CN110020136A - Object recommendation method and relevant device - Google Patents
Object recommendation method and relevant device Download PDFInfo
- Publication number
- CN110020136A CN110020136A CN201711104368.2A CN201711104368A CN110020136A CN 110020136 A CN110020136 A CN 110020136A CN 201711104368 A CN201711104368 A CN 201711104368A CN 110020136 A CN110020136 A CN 110020136A
- Authority
- CN
- China
- Prior art keywords
- user
- retrieval
- target
- classification
- access
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
- G06Q30/0635—Processing of requisition or of purchase orders
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
This application provides a kind of object recommendation methods, this method is in the case where accessing the user of web object is target type user, determine user to the search pattern of web object, different search patterns can reflect user's access different to web object and be intended to, and the target object found according to search pattern can meet user's intention.As it can be seen that object recommendation method provided by the present application can recommend the object for meeting its intention for the target type user with different intentions.In addition, present invention also provides object recommendation relevant device, to guarantee the application and realization of the method in practice.
Description
Technical field
This application involves Internet technical fields, more specifically, being object recommendation method and relevant device.
Background technique
It buys on behalf, is a kind of novel commodity purchasing mode, help practical consumer purchases goods for consumer, and pass through the difference of dealing
Valence or a certain proportion of commission are returned.The commodity bought for consumer are typically from e-commerce platform, electronics
Business platform can recommend it that may feel emerging according to the buying behavior for consumer in order to promote the sale of more commodity to for consumer
The commodity of interest.
The Floor layer Technology of the above business model is achieved in that e-commerce platform is in the numerous user accounts for buying commodity
In identify for consumer's account, and be the commodity for recommending same or like category for consumer that are indicated for consumer's account.This technology
In scheme, the commodity interested for this type buyer of consumer how are determined, be a technical issues that need to address.
Summary of the invention
In view of this, this application provides a kind of object recommendation method, to recommend for the target types user such as the person of buying on behalf
The objects such as dependent merchandise.
In order to achieve the object, technical solution provided by the present application is as follows:
In a first aspect, this application provides a kind of object recommendation methods, comprising:
If the user for accessing web object is target type user, it is determined that lookup of the user to the web object
Mode；
According to the search pattern, target object is searched；
The relevant information of the target object is sent to the user.
Second aspect, this application provides a kind of object recommendation equipment, comprising:
Processor, if the user for accessing web object is target type user, it is determined that the user is to the net
The search pattern of page object；And according to the search pattern, search target object；
Communication interface, for the relevant information of the target object to be sent to the user.
The third aspect, this application provides a kind of object recommendation devices, comprising:
Search pattern determining module, if the user for accessing web object is target type user, it is determined that the use
Search pattern of the family to the web object；
Target object searching module, for searching target object according to the search pattern；
Object information sending module, for the relevant information of the target object to be sent to the user.
From the above technical scheme, this application provides a kind of object recommendation methods, and this method is in access web object
User be target type user in the case where, determine user to web object search pattern, different search patterns is to application
The family access different to web object is intended to, and the access that the target object found according to search pattern can meet user needs
It asks.As it can be seen that object recommendation method provided by the present application can recommend to meet its meaning for the target type user with different intentions
The object of figure.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, 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
The embodiment of application for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the configuration diagram of object recommendation system provided by the present application；
Fig. 2 is train classification models provided by the present application and is identified the flow diagram of user using model；
Fig. 3 is a kind of flow diagram of object recommendation method provided by the present application；
Fig. 4 is another flow diagram of object recommendation method provided by the present application；
Fig. 5 is a kind of structural schematic diagram of object recommendation device provided by the present application；
Fig. 6 is another structural schematic diagram of object recommendation device provided by the present application；
Fig. 7 is a kind of hardware frame schematic diagram of object recommendation equipment provided by the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall in the protection scope of this application.
It buys on behalf in this business model, is related to tripartite's main body, be for consumer, practical buyer and E-business service respectively
Provider.A kind of concrete application scene bought on behalf is that Taobao (a kind of E-business service provider) can be specifically for rural area
User (a kind of practical buyer) issues some merchandise newss, and village school two (a kind of generation consumer) is according to the demand of rural subscribers, at this
It is retrieved in kind merchandise news and buys rural subscribers' needs or interested commodity.
It is server and end respectively as shown in Figure 1, realizing that the Technical Architecture of this business model mainly includes two method, apparatus
End；Wherein server be E-business service provider deployment include extensive stock information equipment, terminal be for consumer
The equipment of one side accesses web object for consumer by terminal login service device, for example, according to practical buyer demand to commodity
Browsed plus purchased the operations such as (addition shopping cart), purchase.It should be noted that server mainly realizes the function of e-commerce
Can, therefore it is properly termed as electric business server or electric business platform.Electric business server is in order to sell more commodity, also for help generation
Consumer more easily finds required commodity, can be to for consumer's Recommendations.As shown in Figure 1, server is needed according to access number
According to, the determining object recommended, and recommended is returned to terminal.
How electric business server is identified in numerous access users for this kind of user of consumer, and determines such user
Interested commodity are all the technical issues that need to address.
Currently, commercial product recommending is a kind of common technology of various electric business servers, user can be collected in electric business service
Behavioral data on device in longer period of time according to behavioral data structuring user's model and portrays user preference, Jin Eryi
The recommendation of corresponding commodity is carried out according to user model and user preference.However, this kind of general recommended technology is not particularly suited for generation
Scene is purchased, main cause is, the actual purchase user number serviced under normal conditions for consumer is more, for the behavior number of consumer
According to the summation for being a large amount of actual purchase user behavior datas.Based on the main cause, if will appear using common recommended technology
Following problems.
First, since the commodity that the historical behavior data for consumer are related to are the remittances of a large amount of actual purchase user demand commodity
Always, it includes commodity classification it is many and diverse and span is larger, therefore select in these diversified commodity the quotient recommended
Product are more difficult.Second, certain is that the commodity recommended for consumer are chosen according to the historical behavior data for consumer, so
And the actual purchase person serviced in behavior is bought on behalf this certain for consumer, it is larger that may to buy behavior on behalf with history not identical,
Therefore history behavioral data and do not have reference.
There is also another recommended technologies in the prior art, i.e. the recommendation based on public account.Public account can be associated with
Multiple sub- accounts of individual construct preference mould respectively according to the historical behavior data of each sub- account of individual for the sub- account of the individual
Type, and collective's preference pattern of public account is formed after the preference pattern of all sub- accounts of individual is weighted, and then basis
Collective's preference pattern provides same commodity for multiple individual.
For example, tourism website is provided with the public account of family, the public account of family can be associated with several kinsfolks
The sub- account of individual, the viewing that website can count each kinsfolk respectively according to the historical data of each sub- account of individual likes
Good model, then the viewing modeling hobbies of family's entirety are generated after the viewing modeling hobbies of each kinsfolk are weighted.Multiple
When kinsfolk has common viewing demand, according to the viewing modeling hobbies of family entirety, recommend associated film for the family.
But this recommended technology based on public account is also not particularly suited for buying recommendation scene, specific unworthiness on behalf
It can be presented as the following.
First, the object that public account is recommended is a group, thus calculate be collective preference pattern.However in generation
It purchases in scene, between the actual purchase person that is serviced for consumer and onrelevant, the historical data of other actual purchase persons can not
Recommendation foundation as certain actual purchase person.Second, individual account associated by public account is for electric business server
Known, electric business server can clearly distinguish which historical data belongs to which individual account, so as to according to ownership
Respective preference pattern is calculated in the historical data of each individual account, however in buying scene on behalf, the reality serviced for consumer
There is no differentiations for the historical data of border buyer, therefore cannot calculate the inclined of each actual purchase person according to respective historical data
Good situation.
The prior art is mainly described above how to account Recommendations, for the user's account how identified for consumer
Number, there is no a kind of relevant technical solutions at present.Existing public account is explicit, if that is, user registers public account
Number, it is public account that register platforms, which can clearly mark the account, and electric business platform receives an access request, according to access account
It can determine with the presence or absence of label using whether the account is public account.But behavior is bought on behalf for electric business platform
It is transparent, there is no the user is not clearly marked is for consumer for user data such as register account number of consumer etc., therefore electric business is flat
After platform receives an access request, directly the user account can not be determined for the account for consumer according to user account.
In this regard, this application provides a kind of user identification methods.In a kind of application scenarios, electric business server be can be used
This method identifies in largely access user for the such user of consumer.But it should be recognized that this method not office
It is limited to buy scene on behalf, as long as the scene of user aid crowd's multi-user access electric business object provided by the server can be with
It is applicable in.Therefore, the user which is identified is not limited to for consumer, is the tool of this user for consumer
The user of the type can be become target type user for ease of description by body example.
User's identification is mainly according to identification model trained in advance, and the process of identification model training mainly includes walking as follows
Rapid A1~A3.
A1: sample of users data are obtained.
Wherein, sample of users data include positive sample user data and negative sample user data.Positive sample user data refers to
Be user data labeled as target type user, it should be noted that the user data labeled as positive sample is specific
Target type user.
For example, being washed in a pan in business scenario in the village of Taobao, infused on the electric business platform of the naughty business in the village personnel Hui in village service station
Volume " village school two " account, and using " village school two " account is that villager carries out the service of buying on behalf, therefore these " village school two " accounts registered
It number is the user data of specific target type user, and then the account data of these " village school two " accounts can be used as positive sample
This user data.Certainly, it is only under a kind of specific business scenario for example, it should not become pair that business is washed in a pan in the above village
The restriction of the application application scenarios.
Negative sample user data is some use for obtaining at random in the user data for not being labeled as target type user
User data.
A2: from sample of users data, user characteristics are extracted.
Wherein, the user characteristics in terms of which are extracted, these aspects are preset.
For example, preset extract user within a certain period of time enliven number of days, average daily liveness, purchase commodity classification
Deng.Wherein, the number of days that enlivens of user within a certain period of time can be, and user accesses object on electric business server within a certain period of time
Number of days；Average daily liveness can be, the average daily duration for accessing object on electric business server.
The user characteristics that user extracts the following aspects are extracted more specifically, for example presetting:
(a) number of days is enlivened in 3 days, 7 days and 15 days；
(b) conclusion of the business odd number and gross turnover in 3 days, 7 days and 15 days；
(c) classification for the commodity clicked plus purchased in 3 days, 7 days and 15 days and buy；
(d) 3 days, 7 days and 15 days daily to enliven duration；
(e) become buying ratio of the retrieval commodity of commodity in total retrieval.
Time span and aspect feature in above example are only to illustrate, and the application is not limited thereto.
It is understood that executing access operation (access for want help a large amount of user of the target types user such as consumer
Operation includes logging in electric business server and accessing the object on electric business server), therefore, target type user compares non-target class
Type user has any one or more in following features, and if active degree wants high, the classification of the object accessed is relatively more,
The conclusion of the business conversion ratio of commodity is higher.Why the feature in above-mentioned example is extracted in setting, is because the feature in terms of these can
The characteristics of reflecting above-mentioned target type user.
A3: training user's feature, to obtain identification model.
Wherein it is possible to be calculated using existing sorting algorithm such as logistic regression (Logistic Regression, referred to as LR)
Method is trained the user characteristics extracted from sample of users data, and the model obtained after training can be to other users
Classify, with identify other users whether be target type user.Therefore, which is properly termed as identification model or classification
Model.
In conclusion as shown in Fig. 2, the process for obtaining model mainly includes three steps, respectively sample collection, feature
Extraction and model training.In addition, some features of feature extraction can be the class for enlivening situation, the accessed object of user of user
Mesh distribution situation, the purchase situation of user, conclusion of the business situation of the accessed object of user etc..Certainly, in practical applications, extract
Feature be not limited thereto.
After training obtains identification model, when needing to identify the type of some user, it can be extracted from the user data
The feature of above-mentioned default aspect, and the user characteristics extracted are inputted in the identification model, which can determine
The user whether be target type user.
In the case where identifying user is target type user, it is some right further to recommend for target type user
As.It should be noted that object can be commodity, it is also possible to other objects.Identify whether user is target type user
Mode can also be other, it is not limited to identified using training pattern.
It should be noted that access of the target type user to web object, may include a variety of situations.With buy application on behalf
May be bought goods for client in some cases for consumer for scene, be random browsing quotient in the case of possible other
Product webpage does experience deposit for subsequent selective purchase.In varied situations, it needs to recommend different types of object for user.
This application provides a kind of object recommendation method, the object that this method is recommended can be commodity, be also possible to it
The article of his property.See Fig. 3, it illustrates a kind of processes of item recommendation method provided by the present application, specifically include following step
Rapid S301~S303.
S301: if the user of access web object is target type user, it is determined that lookup mould of the user to web object
Formula.
Wherein, after receiving user to the access request of web object, the type of the user can be identified.Tool
The recognition methods of body can identify the type of user according to above-mentioned user identification method.If identifying, user belongs to target type
User then may further determine that search pattern of the user to the web object accessed.
The access of web object is intended to it should be noted that search pattern can reflect user, access is intended to can wrap
It includes but is not limited to two kinds, accurate lookup and arbitrarily browsing.For buying application scenarios on behalf, it is accurate search searched for consumer or
The commodity of certain concrete type are chosen, arbitrarily browsing looks in commodity website for consumer, analyzes current commodity
Feature, in order to do experience deposit for some characteristic commodity of other customer recommendations.
In one example, it determines that user is to the specific implementation of the search pattern of web object, obtains user couple
The access path of web object；And according to access path, determine user to the search pattern of web object.Specifically, it accesses
Path includes the Web page module that user is accessed in webpage, and may include the duration and number for accessing these Web page modules.
Wherein, different Web page modules can represent user to the intention of web object, and packet is determined from access path
User can be determined to webpage pair containing which Web page module, and to the access times and access duration of these Web page modules
Being intended that for elephant is any.For example, Web page module may include homepage, retrieved page, object details page etc.；Wherein homepage indicates to use
Family is intended to arbitrarily browse to web object, and retrieved page indicates that user is intended to accurately search to web object, and object is detailed
Feelings page also illustrates that user is intended to accurately search to web object.Therefore, if user's long period or the access of more number are first
Page, then it represents that user's is intended to arbitrarily browse；If user's long period or more number access retrieved page and object details page,
Then indicate that user is intended to accurately to search.
In addition, the object classification of user's access can be determined according to Web page module, it is more according to whether object classification has
Sample can also determine that user is intended to the access of web object.Specifically, if object classification has diversity, then it represents that use
Family is intended to arbitrarily browse to the access of web object, if object classification does not have diversity, then it represents that user is to web object
Access be intended to accurately search.The foundation of this kind of method of determination is, if user is in random browse state, the object of access
It is diversified, if user is generally focused on a small amount of several accurately searching state, the object of access.It therefore can be with
Whether there is diversity according to access object, to determine that the access of user is intended to.
It should be noted that different Web page modules, determines that object class purpose mode is different.For example, headed by Web page module
Page, multiple regions are generally comprised in homepage, different zones correspond to inhomogeneity purpose object, therefore can be browsed according to user
Region determines object classification.For another example, Web page module is retrieved page, then can be extracted according to user in the search terms that retrieved page inputs
Object classification, such as user search " 2017 winters trendy frivolous down jackets schoolgirl money ", then the object classification extracted includes
Down jackets.For another example, Web page module is object details page, it is determined that classification belonging to object in object details page, such as object are detailed
Feelings page is the details page of a cowboy's one-piece dress, then can determine that classification belonging to object is one-piece dress.
By it is described above it is found that user access path the characteristics of can reflect out user and anticipate to the access of web object
Figure.It is intended to according to access, user can be directly determined out to the search pattern of web object.Search pattern indicate be according to why
The rule searching object of sample, the object found are accurate detailed or fuzzy large-scale.If such as access is intended to
Accurate to search, then its corresponding search pattern is accurate searches；If access is intended to arbitrarily browse, corresponding search pattern
For fuzzy search.For ease of description, the search pattern determined is properly termed as target search pattern.
S302: according to search pattern, target object is searched.
It wherein, include preset object search rule in search pattern, object search rule is selected from numerous objects
Some objects are selected as target object.Different search patterns is that the levels of precision of user's recommended is different.It is understood that
It is that, if search pattern is accurate lookup, the user that object search rule can obtain in this access is intended to, and anticipates according to user
Figure selector shares the object of family intention as target object；If search pattern is fuzzy search, object search rule can root
Determine the object with certain features as target object according to history access data.
S303: the relevant information of target object is sent to user.
Wherein it is possible to the relevant information of target object is searched from server, it may further be by the correlation of target object
Information is included in webpage and is sent to user, which can not be individual webpage, can be included in other webpage informations
In same webpage.For buying application scenarios on behalf, it is assumed that the target object determined is 5 one-piece dresses, then can be in shopping cart
Some region of webpage shows the picture of 5 one-piece dresses.
From the above technical scheme, this application provides a kind of object recommendation methods, and this method is in access web object
User be target type user in the case where, determine user to the search pattern of web object, different search patterns can be with
Reflect that user's access different to web object is intended to, the target object found according to search pattern can meet user's meaning
Figure.As it can be seen that object recommendation method provided by the present application can recommend to meet its meaning for the target type user with different intentions
The object of figure.
Preceding to have addressed, user is intended to include accurate lookup to the access of web object and arbitrarily two kinds of browsing, access are intended to
Corresponding search pattern includes accurate lookup and two kinds of fuzzy search.Below to two kinds of search patterns how to search target object into
Row explanation.
If search pattern is accurate lookup, described according to the search pattern, the specific implementation side of target object is searched
Formula includes the following steps B1~B3.
B1: the access data that user generates during this accesses web object are obtained.
Wherein, due to the user that user is target type, and the user of the target type has the special feature that and is, every time to net
Page object the process accurately searched all be it is independent, have no incidence relation with the accurate search procedure of history.To buy applied field on behalf
For scape, for consumer, this accesses webpage to buy certain type of commodity on behalf, the quotient that this commodity bought on behalf was bought on behalf with history
Product and onrelevant, therefore the feature of commodity bought on behalf of history can not buy commodity on behalf for this and provide reference.
Therefore, access data acquired in this step are the access data that user generates in this access process.Specifically
For, user is, with the continuity of time, to be continuously generated user couple within one section of duration to the primary access of web object
The access data of web object.Some time point in access process obtained in the secondary access process before the time point
Data are accessed, determine that user accurately searched is the object of which classification according to these access data.
For buying application scenarios on behalf, it is assumed that perform following access operations when certain access web site commodity of user, browse
Homepage repeatedly inputs search terms, after obtaining search result, clicks certain commodity in search result, certain commodity are put into
Shopping cart has purchased certain commodity.User executes each access operation, and server can obtain accessing data accordingly, such as defeated
What the search terms entered are, click which commodity, add and which commodity purchased, and have purchased which commodity etc..Server is obtaining
After accessing data, following two steps can be executed according to access data.It should be noted that as long as user executes access behaviour
Make, server can obtain accessing data accordingly, and server can according to the preset time interval, repeatedly according to obtained by
Access data be user recommend target object.Normally, the access data that server obtains are more, then the target pair retrieved
It, therefore, can be than the preceding object once recommended per the object once recommended afterwards in a web object access process as more accurate
It is more accurate.
It should be noted that if being led due to deficiency of access data etc. when recommending early period or when certain recommendation
Cause can not get the alternative objects of the condition of satisfaction, then can will access pair with certain characteristics that data determine according to history
As target object.Some of them characteristic may include but be not limited to the low price etc. that disappears fastly.
B2: different types of retrieval source is obtained from access data, and inhomogeneity is retrieved according to different types of retrieval source
The alternative objects of type.
Wherein, access data can indicate which access operation user once performed, and access operation can reflect user and think
What the object to be searched is, therefore can extract a variety of different access operations as retrieval source from access data, is used
Corresponding object is retrieved in retrieval source, and will retrieve object alternately object.
For example, retrieval source include it is following it is several in it is any a variety of: the pointed web object of target type operation, retrieval
The band of position where word, user.Respectively to the retrieval source of the three types and how according to retrieval source obtain alternative objects into
Row explanation.
The retrieval source of first seed type.In buying application scenarios on behalf, target type operation also refers to clicking operation, adds
Enter shopping cart operation, purchase operation etc., and then web object pointed by target type operation refers respectively to, the quotient of click
Product, the commodity that the commodity of shopping cart, purchase are added.User executes the operation of these types to these objects, indicates user to this
A little subject interests, the object that these operations are directed toward are most likely to be user and want the object accurately found.
In these cases, when retrieving alternative objects, the web object pointed with target type operation can be retrieved and had
There is the object of default similarity alternately object.For example, the commodity clicked according to user, the commodity that shopping cart is added or purchase
Commodity, to search the Recommendations of commodity similar with each commodity alternately.When searching, similar calculating can be used
Method come calculate with some object have similarity object.Wherein, similar calculation method can be, but not limited to, and cooperate with
Filtering method, similar calculation method based on content etc..
The retrieval source of second of type.Term is the retrieval content that user inputs in search engine, can also be anti-
The lookup for reflecting user is intended to.Therefore can be using term as a kind of retrieval source, the term once retrieved using user is into one
Step retrieval obtains alternative objects.
If the term that retrieval source includes input can determine inspection corresponding with term when searching alternative objects
Rope classification, and retrieve and belong to the object for retrieving classification alternately object.Specifically, the term of user's input may be
Object with detailed features can determine object in order to find more object similar with the object of user search
The classification belonged to is retrieved using object classification.May include multiple classification items in object classification, classification item may include but
It is not limited to category, brand, crowd, style etc..For example, the term of user's input includes " 2017 winters trendy frivolous natural feather
Take schoolgirl's money ", it is " female _ Bossden _ thin _ down jackets " according to the retrieval classification that the term is determined, thus according to the inspection
Rope classification retrieves alternative objects.
The retrieval source of third seed type.In some cases, the band of position where user can reflect user this thought
What the object to be searched is.For buying application scenarios on behalf, the band of position where user can reflect what user was serviced
The position of client, for example, village wash in a pan for consumer, for consumer and its client serviced usually in a village.In this case,
Behind the band of position where acquisition user, the band of position for the client that it is serviced can be determined, and then according to user search
Commodity and the band of position and the commodity between relationship, can determine that the object that user wants to look up is.
That is, if retrieval source includes the band of position where user, when searching alternative, can according to position
The preset retrieval classification in region is retrieved and belongs to the object for retrieving classification alternately object.
For example, the band of position for consumer is in south, the commodity of retrieval are down jackets, can determine what it was searched
Commodity may be the down jackets of thin money；If the band of position for consumer is in the north, the commodity of retrieval are still down jackets, can be with
Determine that the commodity that it is searched may be thick money down jackets.It for another example, is in Zhejiang for the band of position of consumer, the commodity of retrieval are
Tealeaves, due to rule of thumb knowing that one band of Zhejiang compares preference Dragon Well tea and white tea, then the commodity that can determine that it is wanted to look up can
It can be Dragon Well tea and white tea；If the band of position for consumer is in Fujian, the commodity of retrieval are tealeaves, known to rule of thumb
One band of Fujian compares preference Iron Guanyin, then can determine that the commodity that it is wanted to look up may be Iron Guanyin.
About above-mentioned several different types of retrieval sources, it should be noted that, can when using retrieval source retrieval alternative objects
Corresponding alternative objects are searched in multiple retrieval sources simultaneously.
For example, term can be operated with target type in conjunction with pointed object.Specifically, according to user's input
Term be it is multiple, after obtaining different retrieval classifications according to different terms, whether clicked according to user, plus purchase or purchase
The object of certain retrieval classes now is weighted processing to these retrieval classifications.For example, if user clicks plus purchases or buy certain
The object of a little retrieval classes now, then improve the weight of the retrieval classification, to improve the rank of these retrieval classifications.For weighting
Treated retrieves classification, selects the higher retrieval classification of level weights for searching alternative objects.It for another example, can be by term
In conjunction with the band of position.Related explanation and example may refer to above-mentioned third seed type retrieval source.
B3: the alternative objects that user's access is intended to will accurately be corresponded to and be determined as target object.
Wherein, alternative objects are searched according to retrieval source, can be found without the retrieval source of type different types of
Alternative objects.Due to different types of retrieval source, the levels of precision that reflected user is intended to is also different, therefore is searched
It is also different that the alternative objects reflection user arrived accesses the levels of precision being intended to.
In one example, the accuracy level being arranged in advance for the retrieval source of each type can be obtained；And according to accurate
The sequence of rank from high to low, the selection target object from alternative objects.
Specifically, different accuracy levels can be set for different types of retrieval source in advance, if retrieval source includes target
The band of position where object pointed by type operations, term and user, according to saying for the retrieval source to the three types
It is bright it is found that its be able to reflect user access be intended to levels of precision gradually decrease, therefore the three types retrieval source it is accurate
Rank is from high to low.
To which according to the height of the accuracy level in retrieval source, the alternative objects obtained to retrieval source are ranked up, from sequence
Selection target object in preceding alternative objects.The mode selected, which can be, selects the alternative objects of preset quantity as target pair
As.
In another example, attribute value of the alternative objects on objective attribute target attribute item can be obtained；And according to attribute value
Ordering rule, the selection target object from alternative objects.
Specifically, alternative objects can have attribute value on objective attribute target attribute item, as sales volume, price, conclusion of the business conversion ratio,
Seller's grade and the type of merchandise etc..Alternative objects can be ranked up, and the row of selection according to the ordering rule of some attribute value
Sequence is in preceding or posterior certain alternative objects as target object.For example, the rule compositor according to price from low to high, is then selected
The preceding alternative objects of sequence are selected as target object.
Be explained above user want it is accurate search object in the case where, how for user to recommend relevant object.With
The existing way of recommendation is compared, and used above is that the data that user generates in this access process are used as recommendation foundation, from
And the object recommended is more accurate.In addition, different types of retrieval source accuracy level is different in above method, and preferentially selection essence
The corresponding object in the true higher retrieval source of rank is as recommended, so as to further increase the precision of recommended.
It is intended in addition to accurately searching this access, target type user is also possible to be random to the access intention of web object
Browsing.It is intended to for this access, illustrates how that being intended to corresponding fuzzy search mode according to this access searches target object.
Specifically, if search pattern is fuzzy search, the implementation of target object is searched according to the search pattern,
It can specifically comprise the following steps C1~C3.
C1: the target retrieval classification gone out based on history access data statistics is obtained, wherein target retrieval classification includes following
In any one or more: answer season retrieval classification, fast-selling retrieval classification or new popular retrieval classification.
Wherein, in the case where arbitrarily browsing this access and being intended to, user may be such as popular pair in order to understand some information
As which the object of, Ying Ji or the object of fast sale be, in order to higher be that client chooses object.Therefore, in order to meet user
Demand in the case where this access is intended to can obtain some retrieval classifications, and retrieval classification, which needs to have, answers season characteristic, fast-selling characteristic
Or new popular characteristic.Retrieval classification is to access what data statistics went out according to history, and it may include numerous users that history, which accesses data,
History access data, that is to say, that the retrieval classification is come out from a large amount of user accesses data.
About answering season object.The object of some classes now has apparent seasonal rhythm, this class object is properly termed as answering
Season object, answer season object that can be recommended according to time dimension.The statistics of seasonal merchandise is every according to each classification object
A month sales volume distributional difference is measured.Particularly, can count over one year the sales volume mean value of various classification objects monthly and
Variance, the big classification object of variance can have obvious calendar variation, to be directed to these classification objects, can choose reality
Sales volume is higher than fast-selling season of the month of the certain threshold value of sales volume mean value as the classification object.
About popular objects.Term corresponding for each classification object, can carry out the extraction of keyword, and compare
The keyword difference of different times.When certain keywords were counting and the user for being not present or retrieving is fewer in the past, and
When the retrieval frequency of current point in time is higher, then it is assumed that the keyword has fashion trend, so as to have the keyword
Object be determined as popular objects.
C2: retrieval belongs to the object of target retrieval classification alternately object.
Wherein, retrieval classification is some large-scale object properties, such as one-piece dress.When retrieving alternative objects, retrieval belongs to
In the object of the retrieval classification, such as the one-piece dress retrieved may include fourreau, loose one-piece dress.
C3: selection meets the target object of preset condition from alternative objects.
Wherein, due to target retrieval classification may be it is a variety of, then can be according to certain ordering rule, to different retrieval classifications
Obtained alternative objects are ranked up, and select certain objects as target object after sequence.Sortord is referred to above-mentioned step
Mode in rapid B3, does not repeat herein.
It is explained above in the case where user wants arbitrarily browsing, how for user to recommend relevant object.Recommended
Object be the recommendation of coarseness, but can satisfy the demand that user understands current fashion trend, answers season object and fast-selling object.
The object of recommendation can help target type user to hold the general trend of object, and improving target type user is lead referral
Ability.
In conclusion as shown in figure 4, object recommendation method provided by the present application may include following several portions on the whole
Point.First part, disaggregated model building and user type identification, which can correspond to above-mentioned user identification method.Second
Point, however, it is determined that user type belongs to the user of target type, then determines that user is intended to the access of web object.If access is intended to
Accurately to search, then recommend fine-grained object, if access is intended to arbitrarily browse, recommends the object of coarseness.The part
Process, step B1~B3 and step C1~C3 shown in Fig. 3 can be corresponded to.
The structure of object recommendation device provided by the present application is illustrated below.See Fig. 5, it illustrates the application offers
A kind of object recommendation device structure, specifically include: search pattern determining module 501, target object searching module 502 and right
Image information sending module 503.
Search pattern determining module 501, if the user for accessing web object is target type user, it is determined that described
Search pattern of the user to the web object；
Target object searching module 502, for searching target object according to the search pattern；
Object information sending module 503, for the relevant information of the target object to be sent to the user.
It should be noted that the module in object recommendation device can be pushed away when executing concrete function according to the above object
The correlation method recommended in method executes, and does not repeat herein.
See Fig. 6, it illustrates another structure of object recommendation device provided by the present application, which is including Fig. 5 institute
It can also include user type determining module 504 and disaggregated model training module 505 on the basis of the apparatus structure shown.
User type determining module 504, if for being target type user in the user of access web object, it is determined that institute
Before user is stated to the search pattern of the web object, according to disaggregated model trained in advance, access web object is determined
Whether user is target type user.
Disaggregated model training module 505, for obtaining sample of users data；From the sample of users data, extracts and use
Family feature；Wherein the user characteristics can embody the characteristics of target type user；And the training user characteristics, to obtain
Identification model.
The structure of object recommendation equipment provided by the present application is illustrated below.See Fig. 7, it illustrates the application offers
A kind of object recommendation equipment structure, can specifically include: memory 701, processor 702, communication interface 703 and bus
704。
Memory 701, for storing program instruction and/or data.
Processor 702, by reading the instruction and/or data that store in the memory 701, for executing following behaviour
Make: if the user of access web object is target type user, it is determined that search pattern of the user to the web object；
And according to the search pattern, search target object.
Communication interface 703, for the relevant information of the target object to be sent to the user.
Bus 704, for each hardware component for accessing the monitoring device of data to be coupled.
In one example, the processor determines the user to the search pattern of the web object, comprising: processing
Device, specifically for obtaining the user to the access path of web object；And according to the access path, determine the user
To the search pattern of the web object.
In one example, the search pattern is accurate lookup or fuzzy search.
In one example, if the search pattern is accurate lookup, the processor is looked into according to the search pattern
Looking for target object includes: processor, the access generated during this accesses web object specifically for obtaining the user
Data；Different types of retrieval source is obtained from the access data, and different type is retrieved according to different types of retrieval source
Alternative objects；And it will accurately correspond to the alternative objects that user's access is intended to and be determined as target object.
In one example, the alternative objects that the processor will accurately correspond to that user's access is intended to are determined as mesh
Mark object, comprising: processor, specifically for obtaining in advance as the accuracy level of the retrieval source setting of each type；And according to
The sequence of accuracy level from high to low, the selection target object from the alternative objects.
In one example, the alternative objects that the processor will accurately correspond to that user's access is intended to are determined as mesh
Mark object, comprising: processor, specifically for obtaining attribute value of the alternative objects on objective attribute target attribute item；And according to the category
The ordering rule of property value, the selection target object from the alternative objects.
In one example, the retrieval source include it is following it is several in it is any a variety of: target type operation is pointed
The band of position where web object, term, user.
In one example, the processor retrieves different types of alternative objects, packet according to different types of retrieval source
It includes: processor, if including the pointed web object of target type operation, retrieval and the mesh specifically for the retrieval source
Mark object of the web object pointed by type operations with default similarity alternately object；If the retrieval source includes inspection
Rope word, it is determined that retrieval classification corresponding with the term, retrieval belong to the object for retrieving classification alternately object；
And if the retrieval source includes the band of position where user, basis and the preset retrieval classification in the band of position, inspection
Mermis is in the object alternately object of the retrieval classification.
In one example, if the search pattern is fuzzy search, the processor is looked into according to the search pattern
Look for target object, comprising: processor, specifically for obtaining the target retrieval classification gone out based on history access data statistics, wherein
Target retrieval classification includes any one or more in following: answering season retrieval classification, fast-selling retrieval classification or new popular retrieval
Classification；Retrieval belongs to the object of the target retrieval classification alternately object；And it selects to meet from the alternative objects
The target object of preset condition.
In one example, processor, if being also used in the user of access web object be target type user, it is determined that
Before the user is to the search pattern of the web object, according to disaggregated model trained in advance, access web object is determined
User whether be target type user.
In one example, processor is also used to train classification models；The processor train classification models, comprising: obtain
Obtain sample of users data；From the sample of users data, user characteristics are extracted；Wherein the user characteristics can embody target
The characteristics of type of user；And the training user characteristics, to obtain identification model.
It should be noted that all the embodiments in this specification are described in a progressive manner, each embodiment weight
Point explanation is the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.
It should also be noted that, herein, relational terms such as first and second and the like are used merely to one
Entity or operation are distinguished with another entity or operation, without necessarily requiring or implying between these entities or operation
There are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant are intended to contain
Lid non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including above-mentioned element.
The foregoing description of the disclosed embodiments makes professional and technical personnel in the field can be realized or use the application.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the application.Therefore, the application
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (21)
1. a kind of object recommendation method characterized by comprising
If the user for accessing web object is target type user, it is determined that lookup mould of the user to the web object
Formula；
According to the search pattern, target object is searched；
The relevant information of the target object is sent to the user.
2. object recommendation method according to claim 1, which is characterized in that the determination user is to the webpage pair
The search pattern of elephant, comprising:
The user is obtained to the access path of web object；
According to the access path, determine the user to the search pattern of the web object.
3. object recommendation method according to claim 2, which is characterized in that the search pattern is accurate lookup or fuzzy
It searches.
4. object recommendation method according to claim 3, which is characterized in that if the search pattern is accurate lookup,
It is described according to the search pattern, search target object, comprising:
Obtain the access data that the user generates during this accesses web object；
Different types of retrieval source is obtained from the access data, and different types of according to the retrieval of different types of retrieval source
Alternative objects；
The alternative objects that user's access is intended to will accurately be corresponded to and be determined as target object.
5. object recommendation method according to claim 4, which is characterized in that described accurately to correspond to user's access
The alternative objects of intention are determined as target object, comprising:
It obtains in advance as the accuracy level of the retrieval source setting of each type；
According to the sequence of accuracy level from high to low, the selection target object from the alternative objects.
6. object recommendation method according to claim 4, which is characterized in that described accurately to correspond to user's access
The alternative objects of intention are determined as target object, comprising:
Obtain attribute value of the alternative objects on objective attribute target attribute item；
According to the ordering rule of the attribute value, the selection target object from the alternative objects.
7. object recommendation method according to claim 4, which is characterized in that the retrieval source include it is following it is several in appoint
It anticipates a variety of: the band of position where the pointed web object of target type operation, term, user.
8. object recommendation method according to claim 7, which is characterized in that described to be retrieved according to different types of retrieval source
Different types of alternative objects, comprising:
If the retrieval source includes the pointed web object of target type operation, retrieval and target type operation are signified
To web object there is the object of default similarity alternately object；
If the retrieval source includes term, it is determined that retrieval classification corresponding with the term, retrieval belong to the retrieval
The object of classification alternately object；
If the retrieval source includes the band of position where user, basis and the preset retrieval classification in the band of position, inspection
Mermis is in the object alternately object of the retrieval classification.
9. object recommendation method according to claim 3, which is characterized in that if the search pattern is fuzzy search,
It is described according to the search pattern, search target object, comprising:
The target retrieval classification gone out based on history access data statistics is obtained, wherein target retrieval classification includes any in following
It is one or more: to answer season retrieval classification, fast-selling retrieval classification or new popular retrieval classification；
Retrieval belongs to the object of the target retrieval classification alternately object；
Selection meets the target object of preset condition from the alternative objects.
10. object recommendation method according to claim 1, which is characterized in that if the user in access web object is mesh
Mark type of user, it is determined that before the user is to the search pattern of the web object, further includes:
According to disaggregated model trained in advance, determine whether the user of access web object is target type user.
11. object recommendation method according to claim 10, which is characterized in that the training method packet of the disaggregated model
It includes:
Obtain sample of users data；
From the sample of users data, user characteristics are extracted；Wherein the user characteristics can embody target type user's
Feature；
The training user characteristics, to obtain identification model.
12. a kind of object recommendation equipment characterized by comprising
Processor, if the user for accessing web object is target type user, it is determined that the user is to the webpage pair
The search pattern of elephant；And according to the search pattern, search target object；
Communication interface, for the relevant information of the target object to be sent to the user.
13. object recommendation equipment according to claim 12, which is characterized in that the processor determines the user to institute
State the search pattern of web object, comprising:
Processor, specifically for obtaining the user to the access path of web object；And it according to the access path, determines
Search pattern of the user to the web object.
14. object recommendation equipment according to claim 12, which is characterized in that if the search pattern is accurate lookup,
Then according to the search pattern, search target object includes: the processor
Processor, the access data generated during this accesses web object specifically for obtaining the user；From described
Different types of retrieval source is obtained in access data, and different types of alternative objects are retrieved according to different types of retrieval source；
And it will accurately correspond to the alternative objects that user's access is intended to and be determined as target object.
15. object recommendation equipment according to claim 12, which is characterized in that described in the processor will be corresponded to accurately
User accesses the alternative objects being intended to and is determined as target object, comprising:
Processor, specifically for obtaining in advance as the accuracy level of the retrieval source setting of each type；And according to accuracy level
Sequence from high to low, the selection target object from the alternative objects.
16. object recommendation equipment according to claim 12, which is characterized in that described in the processor will be corresponded to accurately
User accesses the alternative objects being intended to and is determined as target object, comprising:
Processor, specifically for obtaining attribute value of the alternative objects on objective attribute target attribute item；And the row according to the attribute value
Sequence rule, the selection target object from the alternative objects.
17. object recommendation equipment according to claim 14, which is characterized in that the processor is according to different types of inspection
Suo Yuan retrieves different types of alternative objects, comprising:
Processor, if including the pointed web object of target type operation specifically for the retrieval source, retrieval with it is described
Target type operates object of the pointed web object with default similarity alternately object；If the retrieval source includes
Term, it is determined that retrieval classification corresponding with the term, the object that retrieval belongs to the retrieval classification are alternately right
As；And if the retrieval source includes the band of position where user, according to and the preset retrieval classification in the band of position,
Retrieval belongs to the object for retrieving classification alternately object.
18. object recommendation equipment according to claim 12, which is characterized in that if the search pattern is fuzzy search,
Then the processor searches target object according to the search pattern, comprising:
Processor accesses the target retrieval classification that data statistics goes out based on history specifically for obtaining, wherein target retrieval classification
Including any one or more in following: answering season retrieval classification, fast-selling retrieval classification or new popular retrieval classification；Retrieval belongs to
The object of the target retrieval classification alternately object；And selection meets the target of preset condition from the alternative objects
Object.
19. object recommendation equipment according to claim 12, which is characterized in that
Processor, if being also used in the user of access web object be target type user, it is determined that the user is to the net
Before the search pattern of page object, according to disaggregated model trained in advance, determine whether the user of access web object is target
Type of user.
20. object recommendation equipment according to claim 19, which is characterized in that
Processor is also used to train classification models；
The processor train classification models, comprising:
Obtain sample of users data；From the sample of users data, user characteristics are extracted；Wherein the user characteristics being capable of body
The characteristics of existing target type user；And the training user characteristics, to obtain identification model.
21. a kind of object recommendation device characterized by comprising
Search pattern determining module, if the user for accessing web object is target type user, it is determined that the user couple
The search pattern of the web object；
Target object searching module, for searching target object according to the search pattern；
Object information sending module, for the relevant information of the target object to be sent to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711104368.2A CN110020136B (en) | 2017-11-10 | 2017-11-10 | Object recommendation method and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711104368.2A CN110020136B (en) | 2017-11-10 | 2017-11-10 | Object recommendation method and related equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110020136A true CN110020136A (en) | 2019-07-16 |
| CN110020136B CN110020136B (en) | 2023-04-07 |
Family
ID=67186519
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711104368.2A Active CN110020136B (en) | 2017-11-10 | 2017-11-10 | Object recommendation method and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110020136B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116739840A (en) * | 2023-08-14 | 2023-09-12 | 贵州优特云科技有限公司 | Travel package recommendation method, device and storage medium based on multi-objective group optimization |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102411596A (en) * | 2010-09-21 | 2012-04-11 | 阿里巴巴集团控股有限公司 | Information recommendation method and system |
| US20120254310A1 (en) * | 2011-03-29 | 2012-10-04 | Sony Corporation | Content recommendation device, recommended content search method, and program |
| CN103838756A (en) * | 2012-11-23 | 2014-06-04 | 阿里巴巴集团控股有限公司 | Method and device for determining pushed information |
| CN105117418A (en) * | 2015-07-30 | 2015-12-02 | 百度在线网络技术（北京）有限公司 | Search based service information management system and method |
| CN106469382A (en) * | 2015-08-14 | 2017-03-01 | 阿里巴巴集团控股有限公司 | Idle merchandise items information processing method and device |
| US20170277778A1 (en) * | 2016-03-25 | 2017-09-28 | Maruthi Siva P Cherukuri | Personalized guidance and recommendation based on multi-variable user attributes and multi-dimensional schema |
-
2017
- 2017-11-10 CN CN201711104368.2A patent/CN110020136B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102411596A (en) * | 2010-09-21 | 2012-04-11 | 阿里巴巴集团控股有限公司 | Information recommendation method and system |
| US20120254310A1 (en) * | 2011-03-29 | 2012-10-04 | Sony Corporation | Content recommendation device, recommended content search method, and program |
| CN103838756A (en) * | 2012-11-23 | 2014-06-04 | 阿里巴巴集团控股有限公司 | Method and device for determining pushed information |
| CN105117418A (en) * | 2015-07-30 | 2015-12-02 | 百度在线网络技术（北京）有限公司 | Search based service information management system and method |
| CN106469382A (en) * | 2015-08-14 | 2017-03-01 | 阿里巴巴集团控股有限公司 | Idle merchandise items information processing method and device |
| US20170277778A1 (en) * | 2016-03-25 | 2017-09-28 | Maruthi Siva P Cherukuri | Personalized guidance and recommendation based on multi-variable user attributes and multi-dimensional schema |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116739840A (en) * | 2023-08-14 | 2023-09-12 | 贵州优特云科技有限公司 | Travel package recommendation method, device and storage medium based on multi-objective group optimization |
| CN116739840B (en) * | 2023-08-14 | 2023-11-14 | 贵州优特云科技有限公司 | Travel package recommendation method, device and storage medium based on multi-objective group optimization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110020136B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109559208B (en) | Information recommendation method, server and computer readable medium | |
| CN108121737B (en) | Method, device and system for generating business object attribute identifier | |
| Su et al. | A method for discovering clusters of e-commerce interest patterns using click-stream data | |
| US10769702B2 (en) | Recommendations based upon explicit user similarity | |
| US8762226B2 (en) | Item discovery tools and methods for shopping in an electronic commerce environment | |
| US10423999B1 (en) | Performing personalized category-based product sorting | |
| CN111784455A (en) | Article recommendation method and recommendation equipment | |
| CN108960945A (en) | Method of Commodity Recommendation and device | |
| US20110282821A1 (en) | Further Improvements in Recommendation Systems | |
| CN108153791B (en) | Resource recommendation method and related device | |
| CN111523976A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
| US20130006713A1 (en) | Method for aggregating pricing information and assigning a fair market value to goods sold in a peer-to-peer e-commerce transaction | |
| US20190355041A1 (en) | Cognitive fashion product recommendation system and method | |
| CN113792176A (en) | Image evaluation | |
| CN103345695A (en) | Commodity recommendation method and device | |
| JP2012529122A (en) | System and method for learning user genre and style and matching products to user preferences | |
| WO2007117980A2 (en) | Method and system for computerized searching and matching using emotional preference | |
| CN106062743A (en) | Systems and methods for keyword suggestion | |
| CN105138690B (en) | The method and apparatus for determining keyword | |
| CN103886001A (en) | Personalized commodity recommendation system | |
| CN108228579A (en) | Network interaction system | |
| CN106326318B (en) | Searching method and device | |
| WO2012009180A1 (en) | Value maximizing recommendation systems | |
| CN111899047A (en) | Resource recommendation method and device, computer equipment and computer-readable storage medium | |
| US20170186065A1 (en) | System and Method of Product Selection for Promotional Display |
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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40010857 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |