CN109064212A - Price forecasting of commodity method and device - Google Patents
Price forecasting of commodity method and device Download PDFInfo
- Publication number
- CN109064212A CN109064212A CN201810727271.5A CN201810727271A CN109064212A CN 109064212 A CN109064212 A CN 109064212A CN 201810727271 A CN201810727271 A CN 201810727271A CN 109064212 A CN109064212 A CN 109064212A
- Authority
- CN
- China
- Prior art keywords
- commodity
- parameter
- sample data
- price
- function 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
-
- 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
- G06Q30/0283—Price estimation or determination
Abstract
The present invention provides a kind of price forecasting of commodity method and device, is related to data analysis technique field, obtains the target sample data of price forecasting of commodity；Target sample data include: training sample data；Based on default input/output relation function model, the prior density function of each parameter in default input/output relation function model is determined；According to prior density function, training sample data and Bayes' theorem, price forecasting of commodity distribution function model is obtained；New input supplemental characteristic to be predicted is inputted into price forecasting of commodity distribution function model, calculates the output of price forecasting of commodity distribution function model as a result, as the new corresponding commodity projection price of input supplemental characteristic to be predicted.The present invention is based on the sample datas of the parameter relevant to price forecasting of commodity of acquisition, establish price forecasting of commodity distribution function model, it is predicted by output result of the distribution function model to new input data, improves the accuracy and confidence level of price forecasting of commodity result.
Description
Technical field
The present invention relates to data analysis technique fields, more particularly, to a kind of price forecasting of commodity method and device.
Background technique
The prediction technique of commodity price is the basis of Market Forecast Analysis Yu commodity production Authorize to X, is market prediction neck
A major issue in domain plays key effect at many aspects such as commodity production, sale.
The existing price forecasting of commodity method based on neural network algorithm, usually estimates interior weight parameter, biasing first
Parameter and outer weight parameter, and then anticipation function f (x) is obtained, to calculate the output valve of f (x) to new input supplemental characteristic x
Predicted value as commodity price.The shortcomings that this kind of algorithm, is: during solving parameter, gradient descent method is generallyd use, it should
The intrinsic defect of algorithm is readily available local extremum, rather than our global extremums for wanting, therefore, forecasting accuracy and can
Reliability is not high.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of price forecasting of commodity method and devices, based on acquisition with
The sample data of the relevant parameter of price forecasting of commodity establishes price forecasting of commodity distribution function model, passes through the distribution function
Model further predicts the output result of new input data, improves the accuracy of price forecasting of commodity result and credible
Degree helps businessman to automate the decision in many supply chain processes, more accurate requirement forecasting, can greatly optimizing management at
This, reduces timeliness of receiving, promotes the supply chain logistics efficiency of entire society.
In a first aspect, the embodiment of the invention provides a kind of price forecasting of commodity methods, comprising:
Obtain the target sample data of price forecasting of commodity；Target sample data include: the first parameter and the second parameter institute
Corresponding data；Using the first parameter as input parameter；Wherein the first parameter includes: item property parameter, commodity local environment
Parameter；Using the second parameter as output parameter；Wherein the second parameter includes: commodity price；Target sample data include: trained sample
Notebook data and verifying sample data；
Based on default input/output relation function model, each parameter in default input/output relation function model is determined
Prior density function；
According to prior density function, training sample data and Bayes' theorem, price forecasting of commodity distribution function mould is obtained
Type；
New input supplemental characteristic to be predicted is inputted into price forecasting of commodity distribution function model, calculates price forecasting of commodity
The output of distribution function model is as a result, as the new corresponding commodity projection price of input supplemental characteristic to be predicted.
With reference to first aspect, the embodiment of the invention provides the first possible embodiments of first aspect, wherein base
In default input/output relation function model, the prior distribution letter of each parameter in default input/output relation function model is determined
Number, specifically includes:
Default input/output relation function model are as follows:
T=f (x)+ε；Wherein,
Wherein, Ai Tx+biIt indicates to carry out a linear transformation, A to input feature valueiFor interior weight parameter, biFor biasing ginseng
Number, βiFor outer weight parameter, m indicates the number of hidden layer node, and G indicates that nonlinear activation primitive, ε indicate white Gaussian noise；
Wherein, it is 0 that the prior density function of ε, which is mean value, variance σ2Gauss of distribution function；
Interior weight parameter { Ai,j: i=1 ..., m；J=1 ..., p prior density function be set as:
Offset parameter { bi: i=1 ..., m } prior density function is set as:
p(bi)=N (bi|κ1,κ2), i=1 ..., m.；
The prior density function of outer weight parameter is set as:
P (β)=N (β | 0, γ-1I).
Hyper parameter { σ2, γ } prior density function be set as:
P (γ)=gamma (γ | α1,α2)
p(σ2)=Gamma (σ2|α3,α4)；
Wherein, μi,j,ηi,jκ1,κ2,α1,α2,α3,α4For constant.
With reference to first aspect, the embodiment of the invention provides second of possible embodiments of first aspect, wherein non-
Linear activation primitive includes: one of sigmoid function, radial basis function and hyperbolic tangent function.
With reference to first aspect, the embodiment of the invention provides the third possible embodiments of first aspect, wherein root
According to prior density function, training sample data and Bayes' theorem, price forecasting of commodity distribution function model is obtained, it is specific to wrap
It includes:
Training sample data are brought into prior density function, using Bayesian formula calculating parameter variable z=β, A, b,
γ,σ2Posterior distrbutionp, obtain following formula:
Wherein, Joint Distribution p (D, z)=p (t | X, β, σ2,A,b)p(β|γ)p(γ)p(A)p(b)；
The variation that p (z | D) is found by maximizing obvious lower bound ELBO approaches distribution q (z | θ), wherein ELBO is defined as: L
(θ)=Eqθ(z)[logp(z|D)]-Eqθ(z)[logq(z|θ)]；
Price forecasting of commodity distribution function model is calculated using the obtained distribution q (z | θ) that approaches:
P (t | D, z, x)=Eq(z|θ)[p (t | x, z)]；
Wherein, D is training sample data, and z is parametric variable, and x is new input supplemental characteristic.
With reference to first aspect, the embodiment of the invention provides the 4th kind of possible embodiments of first aspect, wherein will
New input supplemental characteristic to be predicted inputs price forecasting of commodity distribution function model, calculates price forecasting of commodity distribution function mould
The output of type is as a result, specifically include:
New input supplemental characteristic to be predicted is inputted into price forecasting of commodity distribution function model, calculates price forecasting of commodity
The expectation of distribution function model, it would be desirable to the output result as price forecasting of commodity distribution function model.
With reference to first aspect, the embodiment of the invention provides the 5th kind of possible embodiments of first aspect, wherein meter
The expectation for calculating price forecasting of commodity distribution function model, specifically includes:
Using Monte Carlo MCMC methodology, the value of approaching of the output valve of price forecasting of commodity distribution function model is calculated, it will
Approach expectation of the value as price forecasting of commodity distribution function model.
With reference to first aspect, the embodiment of the invention provides the 6th kind of possible embodiments of first aspect, wherein
Based on default input/output relation function model, the prior distribution of each parameter in default input/output relation function model is determined
Before function, further includes:
Target sample data are pre-processed, are specifically included:
For the case where there are missing datas in target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；
For the case where there are high dimensional datas in target sample data, feature extraction dimension-reduction treatment is carried out.
With reference to first aspect, the embodiment of the invention provides the 7th kind of possible embodiments of first aspect, wherein into
Row feature extraction dimension-reduction treatment, specifically includes:
Using Linear feature extraction algorithm, sparse dimension reduction is carried out to the data characteristics in target sample data, is specifically included:
Equipped with linear relationship: t=βTX+ ε, is optimized by following formula:
Wherein T=[t1,…,tN]T, X=[x1,…,xN]T；
Suitable regularization parameter λ is selected so that input data x is down to r (< d) dimension from d dimension；Wherein X is N number of sample data
Input data constitute N × d matrix；T is the matrix of N × 1 that output data is constituted；
Alternatively,
Using Gradient learning algorithm, the feature of initial data is screened by the norm size of gradient component, specifically
Include:
Equipped with non-linear relation: t=f (x)+ε is optimized by following formula:
Gradient function is obtained, r significant variable before selecting using the norm value size of the component of gradient function.
With reference to first aspect, the embodiment of the invention provides the 8th kind of possible embodiments of first aspect, wherein
According to prior density function, training sample data and Bayes' theorem, after obtaining price forecasting of commodity distribution function model, also
Include:
Verifying sample data is substituted into price forecasting of commodity distribution function model, to price forecasting of commodity distribution function model
Accuracy verified.
With reference to first aspect, the embodiment of the invention provides the 9th kind of possible embodiments of first aspect, wherein quotient
Product include: house；
Item property parameter includes: house house type, floor space, house geographical location；Commodity local environment parameter includes:
Premises perimeter air quality, premises perimeter establishment type；Commodity price includes: home price.
Second aspect, the embodiment of the present invention also provide a kind of price forecasting of commodity device, comprising:
Data acquisition module, for obtaining the target sample data of price forecasting of commodity；Target sample data include: first
Data corresponding to parameter and the second parameter；Using the first parameter as input parameter；Wherein the first parameter includes: item property ginseng
Number, commodity local environment parameter；Using the second parameter as output parameter；Wherein the second parameter includes: commodity price；Target sample
Data include: training sample data and verifying sample data；
Prior density function determining module, for determining that default input is defeated based on default input/output relation function model
Out in functional relation model each parameter prior density function；
Prediction distribution function establishes module, for obtaining according to prior density function, training sample data and Bayes' theorem
To price forecasting of commodity distribution function model；
Data prediction module, for new input supplemental characteristic to be predicted to be inputted price forecasting of commodity distribution function mould
Type calculates the output of price forecasting of commodity distribution function model as a result, as the new corresponding quotient of input supplemental characteristic to be predicted
Product forecast price.
In conjunction with second aspect, the embodiment of the invention provides the first possible embodiments of second aspect, wherein also
Include:
Data preprocessing module is specifically included for pre-processing to target sample data:
For the case where there are missing datas in target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；
For the case where there are high dimensional datas in target sample data, feature extraction dimension-reduction treatment is carried out.
The third aspect, the embodiment of the present invention also provide a kind of meter of non-volatile program code that can be performed with processor
Calculation machine readable medium, program code make processor execute method described in first aspect.
The embodiment of the present invention bring it is following the utility model has the advantages that
In price forecasting of commodity method provided in an embodiment of the present invention, the target sample of price forecasting of commodity is obtained first
Data；Wherein, target sample data include: data corresponding to the first parameter and the second parameter；Using the first parameter as input
Parameter；Wherein the first parameter includes: item property parameter, commodity local environment parameter；Using the second parameter as output parameter；Its
In the second parameter include: commodity price；Target sample data include: training sample data and verifying sample data；Then, it is based on
Default input/output relation function model determines the prior distribution letter of each parameter in default input/output relation function model
Number；According to prior density function, training sample data and Bayes' theorem, price forecasting of commodity distribution function model is obtained；It will
New input supplemental characteristic to be predicted inputs price forecasting of commodity distribution function model, calculates price forecasting of commodity distribution function mould
The output of type is as a result, as the new corresponding commodity projection price of input supplemental characteristic to be predicted.The embodiment of the present invention is based on adopting
The sample data of the parameter relevant to price forecasting of commodity of collection, establishes price forecasting of commodity distribution function model, passes through this point
Cloth function model further predicts the output result of new input data, improves the accuracy of price forecasting of commodity result
And confidence level, help businessman to automate the decision in many supply chain processes, more accurate requirement forecasting can greatly optimize
Operation cost reduces timeliness of receiving, promotes the supply chain logistics efficiency of entire society.
Other features and advantages of the present invention will illustrate in the following description, also, partly become from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention are in specification, claims
And specifically noted structure is achieved and obtained in attached drawing.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart for price forecasting of commodity method that the embodiment of the present invention one provides；
Fig. 2 is the flow chart for another price forecasting of commodity method that the embodiment of the present invention one provides；
Fig. 3 is the flow chart for another price forecasting of commodity method that the embodiment of the present invention one provides；
Fig. 4 is a kind of structural schematic diagram of price forecasting of commodity device provided by Embodiment 2 of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
The existing price forecasting of commodity method based on neural network algorithm generallys use during solving parameter
Gradient descent method, the intrinsic defect of the algorithm cause forecasting accuracy and confidence level not high.Based on this, the embodiment of the present invention is provided
A kind of price forecasting of commodity method and device, can based on the sample data of the parameter relevant to price forecasting of commodity of acquisition,
Price forecasting of commodity distribution function model is established, by the distribution function model further to the output result of new input data
It is predicted, improves the accuracy and confidence level of price forecasting of commodity result, businessman is helped to automate in many supply chain processes
Decision, more accurate requirement forecasting, can greatly optimizing management cost, reduction receives timeliness, promotes the confession of entire society
Answer chain logistic efficiency.
For convenient for understanding the present embodiment, first to a kind of price forecasting of commodity side disclosed in the embodiment of the present invention
Method describes in detail.
Embodiment one:
It is shown in Figure 1 the embodiment of the invention provides a kind of price forecasting of commodity method, this method comprises:
S11: obtaining the target sample data of price forecasting of commodity, and target sample data include: training sample data and test
Demonstrate,prove sample data.
Wherein, target sample data include: data corresponding to the first parameter and the second parameter；Using the first parameter as defeated
Enter parameter；Wherein the first parameter includes: item property parameter, commodity local environment parameter；Using the second parameter as output parameter；
Wherein the second parameter includes: commodity price.
It should be noted that the commodity in the embodiment of the present invention can be daily necessities, can be house or agricultural product etc.
Other commodity with economic significance.In addition, a part is used to training pattern, referred to as in the target sample data usually acquired
Training sample data, another part are used to verify the accuracy of model, referred to as verifying sample data.
In a kind of specific embodiment, above-mentioned commodity are house, and item property parameter includes: house house type, house
Area, house geographical location etc.；Commodity local environment parameter includes: premises perimeter air quality, premises perimeter establishment type etc.；
Commodity price includes: home price.
For example, the training sample data of the home price prediction obtained are as follows: D={ zi=(xi,ti):xi∈Rd,ti∈R,i
=1 ..., N }, in home price prediction, ziIndicate the related data in i-th of house.Wherein xiIt is the input data of d dimension, i.e.,
The vector that record or the d characteristic related with price of measurement are constituted, for example, in home price prediction, with valence
The relevant characteristic of lattice includes: cities and towns crime rate per capita, age of dwellings, nitrogen oxidation pollutant, low-income groups's ratio, cities and towns non-zero
Sell the ratio of commercial land, house is averaged room number etc..tiIt is 1 dimension output variable, i.e. price.
S12: based on default input/output relation function model, each ginseng in default input/output relation function model is determined
Several prior density functions.
In the present embodiment, input/output relation function model is preset are as follows:
T=f (x)+ε；Wherein,
Wherein, Ai Tx+biIt indicates to carry out a linear transformation, A to input feature valueiFor interior weight parameter, biFor biasing ginseng
Number, βiFor outer weight parameter, m indicates the number of hidden layer node, and G indicates that nonlinear activation primitive, ε indicate white Gaussian noise.
Wherein, it is 0 that the prior density function of ε, which is mean value, variance σ2Gauss of distribution function；
Interior weight parameter { Ai,j: i=1 ..., m；J=1 ..., p prior density function be set as:
Offset parameter { bi: i=1 ..., m } prior density function is set as:
p(bi)=N (bi|κ1,κ2), i=1 ..., m.；
The prior density function of outer weight parameter is set as:
P (β)=N (β | 0, γ-1I).
Hyper parameter { σ2, γ } prior density function be set as:
P (γ)=gamma (γ | α1,α2)
p(σ2)=Gamma (σ2|α3,α4)；
Wherein, μi,j,ηi,jκ1,κ2,α1,α2,α3,α4For constant.
In practical applications, above-mentioned nonlinear activation primitive may is that sigmoid function, radial basis function and hyperbolic
Any one of tangent function, in the present embodiment, G uses sigmoid function:Moreover, training sample data
It is inputted according to format D；To the parameter initialization in random distribution: αi=10-5, i=1 ..., 4；κ1=0, κ2=1；μi,j=0,
ηi,j=1.
Due in data acquisition, not can avoid the generation of noise, it can be assumed that the data t and true letter that obtain
A white Gaussian noise is differed between number f (x), i.e., the mean value of above-mentioned ε, stochastic variable ε are 0, variance σ2.Further, respectively
Determine the prior density function of each parameter in default input/output relation function model, such as above-mentioned interior weight parameter, biasing ginseng
The prior density function of several, outer weight parameter and hyper parameter, so as to subsequent modeling use.
S13: according to prior density function, training sample data and Bayes' theorem, price forecasting of commodity distribution letter is obtained
Exponential model.
Specifically, training sample data D is brought into prior density function, Bayesian formula calculating parameter variable z is utilized
={ β, A, b, γ, σ2Posterior distrbutionp, obtain following formula:
Wherein, Joint Distribution p (D, z)=p (t | X, β, σ2,A,b)p(β|γ)p(γ)p(A)p(b)；
The variation that p (z | D) is found by maximizing obvious lower bound ELBO approaches distribution q (z | θ), wherein ELBO is defined as: L
(θ)=Eqθ(z)[logp(z|D)]-Eqθ(z)[logq(z|θ)]；
Price forecasting of commodity distribution function model is calculated using the obtained distribution q (z | θ) that approaches:
P (t | D, z, x)=Eq(z|θ)[p (t | x, z)]；
Wherein, D is training sample data, and z is parametric variable, and x is new input supplemental characteristic.
S14: new input supplemental characteristic to be predicted is inputted into price forecasting of commodity distribution function model, calculates commodity price
The output of prediction distribution function model is as a result, as the new corresponding commodity projection price of input supplemental characteristic to be predicted.
Specifically, new input supplemental characteristic to be predicted is inputted price forecasting of commodity distribution function model, commodity are calculated
The expectation of price expectation distribution function model, it would be desirable to the output result as price forecasting of commodity distribution function model.
When being estimated using the expectation of prediction distribution output valve, since price forecasting of commodity distribution function model does not have
There is analytical expression, so value will be approached using Monte Carlo (MCMC) method calculating output valve, value will be approached as commodity
The expectation of price expectation distribution function model, that is, the output result of last price forecasting of commodity distribution function model.
MCMC methodology be for shaped like:Integral Problem can not directly calculate that (such as original function does not have resolution table
Up to formula) when a kind of method for using, the probability distribution p (y) of sampling is easy by construction one, computational problem is converted toUnder certain condition, it extracts n sample out, calculatesEstimated value as above-mentioned integral.
As a preferred implementation manner, in above-mentioned steps S12: based on default input/output relation function model, determining
In default input/output relation function model before the prior density function of each parameter, it can also include the following steps, referring to
Shown in Fig. 2:
S21: target sample data are pre-processed.The step specifically includes:
For the case where there are missing datas in target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；
For the case where there are high dimensional datas in target sample data, feature extraction dimension-reduction treatment is carried out.
Above-mentioned carry out feature extraction dimension-reduction treatment, specifically includes the following two kinds mode:
(1) Linear feature extraction algorithm is used, sparse dimension reduction is carried out to the data characteristics in target sample data, it is specific to wrap
It includes:
Equipped with linear relationship: t=βTX+ ε, is optimized by following formula:
Wherein T=[t1,…,tN]T, X=[x1,…,xN]T；
Suitable regularization parameter λ is selected so that input data x is down to r (< d) dimension from d dimension；Wherein X is N number of sample data
Input data constitute N × d matrix；T is the matrix of N × 1 that output data is constituted.
(2) Gradient learning algorithm is used, the feature of initial data is screened by the norm size of gradient component, is had
Body includes:
Equipped with non-linear relation: t=f (x)+ε is optimized by following formula:
Gradient function is obtained, r significant variable before selecting using the norm value size of the component of gradient function.
It is executing step S13: according to prior density function, training sample data and Bayes' theorem, obtaining commodity price
It is further comprising the steps of after prediction distribution function model, shown in Figure 3:
S31: verifying sample data is substituted into price forecasting of commodity distribution function model, to price forecasting of commodity distribution function
The accuracy of model is verified.
Sample data is verified as the input format of above-mentioned training sample data, by verifying sample data, detects quotient
The accuracy of product price expectation distributed model is updated so that the parameter in model is modified and is adjusted by continuous data
And accumulation, the accuracy of above-mentioned model can be made to become higher and higher, so that its confidence level be made to greatly increase.
Method provided by the embodiment of the present invention, the sample number of the factor relevant to price forecasting of commodity based on acquisition
According to, the modeling of price forecasting of commodity model is carried out, it is further defeated to new input data progress by price forecasting of commodity model
The prediction of result out improves the accuracy and confidence level of price forecasting of commodity value, and businessman is helped to automate many supply chain processes
In decision, more accurate requirement forecasting, can greatly optimizing management cost, reduction receives timeliness, promotes entire society
Supply chain logistics efficiency.
Price forecasting of commodity distribution function model provided in the embodiment of the present invention can portray simple function and close
System can more portray the pass between complex data (herein refer to: fluctuating the isomeric data of violent or distinct device acquisition) well
System.Meanwhile provide be price Distribution estimation, rather than simple price evaluation, such benefit are when data contain
When having noise and outlier, prediction result has better robustness, the i.e. minor alteration of sample data, causes algorithm resulting
Prediction result changes also very small.
It should be noted that method provided by the present embodiment is gone back in addition to that can predict the price of extensive stock
It can be applied to other fault pre-alarming fields, when the input parameter in the present embodiment, output parameter are various voltages, current value
When, equipment fault early-warning etc. can be carried out by the output result of prediction distribution function model.
Embodiment two:
The embodiment of the present invention also provides a kind of price forecasting of commodity device, shown in Figure 4, which includes: that data obtain
Modulus block 41, prior density function determining module 42, prediction distribution function establish 43 sum number of module it is predicted that module 44.
Wherein, data acquisition module 41, for obtaining the target sample data of price forecasting of commodity；Target sample data packet
It includes: data corresponding to the first parameter and the second parameter；Using the first parameter as input parameter；Wherein the first parameter includes: quotient
Product property parameters, commodity local environment parameter；Using the second parameter as output parameter；Wherein the second parameter includes: commodity price；
Target sample data include: training sample data and verifying sample data；Prior density function determining module 42, for based on pre-
If input/output relation function model, the prior density function of each parameter in default input/output relation function model is determined；
Prediction distribution function establishes module 43, for obtaining commodity according to prior density function, training sample data and Bayes' theorem
Price expectation distribution function model；Data prediction module 44, for new input supplemental characteristic to be predicted to be inputted commodity price
Prediction distribution function model calculates the output of price forecasting of commodity distribution function model as a result, joining as new input to be predicted
The corresponding commodity projection price of number data.
In addition, in above-mentioned apparatus further include: data preprocessing module 45, for being pre-processed to target sample data,
It specifically includes:
For the case where there are missing datas in target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；For the case where there are high dimensional datas in target sample data, feature extraction dimension-reduction treatment is carried out.
In price forecasting of commodity device provided by the embodiment of the present invention, modules and aforementioned price forecasting of commodity method
Therefore above-mentioned function equally may be implemented in technical characteristic having the same.The specific work process ginseng of modules in the present apparatus
See above method embodiment, details are not described herein.
The computer program product of price forecasting of commodity method provided by the embodiment of the present invention, including store processor
The computer readable storage medium of executable non-volatile program code, the instruction that said program code includes can be used for executing
Previous methods method as described in the examples, specific implementation can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the device of foregoing description
And the specific work process of electronic equipment, it can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
The flow chart and block diagram in the drawings show multiple embodiment method and computer program products according to the present invention
Architecture, function and operation in the cards.In this regard, each box in flowchart or block diagram can represent one
A part of module, section or code, a part of the module, section or code include it is one or more for realizing
The executable instruction of defined logic function.It should also be noted that in some implementations as replacements, function marked in the box
It can also can occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually be substantially parallel
Ground executes, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram
And/or the combination of each box in flow chart and the box in block diagram and or flow chart, it can the function as defined in executing
Can or the dedicated hardware based system of movement realize, or can come using a combination of dedicated hardware and computer instructions real
It is existing.
In the description of the present invention, it should be noted that term " center ", "upper", "lower", "left", "right", "vertical",
The orientation or positional relationship of the instructions such as "horizontal", "inner", "outside" be based on the orientation or positional relationship shown in the drawings, merely to
Convenient for description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation,
It is constructed and operated in a specific orientation, therefore is not considered as limiting the invention.In addition, term " first ", " second ",
" third " is used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance.
In several embodiments provided herein, it should be understood that disclosed systems, devices and methods, it can be with
It realizes by another way.The apparatus embodiments described above are merely exemplary, for example, the division of the unit,
Only a kind of logical function partition, there may be another division manner in actual implementation, in another example, multiple units or components can
To combine or be desirably integrated into another system, or some features can be ignored or not executed.Another point, it is shown or beg for
The mutual coupling, direct-coupling or communication connection of opinion can be through some communication interfaces, device or unit it is indirect
Coupling or communication connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in the executable non-volatile computer-readable storage medium of a processor.Based on this understanding, of the invention
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the present invention
State all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can be with
Store the medium of program code.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (13)
1. a kind of price forecasting of commodity method characterized by comprising
Obtain the target sample data of price forecasting of commodity；The target sample data include: the first parameter and the second parameter institute
Corresponding data；Using first parameter as input parameter；Wherein first parameter includes: item property parameter, commodity
Local environment parameter；Using second parameter as output parameter；Wherein second parameter includes: commodity price；The mesh
Standard specimen notebook data includes: training sample data and verifying sample data；
Based on default input/output relation function model, each parameter in the default input/output relation function model is determined
Prior density function；
According to the prior density function, the training sample data and Bayes' theorem, price forecasting of commodity distribution letter is obtained
Exponential model；
New input supplemental characteristic to be predicted is inputted into the price forecasting of commodity distribution function model, calculates the commodity price
The output of prediction distribution function model is as a result, as the new corresponding commodity projection price of input supplemental characteristic to be predicted.
2. the method according to claim 1, wherein described be based on default input/output relation function model, really
The prior density function of each parameter in the fixed default input/output relation function model, specifically includes:
The default input/output relation function model are as follows:
T=f (x)+ε；Wherein,
Wherein, Ai Tx+biIt indicates to carry out a linear transformation, A to input feature valueiFor interior weight parameter, biFor offset parameter, βi
For outer weight parameter, m indicates the number of hidden layer node, and G indicates that nonlinear activation primitive, ε indicate white Gaussian noise；
Wherein, it is 0 that the prior density function of ε, which is mean value, variance σ2Gauss of distribution function；
Interior weight parameter { Ai,j: i=1 ..., m；J=1 ..., p prior density function be set as:
Offset parameter { bi: i=1 ..., m } prior density function is set as:
p(bi)=N (bi|κ1,κ2), i=1 ..., m.；
The prior density function of outer weight parameter is set as:
P (β)=N (β | 0, γ-1I).
Hyper parameter { σ2, γ } prior density function be set as:
P (γ)=gamma (γ | α1,α2)
p(σ2)=Gamma (σ2|α3,α4)；
Wherein, μi,j,ηi,jκ1,κ2,α1,α2,α3,α4For constant.
3. according to the method described in claim 2, it is characterized in that, the nonlinear activation primitive includes: sigmoid letter
One of number, radial basis function and hyperbolic tangent function.
4. the method according to claim 1, wherein described according to the prior density function, the trained sample
Notebook data and Bayes' theorem obtain price forecasting of commodity distribution function model, specifically include:
The training sample data are brought into the prior density function, using Bayesian formula calculating parameter variable z=β,
A,b,γ,σ2Posterior distrbutionp, obtain following formula:
Wherein, Joint Distribution p (D, z)=p (t | X, β, σ2,A,b)p(β|γ)p(γ)p(A)p(b)；
The variation that p (z | D) is found by maximizing obvious lower bound ELBO approaches distribution q (z | θ), wherein ELBO is defined as: L (θ)
=Eqθ(z)[logp(z|D)]-Eqθ(z)[logq(z|θ)]；
The price forecasting of commodity distribution function model is calculated using the obtained distribution q (z | θ) that approaches:
P (t | D, z, x)=Eq(z|θ)[p (t | x, z)]；
Wherein, D is the training sample data, and z is the parametric variable, and x is new input supplemental characteristic.
5. according to the method described in claim 4, it is characterized in that, described will be described in new input supplemental characteristic to be predicted input
Price forecasting of commodity distribution function model calculates the output of the price forecasting of commodity distribution function model as a result, specifically including:
New input supplemental characteristic to be predicted is inputted into the price forecasting of commodity distribution function model, calculates the commodity price
The expectation of prediction distribution function model, the output knot by the expectation as the price forecasting of commodity distribution function model
Fruit.
6. according to the method described in claim 5, it is characterized in that, described calculate the price forecasting of commodity distribution function model
Expectation, specifically include:
Using Monte Carlo MCMC methodology, the value of approaching of the output valve of the price forecasting of commodity distribution function model is calculated, it will
The expectation that value is approached as the price forecasting of commodity distribution function model.
7. the method according to claim 1, wherein described based on default input/output relation function model,
It determines in the default input/output relation function model before the prior density function of each parameter, further includes:
The target sample data are pre-processed, are specifically included:
For the case where there are missing datas in the target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；
For the case where there are high dimensional datas in the target sample data, feature extraction dimension-reduction treatment is carried out.
8. the method according to the description of claim 7 is characterized in that the progress feature extraction dimension-reduction treatment, specifically includes:
Using Linear feature extraction algorithm, sparse dimension reduction is carried out to the data characteristics in the target sample data, is specifically included:
Equipped with linear relationship: t=βTX+ ε, is optimized by following formula:
Wherein T=[t1,…,tN]T, X=[x1,…,xN]T；
Suitable regularization parameter λ is selected so that input data x is down to r (< d) dimension from d dimension；Wherein X is the defeated of N number of sample data
Enter N × d matrix of data composition；T is the matrix of N × 1 that output data is constituted；
Alternatively,
Using Gradient learning algorithm, the feature of initial data is screened by the norm size of gradient component, is specifically included:
Equipped with non-linear relation: t=f (x)+ε is optimized by following formula:
Gradient function is obtained, r significant variable before selecting using the norm value size of the component of gradient function.
9. the method according to claim 1, wherein described according to the prior density function, the training
Sample data and Bayes' theorem, after obtaining price forecasting of commodity distribution function model, further includes:
The verifying sample data is substituted into the price forecasting of commodity distribution function model, the price forecasting of commodity is distributed
The accuracy of function model is verified.
10. the method according to claim 1, wherein the commodity include: house；
The item property parameter includes: house house type, floor space, house geographical location；The commodity local environment parameter
It include: premises perimeter air quality, premises perimeter establishment type；The commodity price includes: home price.
11. a kind of price forecasting of commodity device characterized by comprising
Data acquisition module, for obtaining the target sample data of price forecasting of commodity；The target sample data include: first
Data corresponding to parameter and the second parameter；Using first parameter as input parameter；Wherein first parameter includes: quotient
Product property parameters, commodity local environment parameter；Using second parameter as output parameter；Wherein second parameter includes:
Commodity price；The target sample data include: training sample data and verifying sample data；
Prior density function determining module, for determining that the default input is defeated based on default input/output relation function model
Out in functional relation model each parameter prior density function；
Prediction distribution function establishes module, for fixed according to the prior density function, the training sample data and Bayes
Reason, obtains price forecasting of commodity distribution function model；
Data prediction module, for new input supplemental characteristic to be predicted to be inputted the price forecasting of commodity distribution function mould
Type calculates the output of the price forecasting of commodity distribution function model as a result, as the new input supplemental characteristic to be predicted
Corresponding commodity projection price.
12. device according to claim 11, which is characterized in that further include:
Data preprocessing module is specifically included for pre-processing to the target sample data:
For the case where there are missing datas in the target sample data, processing is carried out using mean value Substitution Rules or to scarce
It loses data and carries out delete processing；
For the case where there are high dimensional datas in the target sample data, feature extraction dimension-reduction treatment is carried out.
13. a kind of computer-readable medium for the non-volatile program code that can be performed with processor, which is characterized in that described
Program code makes the processor execute the described in any item methods of claims 1 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810727271.5A CN109064212A (en) | 2018-07-04 | 2018-07-04 | Price forecasting of commodity method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810727271.5A CN109064212A (en) | 2018-07-04 | 2018-07-04 | Price forecasting of commodity method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109064212A true CN109064212A (en) | 2018-12-21 |
Family
ID=64819173
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810727271.5A Pending CN109064212A (en) | 2018-07-04 | 2018-07-04 | Price forecasting of commodity method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109064212A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109509039A (en) * | 2018-12-29 | 2019-03-22 | 携程计算机技术（上海）有限公司 | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model |
| CN109886844A (en) * | 2019-03-21 | 2019-06-14 | 中国电建集团昆明勘测设计研究院有限公司 | House registration data based on Bayesian network model is associated with building table method |
| CN112330368A (en) * | 2020-11-16 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Data processing method, system, storage medium and terminal equipment |
| CN113554449A (en) * | 2020-04-23 | 2021-10-26 | 阿里巴巴集团控股有限公司 | Commodity variable prediction method, commodity variable prediction device, and computer-readable medium |
| CN115936757A (en) * | 2022-12-27 | 2023-04-07 | 广东开放大学（广东理工职业学院） | Method and device for predicting psychological price distribution of bidders |
| CN116757730A (en) * | 2023-07-03 | 2023-09-15 | 中科智宏(北京)科技有限公司 | Method, device and storage medium for predicting selling price of electronic commerce commodity |
-
2018
- 2018-07-04 CN CN201810727271.5A patent/CN109064212A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109509039A (en) * | 2018-12-29 | 2019-03-22 | 携程计算机技术（上海）有限公司 | Method for building up and system, the Method of Commodity Recommendation and system of price expectation model |
| CN109886844A (en) * | 2019-03-21 | 2019-06-14 | 中国电建集团昆明勘测设计研究院有限公司 | House registration data based on Bayesian network model is associated with building table method |
| CN109886844B (en) * | 2019-03-21 | 2022-08-12 | 中国电建集团昆明勘测设计研究院有限公司 | House registration data association building chart method based on Bayesian network model |
| CN113554449A (en) * | 2020-04-23 | 2021-10-26 | 阿里巴巴集团控股有限公司 | Commodity variable prediction method, commodity variable prediction device, and computer-readable medium |
| CN112330368A (en) * | 2020-11-16 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Data processing method, system, storage medium and terminal equipment |
| CN112330368B (en) * | 2020-11-16 | 2024-04-09 | 腾讯科技（深圳）有限公司 | Data processing method, system, storage medium and terminal equipment |
| CN115936757A (en) * | 2022-12-27 | 2023-04-07 | 广东开放大学（广东理工职业学院） | Method and device for predicting psychological price distribution of bidders |
| CN115936757B (en) * | 2022-12-27 | 2023-09-22 | 广东开放大学（广东理工职业学院） | Method and device for predicting psychological price distribution of bidders |
| CN116757730A (en) * | 2023-07-03 | 2023-09-15 | 中科智宏(北京)科技有限公司 | Method, device and storage medium for predicting selling price of electronic commerce commodity |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109064212A (en) | Price forecasting of commodity method and device | |
| Sajjad et al. | A novel CNN-GRU-based hybrid approach for short-term residential load forecasting | |
| Boto‐Giralda et al. | Wavelet‐based denoising for traffic volume time series forecasting with self‐organizing neural networks | |
| CN107636693B (en) | Relevance score assignment for artificial neural networks | |
| Desmarais et al. | Micro‐level interpretation of exponential random graph models with application to estuary networks | |
| Santé et al. | Cellular automata models for the simulation of real-world urban processes: A review and analysis | |
| Huang et al. | Spatiotemporal analysis of rural–urban land conversion | |
| Liu et al. | Simulating land-use dynamics under planning policies by integrating artificial immune systems with cellular automata | |
| Matijaš et al. | Load forecasting using a multivariate meta-learning system | |
| Zhang et al. | Time series prediction using Lyapunov exponents in embedding phase space | |
| Kauko | On current neural network applications involving spatial modelling of property prices | |
| Blecic et al. | How much past to see the future: a computational study in calibrating urban cellular automata | |
| Seya et al. | A comparison of residential apartment rent price predictions using a large data set: Kriging versus deep neural network | |
| Giupponi et al. | Spatial assessment of water use efficiency (SDG indicator 6.4. 1) for regional policy support | |
| CN109271465A (en) | A kind of Hydrological Data Analysis and methods of exhibiting based on cloud computing | |
| CN109993753A (en) | The dividing method and device of urban function region in remote sensing image | |
| Wen et al. | Visibility graph for time series prediction and image classification: a review | |
| CN109615414A (en) | House property predictor method, device and storage medium | |
| Ahmadi et al. | Robust methods for identifying optimal reservoir operation strategies using deterministic and stochastic formulations | |
| Wang et al. | Spatial–temporal multi-feature fusion network for long short-term traffic prediction | |
| CN114077912A (en) | Data prediction method and data prediction device | |
| Gong et al. | ART‐P‐MAP neural networks modeling of land‐use change: accounting for spatial heterogeneity and uncertainty | |
| Qian et al. | Short-term demand forecasting for on-demand mobility service | |
| Zhang et al. | Efficient history matching with dimensionality reduction methods for reservoir simulations | |
| Liang et al. | Sponet: solve spatial optimization problem using deep reinforcement learning for urban spatial decision analysis |
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: 20181221 |
|
| RJ01 | Rejection of invention patent application after publication |