CN109034899A - Estimate the method, apparatus and computer readable storage medium of article residual value - Google Patents
Estimate the method, apparatus and computer readable storage medium of article residual value Download PDFInfo
- Publication number
- CN109034899A CN109034899A CN201810831981.2A CN201810831981A CN109034899A CN 109034899 A CN109034899 A CN 109034899A CN 201810831981 A CN201810831981 A CN 201810831981A CN 109034899 A CN109034899 A CN 109034899A
- Authority
- CN
- China
- Prior art keywords
- returned item
- return
- goods
- residual value
- sample
- 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
- G06Q30/0283—Price estimation or determination
-
- 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
- G06Q30/01—Customer relationship services
- G06Q30/015—Providing customer assistance, e.g. assisting a customer within a business location or via helpdesk
- G06Q30/016—After-sales
Landscapes
- Business, Economics & Management (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Marketing (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Present disclose provides a kind of method, apparatus and computer readable storage medium for estimating article residual value, are related to field of artificial intelligence.The method of estimation article residual value therein includes: the return of goods user data using returned item, generates the user characteristics of the returned item；Using the product data of the returned item, the article characteristics of the returned item are generated；The user characteristics and the article characteristics are inputted into deep learning neural network, estimation obtains the residual value of the returned item.The disclosure uses artificial intelligence depth learning technology when estimating article residual value, and considers user's factor and article factor, can accurately and efficiently estimate the residual value of article.
Description
Technical field
This disclosure relates to field of artificial intelligence, in particular to a kind of method, apparatus for estimating article residual value and
Computer readable storage medium.
Background technique
Internet changes the consumption pattern of modern people, in such a way that electric business platform carries out online shopping and traditional
There is very big difference in shopping way, this also proposed very high requirement to the efficiency and process of the after-sale service of electric business platform.
In electric business platform shopping process, commodity after-sale service is to influence user's shopping experience, promote likability and save trade company's spending
Key link.The promotion of after-sale service quality can not only enhance the retention ratio of user, more proposing to the service image of enterprise
Rising has huge facilitation.Therefore, when occurring the case where commodity face goods return and replacement, how to estimate the residual valence of return of goods commodity
Value is most important problem to handle after-sale service list according to the residual value of return of goods commodity.Therefore, return of goods commodity
Residual value is to handle the important indicator for needing to pay close attention in Return Application.
Currently, the method for estimation return of goods commodity residual value depends on manual evaluation, it can only residual valence to commodity
Value carries out rough estimate.This method can not accurately estimate the residual value of return of goods commodity, and electric business sample platform is caused to make discomfort
When after-sale service processing.Meanwhile contact staff faces the return of goods commodity of magnanimity, manual evaluation process takes a long time.
Summary of the invention
How quickly the technical problem that the disclosure solves is, accurately and efficiently to identify target in target group
Quantity.
According to the one aspect of the embodiment of the present disclosure, a kind of method for estimating article residual value is provided, comprising: utilize
The return of goods user data of returned item, generates the user characteristics of returned item；Using the product data of returned item, generates and return goods
The article characteristics of article；User characteristics and article characteristics are inputted into deep learning neural network, estimation obtains returned item
Residual value.
In some embodiments, using the return of goods user data of returned item, the user characteristics for generating returned item include
At least one of following steps: using the credit data of the return of goods user of returned item, the user credit of returned item is generated
Feature；Using the area data of the return of goods user of returned item, user's regionalism of returned item is generated；Utilize returned item
The return of goods user description return of goods opinion data, generate the consumers' opinions feature of returned item.
In some embodiments, the return of goods opinion data described using the return of goods user of returned item generates returned item
Consumers' opinions feature include: to the return of goods user description return of goods text carry out natural language processing, alternatively, to return of goods user clap
The return of goods photo taken the photograph carries out depth characteristic extraction, and the consumers' opinions for obtaining returned item understands feature.
In some embodiments, using the product data of returned item, it includes following for generating the article characteristics of returned item
At least one of step: using the category data of returned item, the category feature of returned item is generated；Utilize returned item
Specification data generates the specification feature of returned item；Using the time critical data of returned item, the time limit for generating returned item is special
Sign；Using the residual value statistical data of the same category article of returned item, the statistical nature of returned item is generated.
In some embodiments, deep learning neural network includes: linear learning sub-network, is configured as study input
Linear relationship between feature vector and the residual value of returned item, output indicate the feature vector of linear relationship；It is non-linear
Learn sub-network, the non-linear relation being configured as between the feature vector of study input and the residual value of returned item is defeated
The feature vector of non-linear relation is indicated out；Full articulamentum is configured as to indicate the feature of linear relationship by activation primitive
Vector and the maps feature vectors for indicating non-linear relation are the residual value of returned item.
In some embodiments, deep learning neural network further include: input layer, the user for being configured as merging input are special
Sign and article characteristics, obtain the merging feature of returned item；It is embedded in learning layer, is configured with the hidden of default fixed quantity
Containing the factor, it can will merge Feature Mapping to implicit factor space and obtain the feature vector of regular length.
In some embodiments, this method further include: using the user data of sample returned item, generate sample return of goods object
The user characteristics of product；Using the product data of sample returned item, the article characteristics of sample returned item are generated；Mark sample moves back
The residual value of cargo product；It is moved back using the user characteristics of sample returned item, the article characteristics of sample returned item and sample
The residual value of cargo product is trained deep learning neural network, enables deep learning neural network according to return of goods object
The user characteristics of product and the article characteristics of returned item, estimation obtain the residual value of returned item.
In some embodiments, if the residual value of mark sample returned item includes: that sample returned item needs to return
Genuine, the then residual value by the difference of the cost price of sample returned item and return of goods logistics cost, as sample returned item；
If sample returned item needs to return supplier, by the difference of the return price of sample returned item and return of goods logistics cost,
Residual value as sample returned item；If sample returned item needs to renovate sale, and sample returned item is corresponding
The difference for losing making-up price Yu return of goods logistics cost, the residual value as sample returned item；If sample returned item needs
Depreciation is wanted to sell, then the residual value by the difference of depreciation selling price and return of goods logistics cost, as sample returned item.
According to the other side of the embodiment of the present disclosure, a kind of device for estimating article residual value is provided, comprising: use
Family feature generation module is configured as the return of goods user data using returned item, generates the user characteristics of returned item；Article
Feature generation module is configured as the product data using returned item, generates the article characteristics of returned item；Residual value is estimated
Module is counted, is configured as user characteristics and article characteristics input deep learning neural network, estimation obtaining returned item
Residual value.
In some embodiments, user characteristics generation module is configured as executing at least one of following steps: utilizing
The credit data of the return of goods user of returned item generates the user credit feature of returned item；It is used using the return of goods of returned item
The area data at family generates user's regionalism of returned item；The return of goods opinion described using the return of goods user of returned item
Data generate the consumers' opinions feature of returned item.
In some embodiments, user characteristics generation module is configured as: being carried out to the return of goods text of return of goods user description
Natural language processing obtains the user of returned item alternatively, carrying out depth characteristic extraction to the return of goods photo of return of goods user shooting
Opinion understands feature.
In some embodiments, article characteristics generation module is configured as executing at least one of following steps: utilizing
The category data of returned item generate the category feature of returned item；Using the specification data of returned item, returned item is generated
Specification feature；Using the time critical data of returned item, the timing characteristics of returned item are generated；Utilize the same category of returned item
The residual value statistical data of article, generates the statistical nature of returned item.
In some embodiments, deep learning neural network includes: linear learning sub-network, is configured as study input
Linear relationship between feature vector and the residual value of returned item, output indicate the feature vector of linear relationship；It is non-linear
Learn sub-network, the non-linear relation being configured as between the feature vector of study input and the residual value of returned item is defeated
The feature vector of non-linear relation is indicated out；Full articulamentum is configured as to indicate the feature of linear relationship by activation primitive
Vector and the maps feature vectors for indicating non-linear relation are the residual value of returned item.
In some embodiments, deep learning neural network further include: input layer, the user for being configured as merging input are special
Sign and article characteristics, obtain the merging feature of returned item；It is embedded in learning layer, is configured with the hidden of default fixed quantity
Containing the factor, it can will merge Feature Mapping to implicit factor space and obtain the feature vector of regular length.
In some embodiments, which further includes neural metwork training module, is configured as: utilizing sample returned item
User data, generate sample returned item user characteristics；Using the product data of sample returned item, generates sample and return goods
The article characteristics of article；Mark the residual value of sample returned item；It is returned goods using the user characteristics of sample returned item, sample
The article characteristics of article and the residual value of sample returned item are trained deep learning neural network, so that depth
Practising neural network can estimate to obtain returned item according to the user characteristics of returned item and the article characteristics of returned item
Residual value.
In some embodiments, neural metwork training module is configured as: if sample returned item needs to return genuine,
Residual value by the difference of the cost price of sample returned item and return of goods logistics cost, as sample returned item；If sample
Returned item needs to return supplier, then by the difference of the return price of sample returned item and return of goods logistics cost, as sample
The residual value of this returned item；If sample returned item needs to renovate sale, core is lost by sample returned item is corresponding
The difference of price lattice and return of goods logistics cost, the residual value as sample returned item；If sample returned item needs depreciation
Sale, the then residual value by the difference of depreciation selling price and return of goods logistics cost, as sample returned item.
According to the another aspect of the embodiment of the present disclosure, the device of another estimation article residual value is provided, comprising:
Memory；And it is coupled to the processor of memory, processor is configured as based on instruction stored in memory, before execution
The method for the estimation article residual value stated.
According to another aspect of the embodiment of the present disclosure, a kind of computer readable storage medium is provided, wherein computer
Readable storage medium storing program for executing is stored with computer instruction, and the side of estimation article residual value above-mentioned is realized in instruction when being executed by processor
Method.
The disclosure uses artificial intelligence deep learning technology when estimating article residual value, and considers user's factor
And article factor, it can accurately and efficiently estimate the residual value of article.
By the detailed description referring to the drawings to the exemplary embodiment of the disclosure, the other feature of the disclosure and its
Advantage will become apparent.
Detailed description of the invention
In order to illustrate more clearly of the embodiment of the present disclosure or technical solution in the prior art, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Disclosed some embodiments without any creative labor, may be used also for those of ordinary skill in the art
To obtain other drawings based on these drawings.
Fig. 1 shows the flow diagram being trained to deep learning neural network.
Fig. 2 shows the structural schematic diagrams for the deep learning neural network that the disclosure uses.
Fig. 3 shows the flow diagram of one embodiment of the method for disclosure estimation returned item residual value.
Fig. 4 shows the structural schematic diagram of the device of the estimation article residual value of an embodiment of the present disclosure.
Fig. 5 shows the structural schematic diagram of the device of the estimation article residual value of the disclosure another embodiment.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present disclosure, the technical solution in the embodiment of the present disclosure is carried out clear, complete
Site preparation description, it is clear that described embodiment is only disclosure a part of the embodiment, instead of all the embodiments.Below
Description only actually at least one exemplary embodiment be it is illustrative, never as to the disclosure and its application or making
Any restrictions.Based on the embodiment in the disclosure, those of ordinary skill in the art are not making creative work premise
Under all other embodiment obtained, belong to the disclosure protection range.
Inventor analyzes the correlation technique of the residual value of estimation return of goods commodity.Inventor is the study found that visitor
Take personnel be connected to Return Application phone after sale or generally taking two kinds of estimation methods after Return Application list after sale: one is masters
The estimation technique is seen, judges whether commodity are lost according to the essential information of the commodity in Return Application list first, then according to after sale
Return Application phone and return of goods user, which link up, to be verified commodity and loses situation, finally according to history losing situation and go through with commodity
The residual value of the history empirical analysis return of goods commodity；Another kind is mean value Evaluation Method, it is first determined is determined according to Return Application list
The category of return of goods commodity, then using businessman lose determine human assessment with category return of goods commodity residual mean value as the return of goods
The residual value of commodity.
Former the relevant technologies need staff to carry out subjective judgement, will lead to the larger situation of assessment errors.It is different
When staff estimates identical returned item, estimated result is likely to occur relatively large deviation.In terms of assessment cycle, exists and artificially estimate
The problem of counting inefficiency, single-item analysis speed can exceed that 20 minutes/part.In the case where high concurrent big flow user rushes to purchase request,
It is difficult to meet the processing speed requirement of customer service list.Latter the relevant technologies cannot then be estimated for the feature of different Return Application lists
The residual value for counting return of goods commodity, it is larger to also result in assessment errors.The complicated inspection of return of goods commodity or tune are needed encountering
In the case where integral quotient product residual value standard, both the relevant technologies all have that executory cost is high, are difficult to flexibly to be applicable in.
Meanwhile the residual value of return of goods commodity often relates to various influence factors, and there may be intersect to close for different affecting factors
System often can not accurately be estimated by the method that staff carries out manual analysis.
Deep learning is a kind of important technology in machine learning, can be moved back for difference by deep learning neural network
The case where goods request slip and feature preferably analyze the relationship between multi-data source, carry out more accurately and efficiently estimation and return goods
The residual value of commodity.For ease of understanding, the disclosure carries out tool example description to returned item by taking the return of goods commodity of electric business as an example.
Fig. 1 is combined to describe the process being trained to deep learning neural network first.
Fig. 1 shows the flow diagram being trained to deep learning neural network.As shown in Figure 1, in the present embodiment
Training process include step S102~step S108.
In step s 102, using the user data of sample returned item, the user characteristics of sample returned item are generated.
The return of goods user identifier UID of return of goods commodity can be determined according to Return Application list.Pass through return of goods user identifier UID energy
The credit data (such as credit grade) of return of goods user, the area data (province where such as) of return of goods user are enough obtained, is passed through
Return Application list can also determine the return of goods user description return of goods opinion data (such as return of goods text, return of goods voice, return goods shine
Piece).
Using the credit data of the return of goods user of returned item, the user credit feature of returned item can be generated.It is moving back
In goods service, it is understood that there may be some users hesitate after often there is lower to commodity satisfaction or frequent purchase and the feelings returned goods
Condition, although in general user applies returning goods in this case, commodity lose situation is relatively light, residual values of commodity compared with
It is high.If the preferable user of credit applies returning goods, then it is likely used only to return of goods commodity lose that situation is heavier, residual valences of commodity
It is worth lower.
Using the area data of the return of goods user of returned item, user's regionalism of returned item can be generated.Area
Feature is specifically as follows province of receiving, city of receiving etc..The natural damage of some commodity lost in situation and logistics transportation
It is bad and consume it is related, it is some to have inconvenient traffic benefit or transport is easier to encounter the area of receiving of situation of jolting and is easier to cause commodity
Lose.Therefore, regionalism will cause influence to the residual value of return of goods commodity.
The return of goods opinion data described using the return of goods user of returned item, the consumers' opinions that returned item can be generated are special
Sign.The return of goods opinion data of return of goods user description can be description text, the description multi-modal datas such as voice and return of goods picture.
User often submits the return of goods opinion during carrying out Return Application, and describes return of goods reason by filling in Return Application reason
Or commodity lose situation.For example, natural language processing is carried out to the return of goods text of return of goods user description, alternatively, using it
Its neural network model carries out depth characteristic extraction to the return of goods photo that return of goods user shoots, and obtains user's meaning of returned item
See and understands feature.
For concrete example, the credit grade of return of goods user is A, and province where return of goods user is B, return of goods user description
Return of goods opinion data is return of goods text, and carrying out the description vocabulary obtained after natural language processing to return of goods text is C.Due to A, B,
C is classifying type data, carries out one-hot coding (one-hot) to A, B, C respectively and obtains user credit feature U1, user area spy
Levy U2, consumers' opinions feature U3.User characteristics (U1, U2, U3) is can be generated into U1, U2, U3 combination.
In step S104, using the product data of sample returned item, the article characteristics of sample returned item are generated.
For example, can determine the quantity in stock unit marks SKUID and commodity sign of return of goods commodity according to Return Application list
PID.The category data (such as brand, classification) of returned item, the rule of returned item can be obtained by the SKUID of return of goods commodity
Lattice data (such as volume, weight), the residual value statistical data (such as residual mean value with category article) with category article,
The time critical data (such as number of days in the shelf-life) of returned item can be obtained by commodity sign PID.
Using the category data of returned item, the category feature of returned item can be generated.The residual value of return of goods commodity
In general be positively correlated with brand quality, the residual values of return of goods commodity also as merchandise classification variation and change, such as day
Articles and fresh commodity often have damage rate more higher than other classification commodity.
Using the specification data of returned item, the specification feature of returned item can be generated.In general, return of goods commodity
Residual value may have certain incidence relation with volume, weight.
Using the time critical data of returned item, the timing characteristics of returned item can be generated.Utilize the purchase date, production date
Phase, Return Application date and shelf-life can determine the remaining shelf-life number of days of return of goods commodity, more than the commodity of quality guarantee period
Often there is lower residual value, the residual value of the remaining more commodity of shelf-life number of days may be higher.
Using the residual value statistical data of the same category article of returned item, the statistics that returned item can be generated is special
Sign.The residual value mean value of generic commodity is belonged to brand article residual value mean value, same period commodity residual value mean value
With the residual value statistical data of category article.If the same category commodity of commodity a within a certain period of time as occurred n times altogether
It returns goods, wherein n1Secondary residual value is x, n2Secondary residual value is y, n3Secondary residual value is z, wherein n1+n2+n3=N, then commodity
The residual value mean value of a is (n1*x+n2*y+n3*z)/N。
For concrete example, the category of returned item is D, and the specification of returned item is E, the day in the returned item shelf-life
Number is F, the residual mean value of the same category article of returned item is G.Since D, E, F are classifying type data, A, B, C are carried out respectively
One-hot coding obtains category feature U4, specification feature U5, timing characteristics U6；Since F is numeric type data, G is normalized
Obtain statistical nature U7.User characteristics (U4, U5, U6, U7) is can be generated into U4, U5, U6, U7 combination.
Step S106 marks the residual value of sample returned item.
In annotation process, the return of goods commodity of (completely new) are lost for nothing, it can be by cost price directly as residual value.It is right
In there is the return of goods commodity lost, can first determine whether the classification of return of goods commodity, by some classifications lose return of goods commodity (such as
Disposable product class, fresh class, opening class etc. commodity) residual value be directly determined as 0, without calculating commodity residual value.
For there are the other classification return of goods commodity of the overwhelming majority lost, point situation is needed to calculate the residual value of commodity.
If sample returned item needs to return genuine, by the difference of the cost price of sample returned item and return of goods logistics cost
Value, the residual value as sample returned item.If sample returned item needs to return supplier, by sample returned item
The difference of return price and return of goods logistics cost, the residual value as sample returned item.If sample returned item needs turn over
New sale, then by the corresponding difference for losing making-up price Yu return of goods logistics cost of sample returned item, as sample return of goods object
The residual value of product.If sample returned item needs depreciation to sell, by the difference of depreciation selling price and return of goods logistics cost,
Residual value as sample returned item.When determining specific return of goods logistics cost, it can be determined according to Return Application list
Sales order number determines logistics information numbering inquiry logistics order by sales order.
Step S108 is moved back using the user characteristics of sample returned item, the article characteristics of sample returned item and sample
The residual value of cargo product is trained deep learning neural network, enables deep learning neural network according to return of goods object
The user characteristics of product and the article characteristics of returned item, estimation obtain the residual value of returned item.
Fig. 2 shows the structural schematic diagrams for the deep learning neural network that the disclosure uses.As shown in Fig. 2, the disclosure is adopted
Deep learning neural network is that (depth residual value estimates network, Deep Residual Price Estimation to DRPEN
Network), following layers is specifically included.
(1) input layer can specifically include user characteristics input layer and article characteristics input layer, for merging input
User characteristics and article characteristics obtain the merging feature of returned item.In this layer, pretreated various features meeting is carried out
Merge combination according to the association of user characteristics and product features, generate new merging feature (U1, U2, U3, U4, U5, U6,
U7)。
(2) it is embedded in learning layer, for the full articulamentum with Embedding learning function.By presetting the hidden of fixed quantity
Containing the factor, it can will merge Feature Mapping to implicit factor space and obtain the feature vector of regular length, to original sharing feature
Again it is expressed in space.
(3) Implication learning layer specifically includes linear learning sub-network for the full articulamentum with Drop Out learning function
With Nonlinear Learning sub-network.Relationship of the feature vector for the regular length that different samples generate in implicit factor space is substantially
It is divided into linear relationship and two kinds of non-linear relation.Therefore, Implication learning layer is by two sub- network association structure compositions: a subnet
Network is the linear learning sub-network of multilayered structure, learns the non-linear relation between cross feature；Another sub-network is the double-deck knot
The Nonlinear Learning sub-network of structure.Wherein, linear learning sub-network is configured as the feature vector and returned item of study input
Residual value between linear relationship, output indicate linear relationship feature vector；Nonlinear Learning sub-network, is configured as
Learn the non-linear relation between the feature vector of input and the residual value of returned item, output indicates the spy of non-linear relation
Levy vector.Also include in the Nonlinear Learning sub-network of double-layer structure CT (Cross Transformation, cross over transition layer),
It is rearranged for the position to the implicit factor, to intersect study to user credit feature, user's regionalism, user
Opinion feature, the specification feature of returned item, the timing characteristics of returned item, the statistical nature of returned item, returned item
The combined situation of category feature.
(4) full articulamentum is configured as to indicate the feature vector and table of linear relationship by activation primitive Sigmoid
The maps feature vectors for showing non-linear relation are the residual value of returned item.Full articulamentum is in entire deep learning neural network
In play the role of classifier, each node of full articulamentum is connected with upper one layer of all nodes, for front is mentioned
The feature got carries out COMPREHENSIVE CALCULATING.
It is above-mentioned can iteration update deep learning network model, can not only learn simultaneously input feature and returned item
Residual value between linear relationship and non-linear relation, additionally it is possible to new data generation after be iterated update.
(2) test phase of deep learning neural network
One embodiment of the method for disclosure estimation returned item residual value is described below with reference to Fig. 3.
Fig. 3 shows the flow diagram of one embodiment of the method for disclosure estimation returned item residual value.Such as
Shown in Fig. 3, the method in the present embodiment includes step S302~step S308.
In step s 302, using the return of goods user data of returned item, the user characteristics of returned item are generated.It is specific real
Now it is referred to step S102.
In step s 304, using the product data of returned item, the article characteristics of returned item are generated；Specific implementation can
Referring to step S104.
In step S306, user characteristics and article characteristics are inputted into deep learning neural network, estimation is returned goods
The residual value of article.
270,000 forms datas are amounted to as training data for example, can choose from Return Application list, are chosen 30,000 forms datas and are made
For test data, training test ratio is 9:1.
Above-described embodiment consider during merchandise return in terms of commodity the residual value customer-side and article that are related to because
Element can carry out Effective selection and integrated treatment to related data, to more accurately and efficiently estimate the residual valence of article
Value.Through technical solution provided by the present disclosure, before goods return and replacement generation, it will be able to estimate commodity in advance to a certain extent
Residual value, and assisted effectively in view of different service single pairs answer the otherness of situation based on multi-source data in estimation procedure
The processing customer service of contact staff's intelligence promotes the after-sale service experience of user.
The device of the estimation article residual value of an embodiment of the present disclosure is described below with reference to Fig. 4.
Fig. 4 shows the structural schematic diagram of the device of the estimation article residual value of an embodiment of the present disclosure.Such as Fig. 4 institute
Show, the device 40 of the estimation article residual value in the present embodiment includes:
User characteristics generation module 402 is configured as the return of goods user data using returned item, generates returned item
User characteristics；
Article characteristics generation module 404 is configured as the product data using returned item, generates the article of returned item
Feature；
Residual value estimation module 406 is configured as user characteristics and article characteristics input deep learning nerve net
Network, estimation obtain the residual value of returned item.
In some embodiments, user characteristics generation module 402 is configured as executing at least one of following steps: benefit
With the credit data of the return of goods user of returned item, the user credit feature of returned item is generated；Utilize the return of goods of returned item
The area data of user generates user's regionalism of returned item；It is anticipated using the return of goods that the return of goods user of returned item describes
See data, generates the consumers' opinions feature of returned item.
In some embodiments, user characteristics generation module 402 is configured as: to the return of goods user description return of goods text into
Row natural language processing obtains the use of returned item alternatively, carrying out depth characteristic extraction to the return of goods photo of return of goods user shooting
Family opinion understands feature.
In some embodiments, article characteristics generation module 404 is configured as executing at least one of following steps: benefit
With the category data of returned item, the category feature of returned item is generated；Using the specification data of returned item, return of goods object is generated
The specification feature of product；Using the time critical data of returned item, the timing characteristics of returned item are generated；Utilize the same product of returned item
The residual value statistical data of class article, generates the statistical nature of returned item.
In some embodiments, deep learning neural network includes: linear learning sub-network, is configured as study input
Linear relationship between feature vector and the residual value of returned item, output indicate the feature vector of linear relationship；It is non-linear
Learn sub-network, the non-linear relation being configured as between the feature vector of study input and the residual value of returned item is defeated
The feature vector of non-linear relation is indicated out；Full articulamentum is configured as to indicate the feature of linear relationship by activation primitive
Vector and the maps feature vectors for indicating non-linear relation are the residual value of returned item.
In some embodiments, deep learning neural network further include: input layer, the user for being configured as merging input are special
Sign and article characteristics, obtain the merging feature of returned item；It is embedded in learning layer, is configured with the hidden of default fixed quantity
Containing the factor, it can will merge Feature Mapping to implicit factor space and obtain the feature vector of regular length.
In some embodiments, which further includes neural metwork training module 400, is configured as: being moved back using sample
The user data of cargo product generates the user characteristics of sample returned item；Using the product data of sample returned item, sample is generated
The article characteristics of this returned item；Mark the residual value of sample returned item；Utilize the user characteristics of sample returned item, sample
The article characteristics of this returned item and the residual value of sample returned item are trained deep learning neural network, so that
Deep learning neural network can be estimated to be returned goods according to the user characteristics of returned item and the article characteristics of returned item
The residual value of article.
In some embodiments, neural metwork training module 400 is configured as: if sample returned item needs to return original
Factory, the then residual value by the difference of the cost price of sample returned item and return of goods logistics cost, as sample returned item；If
Sample returned item needs to return supplier, then by the difference of the return price of sample returned item and return of goods logistics cost, makees
For the residual value of sample returned item；If sample returned item needs to renovate sale, by the corresponding folding of sample returned item
Damage the difference of making-up price and return of goods logistics cost, the residual value as sample returned item；If sample returned item needs
Depreciation sale, the then residual value by the difference of depreciation selling price and return of goods logistics cost, as sample returned item.
Above-described embodiment consider during merchandise return in terms of commodity the residual value customer-side and article that are related to because
Element can carry out Effective selection and integrated treatment to related data, to more accurately and efficiently estimate the residual valence of article
Value.Through technical solution provided by the present disclosure, before goods return and replacement generation, it will be able to estimate commodity in advance to a certain extent
Residual value, and assisted effectively in view of different service single pairs answer the otherness of situation based on multi-source data in estimation procedure
The processing customer service of contact staff's intelligence promotes the after-sale service experience of user.
Fig. 5 shows the structural schematic diagram of the device of the estimation article residual value of the disclosure another embodiment.Such as Fig. 5
Shown, the device 50 of the estimation article residual value of the embodiment includes: memory 510 and is coupled to the memory 510
Processor 520, processor 520 are configured as executing any one aforementioned embodiment based on the instruction being stored in memory 510
In estimation article residual value method.Wherein, memory 510 for example may include system storage, fixation it is non-volatile
Storage medium etc..System storage be for example stored with operating system, application program, Boot loader (Boot Loader) with
And other programs etc..
Estimate that the device 50 of article residual value can also connect including input/output interface 530, network interface 540, storage
Mouth 550 etc..It can for example be connected by bus 560 between these interfaces 530,540,550 and memory 510 and processor 520
It connects.Wherein, input/output interface 530 is display, the input-output equipment such as mouse, keyboard, touch screen provide connecting interface.Net
Network interface 540 provides connecting interface for various networked devices.The external storages such as memory interface 550 is SD card, USB flash disk provide company
Connection interface.
The disclosure further includes a kind of computer readable storage medium, is stored thereon with computer instruction, and the instruction is processed
Device realizes the estimation article residual value in any one aforementioned embodiment method when executing.
It should be understood by those skilled in the art that, embodiment of the disclosure can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the disclosure
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the disclosure, which can be used in one or more,
The calculating implemented in non-transient storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) can be used
The form of machine program product.
The disclosure is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present disclosure
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
The foregoing is merely the preferred embodiments of the disclosure, not to limit the disclosure, all spirit in the disclosure and
Within principle, any modification, equivalent replacement, improvement and so on be should be included within the protection scope of the disclosure.
Claims (18)
1. a kind of method for estimating article residual value, comprising:
Using the return of goods user data of returned item, the user characteristics of the returned item are generated；
Using the product data of the returned item, the article characteristics of the returned item are generated；
The user characteristics and the article characteristics are inputted into deep learning neural network, estimation obtains the returned item
Residual value.
2. the method for claim 1, wherein return of goods user data using returned item, generates the return of goods
The user characteristics of article at least one of include the following steps:
Using the credit data of the return of goods user of returned item, the user credit feature of the returned item is generated；
Using the area data of the return of goods user of returned item, user's regionalism of the returned item is generated；
The return of goods opinion data described using the return of goods user of returned item, generates the consumers' opinions feature of the returned item.
3. method according to claim 2, wherein the return of goods opinion number that the return of goods user using returned item describes
According to the consumers' opinions feature for generating the returned item includes:
Natural language processing is carried out to the return of goods text of return of goods user description, alternatively, shining the return of goods of return of goods user shooting
Piece carries out depth characteristic extraction, and the consumers' opinions for obtaining the returned item understands feature.
4. the method for claim 1, wherein product data using the returned item, generates the return of goods
The article characteristics of article at least one of include the following steps:
Using the category data of returned item, the category feature of the returned item is generated；
Using the specification data of returned item, the specification feature of the returned item is generated；
Using the time critical data of returned item, the timing characteristics of the returned item are generated；
Using the residual value statistical data of the same category article of returned item, the statistical nature of the returned item is generated.
5. the method for claim 1, wherein the deep learning neural network includes:
Linear learning sub-network is configured as the linear pass between the feature vector of study input and the residual value of returned item
System, output indicate the feature vector of the linear relationship；
Nonlinear Learning sub-network is configured as non-thread between the feature vector of study input and the residual value of returned item
Sexual intercourse, output indicate the feature vector of the non-linear relation；
Full articulamentum is configured as by activation primitive that the feature vector and expression that indicate the linear relationship is described non-thread
The maps feature vectors of sexual intercourse are the residual value of returned item.
6. method as claimed in claim 5, wherein the deep learning neural network further include:
Input layer is configured as merging the user characteristics of input and the article characteristics, obtains the returned item
Merge feature；
It is embedded in learning layer, is configured with the implicit factor of default fixed quantity, it can be by the merging Feature Mapping to hidden
The feature vector of regular length is obtained containing factor space.
7. the method for claim 1, wherein the method also includes:
Using the user data of sample returned item, the user characteristics of sample returned item are generated；
Using the product data of sample returned item, the article characteristics of sample returned item are generated；
Mark the residual value of sample returned item；
Utilize the residual valence of the user characteristics of sample returned item, the article characteristics of sample returned item and sample returned item
Value is trained the deep learning neural network, enables the deep learning neural network according to the use of returned item
The article characteristics of family feature and returned item, estimation obtain the residual value of returned item.
8. the method for claim 7, wherein it is described mark sample returned item residual value include:
If sample returned item needs to return genuine, by the difference of the cost price of sample returned item and return of goods logistics cost,
Residual value as sample returned item；
If sample returned item needs to return supplier, by the difference of the return price of sample returned item and return of goods logistics cost
Value, the residual value as sample returned item；
If sample returned item needs to renovate sale, by sample returned item it is corresponding lose making-up price and return of goods logistics at
This difference, the residual value as sample returned item；
If sample returned item needs depreciation to sell, by the difference of depreciation selling price and return of goods logistics cost, as sample
The residual value of returned item.
9. a kind of device for estimating article residual value, comprising:
User characteristics generation module is configured as the return of goods user data using returned item, generates the use of the returned item
Family feature；
Article characteristics generation module is configured as the product data using the returned item, generates the object of the returned item
Product feature；
Residual value estimation module is configured as the user characteristics and article characteristics input deep learning nerve net
Network, estimation obtain the residual value of the returned item.
10. device as claimed in claim 9, wherein the user characteristics generation module is configured as executing in following steps
At least one:
Using the credit data of the return of goods user of returned item, the user credit feature of the returned item is generated；
Using the area data of the return of goods user of returned item, user's regionalism of the returned item is generated；
The return of goods opinion data described using the return of goods user of returned item, generates the consumers' opinions feature of the returned item.
11. device as claimed in claim 10, wherein the user characteristics generation module is configured as:
Natural language processing is carried out to the return of goods text of return of goods user description, alternatively, shining the return of goods of return of goods user shooting
Piece carries out depth characteristic extraction, and the consumers' opinions for obtaining the returned item understands feature.
12. device as claimed in claim 9, wherein the article characteristics generation module is configured as executing in following steps
At least one:
Using the category data of returned item, the category feature of the returned item is generated；
Using the specification data of returned item, the specification feature of the returned item is generated；
Using the time critical data of returned item, the timing characteristics of the returned item are generated；
Using the residual value statistical data of the same category article of returned item, the statistical nature of the returned item is generated.
13. device as claimed in claim 9, wherein the deep learning neural network includes:
Linear learning sub-network is configured as the linear pass between the feature vector of study input and the residual value of returned item
System, output indicate the feature vector of the linear relationship；
Nonlinear Learning sub-network is configured as non-thread between the feature vector of study input and the residual value of returned item
Sexual intercourse, output indicate the feature vector of the non-linear relation；
Full articulamentum is configured as by activation primitive that the feature vector and expression that indicate the linear relationship is described non-thread
The maps feature vectors of sexual intercourse are the residual value of returned item.
14. device as claimed in claim 13, wherein the deep learning neural network further include:
Input layer is configured as merging the user characteristics of input and the article characteristics, obtains the returned item
Merge feature；
It is embedded in learning layer, is configured with the implicit factor of default fixed quantity, it can be by the merging Feature Mapping to hidden
The feature vector of regular length is obtained containing factor space.
15. device as claimed in claim 9, wherein described device further includes neural metwork training module, is configured as:
Using the user data of sample returned item, the user characteristics of sample returned item are generated；
Using the product data of sample returned item, the article characteristics of sample returned item are generated；
Mark the residual value of sample returned item；
Utilize the residual valence of the user characteristics of sample returned item, the article characteristics of sample returned item and sample returned item
Value is trained the deep learning neural network, enables the deep learning neural network according to the use of returned item
The article characteristics of family feature and returned item, estimation obtain the residual value of returned item.
16. device as claimed in claim 9, wherein the neural metwork training module is configured as:
If sample returned item needs to return genuine, by the difference of the cost price of sample returned item and return of goods logistics cost,
Residual value as sample returned item；
If sample returned item needs to return supplier, by the difference of the return price of sample returned item and return of goods logistics cost
Value, the residual value as sample returned item；
If sample returned item needs to renovate sale, by sample returned item it is corresponding lose making-up price and return of goods logistics at
This difference, the residual value as sample returned item；
If sample returned item needs depreciation to sell, by the difference of depreciation selling price and return of goods logistics cost, as sample
The residual value of returned item.
17. a kind of device for estimating article residual value, comprising:
Memory；And
It is coupled to the processor of the memory, the processor is configured to the instruction based on storage in the memory,
Execute the method such as estimation article residual value described in any item of the claim 1 to 8.
18. a kind of computer readable storage medium, wherein the computer-readable recording medium storage has computer instruction, institute
State the method realized when instruction is executed by processor such as estimation article residual value described in any item of the claim 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810831981.2A CN109034899A (en) | 2018-07-26 | 2018-07-26 | Estimate the method, apparatus and computer readable storage medium of article residual value |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810831981.2A CN109034899A (en) | 2018-07-26 | 2018-07-26 | Estimate the method, apparatus and computer readable storage medium of article residual value |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109034899A true CN109034899A (en) | 2018-12-18 |
Family
ID=64646488
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810831981.2A Pending CN109034899A (en) | 2018-07-26 | 2018-07-26 | Estimate the method, apparatus and computer readable storage medium of article residual value |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109034899A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503368A (en) * | 2019-08-13 | 2019-11-26 | 蚌埠聚本电子商务产业园有限公司 | A kind of handling return method and system for e-commerce |
| CN115081712A (en) * | 2022-06-21 | 2022-09-20 | 南京航空航天大学 | Delivery scheduling and resource optimization method for passenger plane collaborative development double-layer project structure |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1828663A (en) * | 2005-02-25 | 2006-09-06 | 科帕特公司 | Systems and methods for determining vehicle salvage value |
| CN101628572A (en) * | 2009-07-23 | 2010-01-20 | 烟台麦特电子有限公司 | Automobile maintenance reminder and residual value calculation display method and vehicle-mounted device |
| CN104077703A (en) * | 2014-06-19 | 2014-10-01 | 五八同城信息技术有限公司 | Second-hand product condition calculation method |
| US8949050B2 (en) * | 2011-12-16 | 2015-02-03 | Basen Corporation | Smartgrid energy-usage-data storage and presentation systems, devices, protocol, and processes including a visualization, and load fingerprinting process |
-
2018
- 2018-07-26 CN CN201810831981.2A patent/CN109034899A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1828663A (en) * | 2005-02-25 | 2006-09-06 | 科帕特公司 | Systems and methods for determining vehicle salvage value |
| CN101628572A (en) * | 2009-07-23 | 2010-01-20 | 烟台麦特电子有限公司 | Automobile maintenance reminder and residual value calculation display method and vehicle-mounted device |
| US8949050B2 (en) * | 2011-12-16 | 2015-02-03 | Basen Corporation | Smartgrid energy-usage-data storage and presentation systems, devices, protocol, and processes including a visualization, and load fingerprinting process |
| CN104077703A (en) * | 2014-06-19 | 2014-10-01 | 五八同城信息技术有限公司 | Second-hand product condition calculation method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503368A (en) * | 2019-08-13 | 2019-11-26 | 蚌埠聚本电子商务产业园有限公司 | A kind of handling return method and system for e-commerce |
| CN115081712A (en) * | 2022-06-21 | 2022-09-20 | 南京航空航天大学 | Delivery scheduling and resource optimization method for passenger plane collaborative development double-layer project structure |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ika et al. | The “re-meaning” of project success: Updating and recalibrating for a modern project management | |
| Bhat et al. | Antecedents of tourist loyalty to tourist destinations: a mediated-moderation study | |
| Chinomona | The influence of brand experience on brand satisfaction, trust and attachment in South Africa | |
| CN109582876A (en) | Tourism industry user portrait building method, device and computer equipment | |
| Averina et al. | Impact of digital technologies on the company’s business model | |
| CN113902522A (en) | Patent recommendation method and terminal based on graph neural network | |
| Lazo-Amado et al. | e-Business model to optimise sales through digital marketing in a Peruvian Company | |
| Thorn et al. | Internal technology valuation: real world issues | |
| Calderón et al. | SMEs’ dynamic learning capabilities in international public procurement | |
| CN109034899A (en) | Estimate the method, apparatus and computer readable storage medium of article residual value | |
| US8019638B1 (en) | Dynamic construction of business analytics | |
| Wang et al. | A Bayesian combination forecasting model for retail supply chain coordination | |
| O’Meara et al. | Quantum software architecture blueprints for the cloud: Overview and application to peer-2-peer energy trading | |
| Zema et al. | Deep learning and forecasting in practice: an alternative costs case | |
| CN118096205A (en) | Electronic commerce marketing intelligent decision-making method and system based on multi-modal learning | |
| Yang et al. | Investigating the crucial role of logistics service quality in customer satisfaction for fresh e-commerce: A mutually validating method based on SERVQUAL and service encounter theory | |
| Yamusa et al. | Developing a machine learning integrated e-procurement system for Nigerian public procuring entities | |
| Fontana | The AI-First Company: How to Compete and Win with Artificial Intelligence | |
| Roe et al. | Optimizing supply chain performance: information sharing and coordinated management | |
| Shahryari Nia et al. | A Hybrid Approach to Develop a Structural Model of Factors Affecting Supply Chain Collaboration in Home Appliance Industry | |
| Lin et al. | An assessment framework for the purchase of 3D CAD software in manufacturing industries | |
| Sajedikhah et al. | Historical effect of verified purchases and ordinary users’ testimonials on the sales rank of experience and search goods | |
| KR102453673B1 (en) | System for sharing or selling machine learning model and operating method thereof | |
| Ajay et al. | Analyzing and Predicting the Sales Forecasting using Modified Random Forest and Decision Tree Algorithm | |
| Wahdiniwaty et al. | Comparative analysis of software quality model in the selection of marketplace E-commerce |
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: 20181218 |
|
| RJ01 | Rejection of invention patent application after publication |