
Shape parameter is adjusted.
The method for determining merchandise valuation according to the application is described above with reference to Fig. 1 to Fig. 7.Below in conjunction with figure
8 to Fig. 9 descriptions are according to the application for determining the device of Sales Volume of Commodity predicted value.
Fig. 8 is shown according to the application for determining device (the hereinafter referred to Sales Volume of Commodity of Sales Volume of Commodity predicted value
Predicted value determining device 800) block diagram.As shown in figure 8, Sales Volume of Commodity predicted value determining device 800 includes that sales volume data obtain
Take unit 810, sample set generation unit 820, training unit 830 and Sales Volume of Commodity predicted value determination unit 840.
Sales data acquiring unit 810 is used to obtain the article sales data in predetermined amount of time, the merchandise sales number
According to including at least sales date and sales volume.The operation of sales data acquiring unit 810 is similar to above with reference to the block in Fig. 1
The operation of S110.
Sample set generation unit 820 is used to generate training sample set based on acquired article sales data.Sample set is raw
The operation above with reference to the block S120 in Fig. 1 is similar at the operation of unit 820.
Training unit 830 is for instructing one or more machine learning models using training sample set generated
Practice.The operation of training unit 830 is similar to the operation above with reference to the block S130 in Fig. 1.
Predicting unit 840 is for obtaining Sales Volume of Commodity prediction based on one or more machine learning models after training
Value.The operation of predicting unit 840 is similar to the operation above with reference to the block S140 in Fig. 1.
Optionally, in addition, according to one embodiment of the application, sample set generation unit 820 be can be also used for based on institute
The article sales data of acquisition generates test sample collection.Sales Volume of Commodity predicted value determining device 800 can also include model parameter
Adjustment unit 850, for being joined using training sample set generated come the model to one or more of machine learning models
Number is adjusted.According to one embodiment of the application, the model parameter adjustment be can be using first heuristic algorithm and/or shellfish
Leaf this optimisation technique is realized.The member heuristic algorithm may include genetic algorithm.
Optionally, in addition, according to one embodiment of the application, Sales Volume of Commodity predicted value determining device 800 can also be wrapped
Data pre-processing unit is included, for carrying out data prediction to acquired article sales data.The data prediction can be with
It is screened out including abnormal data and/or missing values is filled up.
Fig. 9 shows the block diagram of an implementation example of the sample set generation unit 820 in Fig. 8.As shown in figure 8, sample
This collection generation unit 820 may include characteristic conversion module 821 and sample set generation module 825.
Characteristic conversion module 821 is used to acquired article sales data being converted to model available feature data.
The operation of characteristic conversion module 821 is similar to the operation of the block S121 referring to described in Fig. 3.
Sample set generation module 825 be used for be based on the model available feature data, generate training sample set and.Sample set
The operation of generation module 825 is similar to the operation of the block S123 referring to described in Fig. 3.
Optionally, in addition, sample set generation unit 820 can also include Feature Dimension Reduction processing module 823.At Feature Dimension Reduction
Module 823 is managed to be used to carry out Feature Dimension Reduction processing to the model available feature data.Then, 825 base of sample set generation module
In by the Feature Dimension Reduction treated model available feature data, training sample set is generated.
Above with reference to Fig. 1 to Fig. 9, to the reality according to the method and apparatus for determining Sales Volume of Commodity predicted value of the application
Example is applied to be described.Sales Volume of Commodity predicted value determining device above can use hardware realization, can also using software or
The combination of person's hardware and software is realized.
In this application, Sales Volume of Commodity predicted value determining device, which can use, calculates equipment realization.Figure 10 shows basis
The block diagram of the calculating equipment 1000 for determining Sales Volume of Commodity predicted value of embodiments herein.According to one embodiment,
Calculating equipment 1000 may include one or more processors 1010, which executes can in computer
It reads to store in storage medium (that is, memory 1020) or one or more computer-readable instructions of coding is (that is, above-mentioned with software
The element that form is realized).
In one embodiment, computer executable instructions are stored in memory 1020, make one when implemented
Or multiple processors 1010: obtaining the article sales data in predetermined amount of time, and the article sales data includes at least sale
Date and sales volume；Based on acquired article sales data, training sample set is generated；Come using training sample set generated
One or more machine learning models are trained；And it is obtained based on one or more machine learning models after training
Sales Volume of Commodity predicted value.
It should be understood that the computer executable instructions stored in memory 1020 make one or more places when implemented
It manages device 1010 and carries out the above various operations and functions described in conjunction with Fig. 1-9 in each embodiment of the application.
According to one embodiment, a kind of program product of such as non-transitory machine readable media is provided.Non-transitory
Machine readable media can have instruction (that is, above-mentioned element realized in a software form), which when executed by a machine, makes
It obtains machine and executes the above various operations and functions described in conjunction with Fig. 1-9 in each embodiment of the application.Specifically, Ke Yiti
For being furnished with the system or device of readable storage medium storing program for executing, store on the readable storage medium storing program for executing any in realization above-described embodiment
The software program code of the function of embodiment, and read and execute the computer of the system or device or processor and be stored in
Instruction in the readable storage medium storing program for executing.
In this case, it is real that any one of above-described embodiment can be achieved in the program code itself read from readable medium
The function of example is applied, therefore the readable storage medium storing program for executing of machine readable code and storage machine readable code constitutes of the invention one
Point.
The embodiment of readable storage medium storing program for executing include floppy disk, hard disk, magneto-optic disk, CD (such as CD-ROM, CD-R, CD-RW,
DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), tape, non-volatile memory card and ROM.It selectively, can be by communication network
Network download program code from server computer or on cloud.
It will be appreciated by those skilled in the art that each embodiment disclosed above can be in the situation without departing from invention essence
Under make various changes and modifications.Therefore, protection scope of the present invention should be defined by the appended claims.
It should be noted that step and unit not all in above-mentioned each process and each system construction drawing is all necessary
, certain step or units can be ignored according to the actual needs.Each step execution sequence be not it is fixed, can be according to need
It is adjusted.Apparatus structure described in the various embodiments described above can be physical structure, be also possible to logical construction, that is, have
A little units may be realized by same physical entity, be realized alternatively, some units may divide by multiple physical entities, alternatively, can be with
It is realized jointly by certain components in multiple autonomous devices.
In the above various embodiments, hardware cell or module mechanically or can be realized electrically.For example, one
Hardware cell, module or processor may include permanent dedicated circuit or logic (such as special processor, FPGA or
ASIC) corresponding operating is completed.Hardware cell or processor can also include programmable logic or circuit (such as general processor or
Other programmable processors), interim setting can be carried out by software to complete corresponding operating.Concrete implementation mode is (mechanical
Mode or dedicated permanent circuit or the circuit being temporarily arranged) it can be determined based on cost and temporal consideration.
The specific embodiment illustrated above in conjunction with attached drawing describes exemplary embodiment, it is not intended that may be implemented
Or fall into all embodiments of the protection scope of claims." exemplary " meaning of the term used in entire this specification
Taste " be used as example, example or illustration ", be not meant to than other embodiments " preferably " or " there is advantage ".For offer pair
The purpose of the understanding of described technology, specific embodiment include detail.However, it is possible in these no details
In the case of implement these technologies.In some instances, public in order to avoid the concept to described embodiment causes indigestion
The construction and device known is shown in block diagram form.
The foregoing description of present disclosure is provided so that any those of ordinary skill in this field can be realized or make
Use present disclosure.To those skilled in the art, the various modifications carried out to present disclosure are apparent
, also, can also answer generic principles defined herein in the case where not departing from the protection scope of present disclosure
For other modifications.Therefore, present disclosure is not limited to examples described herein and design, but disclosed herein with meeting
Principle and novel features widest scope it is consistent.
Claims (18)
1. a kind of method for determining Sales Volume of Commodity predicted value, comprising:
The article sales data in predetermined amount of time is obtained, the article sales data includes at least sales date and sales volume；
Based on acquired article sales data, training sample set is generated；
One or more machine learning models are trained using training sample set generated；And
Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training.
2. the method for claim 1, wherein using training sample set generated come to one or more machine learning
Model be trained be realized based on bagging algorithm, and
Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training includes:
Obtain the candidate commodity Method for Sales Forecast under each machine learning model in one or more of machine learning models
Value；And
The average value for calculating acquired each candidate commodity Method for Sales Forecast value, as Sales Volume of Commodity predicted value.
3. the method for claim 1, wherein using training sample set generated come to one or more machine learning
Model be trained be realized based on boosting algorithm, and
Sales Volume of Commodity predicted value is obtained based on one or more machine learning models after training includes:
Obtain the candidate commodity Method for Sales Forecast under each machine learning model in one or more of machine learning models
Value and corresponding prediction weighted value；And
Weight adduction is carried out to acquired each candidate commodity Method for Sales Forecast value, to obtain Sales Volume of Commodity predicted value.
4. the method for claim 1, wherein the machine learning model includes at least one of following: random gloomy
Woods model, extreme gradient decline model, quantile Random Forest model and supporting vector machine model.
5. generating training sample set includes: the method for claim 1, wherein based on acquired article sales data
Acquired article sales data is converted into model available feature data；And
Based on the model available feature data, training sample set is generated.
6. method as claimed in claim 5, wherein based on acquired article sales data, generate training sample set and also wrap
It includes:
Feature Dimension Reduction processing is carried out to the model available feature data, and
Based on the model available feature data, generating training sample set includes:
Based on by the Feature Dimension Reduction treated model available feature data, training sample set is generated.
7. method as claimed in claim 6, wherein the Feature Dimension Reduction, which is handled, includes:
The characteristic information unrelated with sales volume is removed from the model available feature data；
The characteristic information that the degree of correlation is more than predetermined threshold is removed from the model available feature data；And/or
Characteristic information elimination is carried out to the model available feature information using random forests algorithm.
8. the method as described in claim 1, further includes:
Data prediction is carried out to acquired article sales data.
9. the method as described in claim 1, further includes:
The model parameter of one or more of machine learning models is adjusted using training sample set generated.
10. method as claimed in claim 9, wherein the model parameter adjustment is using first heuristic algorithm and/or Bayes
Optimisation technique is realized.
11. the method for claim 1, wherein the article sales data further includes at least one of following:
Selling price；
Sales promotion information；
Macroeconomic data information；With
Weather information.
12. a kind of for determining the device of Sales Volume of Commodity predicted value, comprising:
Sales data acquiring unit, for obtaining the article sales data in predetermined amount of time, the article sales data is at least
Including sales date and sales volume；
Sample set generation unit, for generating training sample set based on acquired article sales data；
Training unit, for being trained using training sample set generated to one or more machine learning models；With
And
Predicting unit, for obtaining Sales Volume of Commodity predicted value based on one or more machine learning models after training.
13. device as claimed in claim 12, wherein the training unit is based on bagging algorithm or boosting algorithm
To train one or more of machine learning models.
14. device as claimed in claim 12, wherein the sample set generation unit includes:
Characteristic conversion module, for acquired article sales data to be converted to model available feature data；And
Sample set generation module generates training sample set for being based on the model available feature data.
15. device as claimed in claim 14, wherein the sample set generation unit further include:
Feature Dimension Reduction processing module is used to carry out Feature Dimension Reduction processing to the model available feature data, and
The sample set generation module is used for:
Based on by the Feature Dimension Reduction treated model available feature data, training sample set is generated.
16. device as claimed in claim 12, further includes:
Model parameter adjustment unit, for using training sample set generated come to one or more of machine learning models
Model parameter be adjusted.
17. a kind of calculating equipment, comprising:
One or more processors, and
The memory coupled with one or more of processors, the memory store instruction, when described instruction is by described one
When a or multiple processors execute, so that one or more of processors are executed as described in any in claims 1 to 11
Method.
18. a kind of non-transitory machinable medium, is stored with executable instruction, described instruction makes upon being performed
The machine executes the method as described in any in claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810466256.XA CN110503447A (en) | 2018-05-16 | 2018-05-16 | For determining the method and device of Sales Volume of Commodity predicted value |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810466256.XA CN110503447A (en) | 2018-05-16 | 2018-05-16 | For determining the method and device of Sales Volume of Commodity predicted value |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110503447A true CN110503447A (en) | 2019-11-26 |
Family
ID=68583597
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810466256.XA Pending CN110503447A (en) | 2018-05-16 | 2018-05-16 | For determining the method and device of Sales Volume of Commodity predicted value |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110503447A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105160A (en) * | 2019-12-20 | 2020-05-05 | 北京工商大学 | Steel quality prediction method based on tendency heterogeneous bagging algorithm |
| CN111242531A (en) * | 2020-01-03 | 2020-06-05 | 秒针信息技术有限公司 | Raw material management and control method and device and electronic equipment |
| CN111242532A (en) * | 2020-01-03 | 2020-06-05 | 秒针信息技术有限公司 | Purchasing management method and device and electronic equipment |
| CN112884520A (en) * | 2021-03-09 | 2021-06-01 | 拉扎斯网络科技（上海）有限公司 | Object processing method, device, electronic equipment, storage medium and program product |
| CN113674041A (en) * | 2020-05-15 | 2021-11-19 | 浙江大搜车软件技术有限公司 | Vehicle quotation method, device, system and computer readable storage medium |
| CN113743969A (en) * | 2020-05-29 | 2021-12-03 | 北京京东智能城市大数据研究院 | Method, device, equipment and storage medium for sales data prediction |
| CN113781120A (en) * | 2021-09-14 | 2021-12-10 | 北京京东振世信息技术有限公司 | Construction method of sales amount prediction model and sales amount prediction method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102106730A (en) * | 2011-03-16 | 2011-06-29 | 上海交通大学 | Method for processing electroencephalogram signal and detecting alertness based on fractal characteristics |
| CN106204086A (en) * | 2015-05-06 | 2016-12-07 | 阿里巴巴集团控股有限公司 | The method for early warning of Sales Volume of Commodity and device |
| CN106504030A (en) * | 2016-11-10 | 2017-03-15 | 重庆医药数据信息科技有限公司 | The support vector machine medicine Forecasting Methodology of the flower pollination optimized algorithm based on chaos sine-mapping |
| CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
| CN107180362A (en) * | 2017-05-03 | 2017-09-19 | 浙江工商大学 | Retail commodity sales forecasting method based on deep learning |
| CN107274231A (en) * | 2017-06-29 | 2017-10-20 | 北京京东尚科信息技术有限公司 | Data predication method and device |
-
2018
- 2018-05-16 CN CN201810466256.XA patent/CN110503447A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102106730A (en) * | 2011-03-16 | 2011-06-29 | 上海交通大学 | Method for processing electroencephalogram signal and detecting alertness based on fractal characteristics |
| CN106204086A (en) * | 2015-05-06 | 2016-12-07 | 阿里巴巴集团控股有限公司 | The method for early warning of Sales Volume of Commodity and device |
| CN106504030A (en) * | 2016-11-10 | 2017-03-15 | 重庆医药数据信息科技有限公司 | The support vector machine medicine Forecasting Methodology of the flower pollination optimized algorithm based on chaos sine-mapping |
| CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
| CN107180362A (en) * | 2017-05-03 | 2017-09-19 | 浙江工商大学 | Retail commodity sales forecasting method based on deep learning |
| CN107274231A (en) * | 2017-06-29 | 2017-10-20 | 北京京东尚科信息技术有限公司 | Data predication method and device |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105160A (en) * | 2019-12-20 | 2020-05-05 | 北京工商大学 | Steel quality prediction method based on tendency heterogeneous bagging algorithm |
| CN111242531A (en) * | 2020-01-03 | 2020-06-05 | 秒针信息技术有限公司 | Raw material management and control method and device and electronic equipment |
| CN111242532A (en) * | 2020-01-03 | 2020-06-05 | 秒针信息技术有限公司 | Purchasing management method and device and electronic equipment |
| CN113674041A (en) * | 2020-05-15 | 2021-11-19 | 浙江大搜车软件技术有限公司 | Vehicle quotation method, device, system and computer readable storage medium |
| CN113674041B (en) * | 2020-05-15 | 2023-11-21 | 浙江大搜车软件技术有限公司 | Vehicle quotation method, device, system and computer readable storage medium |
| CN113743969A (en) * | 2020-05-29 | 2021-12-03 | 北京京东智能城市大数据研究院 | Method, device, equipment and storage medium for sales data prediction |
| CN112884520A (en) * | 2021-03-09 | 2021-06-01 | 拉扎斯网络科技（上海）有限公司 | Object processing method, device, electronic equipment, storage medium and program product |
| CN113781120A (en) * | 2021-09-14 | 2021-12-10 | 北京京东振世信息技术有限公司 | Construction method of sales amount prediction model and sales amount prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110503447A (en) | For determining the method and device of Sales Volume of Commodity predicted value | |
| CN109325640B (en) | User value prediction method, device, storage medium and equipment | |
| CN109376267B (en) | Method and apparatus for generating a model | |
| EP3545471A1 (en) | Distributed clinical workflow training of deep learning neural networks | |
| CN105824806B (en) | A kind of quality evaluating method and device of public's account | |
| US10537801B2 (en) | System and method for decision making in strategic environments | |
| CN112070310A (en) | Loss user prediction method and device based on artificial intelligence and electronic equipment | |
| US20190392295A1 (en) | Information processing device, method, and program that use deep learning | |
| CN110362639A (en) | A kind of method for prewarning risk, device, electronic equipment calculating analysis based on demographic associations | |
| CN113379071B (en) | Noise label correction method based on federal learning | |
| Pourbafrani et al. | SIMPT: process improvement using interactive simulation of time-aware process trees | |
| CN111159241B (en) | Click conversion estimation method and device | |
| Niemann et al. | Computational Eurotransplant kidney allocation simulations demonstrate the feasibility and benefit of T-cell epitope matching | |
| Bottero et al. | Valuing the impact of social housing renovation programs: An application of the Social Return on Investment (SROI) | |
| Yates et al. | Evaluation of synthetic aerial imagery using unconditional generative adversarial networks | |
| Pourbafrani et al. | Interactive process improvement using simulation of enriched process trees | |
| CN112308623A (en) | High-quality client loss prediction method and device based on supervised learning and storage medium | |
| Stouky et al. | Improving software automation testing using jenkins, and machine learning under big data | |
| CN115829159A (en) | Social media fan newly-increased prediction method, device, equipment and storage medium | |
| CN116954627A (en) | Parameter prediction method, device, equipment, medium and program product for game product | |
| Rabbi et al. | An intelligent crowd-worker selection approach for reliable content labeling of food images | |
| Loyola et al. | Population dynamics in open source communities: an ecological approach applied to Github | |
| WO2015049802A1 (en) | Database generation device and database generation method | |
| Khamsan et al. | Handling highly imbalanced output class label: a case study on Fantasy Premier League (FPL) virtual player price changes prediction using machine learning | |
| JP2018097726A (en) | Determining apparatus, determining method, determining program, generating apparatus, generating method, and generating program |
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: 20191126 |
|
| RJ01 | Rejection of invention patent application after publication |