CN109903127A - A kind of group recommending method, device, storage medium and server - Google Patents
A kind of group recommending method, device, storage medium and server Download PDFInfo
- Publication number
- CN109903127A CN109903127A CN201910113590.1A CN201910113590A CN109903127A CN 109903127 A CN109903127 A CN 109903127A CN 201910113590 A CN201910113590 A CN 201910113590A CN 109903127 A CN109903127 A CN 109903127A
- Authority
- CN
- China
- Prior art keywords
- group
- label
- user
- data
- recommended
- 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
Abstract
The invention discloses a kind of group recommending method, device, storage medium and servers, which comprises data collection steps obtain business datum, behavioral data and third party's data；Label engineering step is respectively that user and article label according to business datum, behavioral data and third party's data, and the label includes behavior label, attribute tags and third party's label；Group's recommendation step is recommended according to the label of each user in the label of article and group to be recommended to group to be recommended.The present invention generates label by introducing third party's data, pass through the hobby of each user in tag expression group, suitable label is selected according to the label of user each in group, according to the corresponding article of the label recommendations of the user of selection, it realizes that the group for taking into account each user preference in group is recommended, solves the problems, such as that current group recommending method is difficult to carry out group's recommendation in cold start-up.
Description
Technical field
The present invention relates to Internet technical field, in particular to a kind of group recommending method, device, storage medium and clothes
Business device.
Background technique
Recommend it can be found that user's content of interest and long-tail content, are widely used, such as Amazon in commercial field
Books recommendation, the commercial product recommending of Taobao etc..The essence of recommendation is precisely to excavate the potential and required individual demand of user,
But in actual application scenarios, the case where group formed to multiple users is recommended can be also faced, as restaurant menus is recommended, electricity
Shadow is recommended, tourist famous-city is recommended etc..Recommendation towards group, ultimate challenge are how to take into account each user in group
Preference, to provide optimal recommendation for group.
Current group recommending method can adopt the explicit preferences (such as to the scoring of article) or implicit inclined of schooling group internal user
Good (such as system action) is then that group generates recommendation results by proposed algorithm.It is however recommended to algorithm will in view of preference and
Predict the fusion of scoring, such as everyone prediction is scored averagely in group, and this requires users behavioral data, stays in system
Article is evaluated after lower operation note, such as purchase article, ware is such as repeatedly clicked and brings shopping cart into, if
When user has just registered or article is just added, without any behavior record, that is, the case where cold start-up, proposed algorithm be just difficult into
Row group is recommended.In this regard, needing a kind of method for carrying out group's recommendation in cold start-up in actual use.
Summary of the invention
The present invention provides a kind of group recommending method, device, storage medium and servers, it is intended to solve current group
Recommended method is difficult to the problem of carrying out group's recommendation in cold start-up.
To achieve the above object, the present invention provides a kind of group recommending methods, comprising:
Data collection steps obtain business datum, behavioral data and third party's data；
Label engineering step is respectively that user and article label according to business datum, behavioral data and third party's data,
The label includes behavior label, attribute tags and third party's label；
Group's recommendation step, according to the label of each user in the label of article and group to be recommended to group to be recommended
Recommended.
Compared with prior art, a kind of group recommending method disclosed by the invention, difference capturing service data, behavioral data
It labels according to the collected data to user and article with third party's data, for the user in same group, according in group
The label of each user selects suitable label, and according to the corresponding article of the label recommendations of the user of selection, realization takes into account group
The group of interior each user preference is recommended.The present invention labels to user and article, is not rely on single data, even if row
Do not have for data, can also be labelled according to business datum and third party's data to user, then according to the case where label to
Family is recommended, and not simple dependent behavior data recommend user in group, will not be in default of behavioral data, cold
It is difficult to be recommended in the case of starting, solves the problems, such as to carry out group's recommendation in cold start-up.This kind of recommendation side, group
Method introduces the reference of third party's data, is labelled according to business datum, behavioral data and third party's data to user, the mark
Label include behavior label, attribute tags and third party's label, can be entirely square by behavior label, attribute tags and third party's label
Position various dimensions portray user and article, and when carrying out user's recommendation, the label for analyzing each user in group is selected properly
Label, the preference of each user can be looked after, be can choose out according to suitable label and meet each user happiness in group
Good article, and recommended.The present invention generates label by introducing third party's data, passes through each use in tag expression group
The hobby at family selects suitable label according to the label of user each in group, corresponding according to the label recommendations of the user of selection
Article, realize take into account each user preference in group group recommend, solve current group recommending method and be cold-started
In the case of be difficult to the problem of carrying out group's recommendation.
Further, the business datum includes demographics basic information data, article basic information data and affairs
Record data；The demographics basic information data and article basic information data determine the acquisition theme of third party's data.
In a preferred embodiment of the invention, the themes of third party's data by demographics basic information data and
Article basic information data determines, since network data is various, it is obviously impossible to obtain whole network datas, so need to select
Suitable third party's data are selected to obtain.In this regard, passing through the demographics basic information number in business datum in the present embodiment
According to the acquisition theme for determining third party's data with article basic information data, third party's data are selectively obtained, solve this
It is various for network data in scheme to be difficult to the problem of effectively obtaining.
Further, third party's data divide the acquisition of theme by web crawlers and save the open letter on internet
Cease data.
In a preferred embodiment of the invention, third party's data divide the acquisition and preservation of theme by web crawlers
Public information data on internet grab most emerging information needed on internet since web crawlers has as soon as possible,
The advantages of discovery page as early as possible is repeated and is eliminated, so divide the acquisition of theme by web crawlers in the present embodiment and save mutual
Public information data in networking are convenient for subsequent label engineering, solve asking for the acquisition process of network data in this programme
Topic.
Further, the label engineering step, comprising:
It is respectively that user and article generate behavior label according to transaction journal data and behavioral data；
Attribute tags are generated for user according to demographics basic information data；
It is that article generates attribute tags according to article basic information data；
It is respectively that user and article generate third party's label according to third party's data.
In a preferred embodiment of the invention, give a kind of specific label and generate scheme, due to according to data into
Row mark is signed with the way of many maturations, or statistics or cluster, but introduces a variety of data in face of this programme, needs to generate more
Kind label, the difficulty that label can be brought to generate provide a kind of specific label generation scheme to this present embodiment, have cleared three classes
The relationship of data output label generates different labels according to different data, solves the problems, such as that the label in this programme generates.
Further, group's recommendation step, comprising:
When there are common labels in the label of user each in group to be recommended, and there are corresponding for the common label
Article is then recommended according to the sequence of common label to group to be recommended.
In a preferred embodiment of the invention, a kind of recommendation side, group specifically based on third party's label is given
Method, by selecting common tag present in the label of each user in group to be recommended (common label) as suitable mark
Label select corresponding article according to this suitable label, recommend the user in group, look after the inclined of each user
It is good, it is ensured that the article of recommendation is suitble to each user in group, is completed by label to user preference each in group
Treatment, and since article and user all have label, it is determined that it is same can to correspond to searching for the label of suitable user
The label of article then recommends the article of the label, it is ensured that recommendation meets user preference.This embodiment gives a kind of tools
The group recommending method based on third party's label of body is solved and how to be carried out in case of a cold start by third party's label
The problem of meeting group's recommendation of each user preference in the masses.
Further, group's recommendation step, further includes:
When in the label of user each in group to be recommended be not present common label or the common label there is no pair
The article answered then finds maximum similar group according to the label of the user of group to be recommended, instructs using maximum similar group as sample
Practice output model, is recommended using the output model that training obtains to group to be recommended.
In a preferred embodiment of the invention, give when there is no altogether in the label of user each in group to be recommended
Group recommending method when corresponding article is not present in same label or the common label, solves in previous embodiment and needs
There is the problem of common tag (common label) and corresponding article could be recommended, that there is no common tags is (common
Label) or when corresponding article, select group similar with group to be recommended as sample, by sample training output model, then
It is gone to predict the article that group to be measured needs with the output model that training is completed, group to be measured be recommended, solving is not having
In the case of common tag (common label) or corresponding article the problem of carrying out group's recommendation by label.
Further, in the group of the similar group of the maximum each user be user in group to be recommended maximum phase
Like user, each user in group to be recommended is the maximum similar users of each user in the group of maximum similar group,
The maximum similar users are greater than threshold value with the ratio of user's label having the same by compared with.
In a preferred embodiment of the invention, a kind of side for selecting group similar with group to be measured sample is given
Method selects maximum similar group by maximum similar users, it is ensured that the similarity of group to be measured and sample group ensures
Training effect.Solves the problems, such as samples selection when trained output model.
The present invention also provides a kind of group's recommendation apparatus, comprising:
Data acquisition unit, for obtaining business datum, behavioral data and third party's data；
Label engineering unit, for being respectively that user and article are beaten according to business datum, behavioral data and third party's data
Label, the label include behavior label, attribute tags and third party's label；
Group's recommendation unit, the label for each user in the label and group to be recommended according to article is to be recommended
Group is recommended.
The present invention also provides a kind of computer readable storage medium, the computer readable storage medium includes the meter of storage
Calculation machine program；Wherein, the equipment where the computer program controls the computer readable storage medium at runtime executes
Group recommending method as described in any one of the above embodiments.
The present invention also provides a kind of server recommended for group, including processor, memory and it is stored in described
In memory and it is configured as the computer program executed by the processor, the processor is executing the computer program
Shi Shixian group recommending method as described in any one of the above embodiments.
Compared with prior art, a kind of group recommending method, device, storage medium and server disclosed by the invention, point
Other capturing service data, behavioral data and third party's data according to the collected data label to user and article, for same
User in group selects suitable label according to the label of user each in group, according to the label recommendations of the user of selection
Corresponding article realizes that the group for taking into account each user preference in group is recommended.The present invention labels to user and article, not
Dependent on single data, even if behavioral data does not have, can also be labelled according to business datum and third party's data to user,
Then user is recommended according to the case where label, not simple dependent behavior data recommend user in group, no
It can be difficult to be recommended in cold start-up in default of behavioral data, solve the progress group in cold start-up and push away
The problem of recommending.This kind of group recommending method introduces the reference of third party's data, according to business datum, behavioral data and third party
Data label to user, and the label includes behavior label, attribute tags and third party's label, pass through behavior label, attribute
Label and third party's label can portray user and article with comprehensive various dimensions, when carrying out user's recommendation, analyze in group
The label of each user selects suitable label, can look after the preference of each user, can be selected according to suitable label
The article for meeting each user preferences in group is selected out, and is recommended.The present invention generates label by introducing third party's data,
By the hobby of each user in tag expression group, suitable label is selected according to the label of user each in group, according to
The corresponding article of the label recommendations of the user of selection realizes that the group for taking into account each user preference in group is recommended, solves mesh
Preceding group recommending method is difficult to the problem of carrying out group's recommendation in cold start-up.
Detailed description of the invention
Fig. 1 is a kind of flow chart of one embodiment of group recommending method of the present invention；
Fig. 2 is a kind of flow chart of group's recommendation step one embodiment of group recommending method of the present invention；
Fig. 3 is a kind of structural block diagram of one embodiment of group's recommendation apparatus of the present invention；
Fig. 4 is a kind of one embodiment structural block diagram for the server recommended for group of the present invention.
Specific embodiment
As shown in Figure 1, a kind of group recommending method of the present invention, comprising: data collection steps, acquisition business datum,
Behavioral data and third party's data；Label engineering step is respectively user according to business datum, behavioral data and third party's data
It labels with article, the label includes behavior label, attribute tags and third party's label；Group's recommendation step, according to article
Label and group to be recommended in the label of each user recommend to group to be recommended.
Compared with prior art, a kind of group recommending method disclosed by the invention, difference capturing service data, behavioral data
It labels according to the collected data to user and article with third party's data, for the user in same group, according in group
The label of each user selects suitable label, and according to the corresponding article of the label recommendations of the user of selection, realization takes into account group
The group of interior each user preference is recommended.The present invention labels to user and article, is not rely on single data, even if row
Do not have for data, can also be labelled according to business datum and third party's data to user, then according to the case where label to
Family is recommended, and not simple dependent behavior data recommend user in group, will not be in default of behavioral data, cold
It is difficult to be recommended in the case of starting, solves the problems, such as to carry out group's recommendation in cold start-up.This kind of recommendation side, group
Method introduces the reference of third party's data, is labelled according to business datum, behavioral data and third party's data to user, the mark
Label include behavior label, attribute tags and third party's label, can be entirely square by behavior label, attribute tags and third party's label
Position various dimensions portray user and article, and when carrying out user's recommendation, the label for analyzing each user in group is selected properly
Label, the preference of each user can be looked after, be can choose out according to suitable label and meet each user happiness in group
Good article, and recommended.The present invention generates label by introducing third party's data, passes through each use in tag expression group
The hobby at family selects suitable label according to the label of user each in group, corresponding according to the label recommendations of the user of selection
Article, realize take into account each user preference in group group recommend, solve current group recommending method and be cold-started
In the case of be difficult to the problem of carrying out group's recommendation.
In a specific embodiment of the invention, when courseware is recommended to teacher by CVTE Xi Wo institute:
1) behavioral data that acquisition teacher leaves on uncommon fertile college websites, and the business datum of record itself, according to
Third party's data that the region attribute of teacher's registration goes crawler to be the theme with area acquire local religion such as from the Educational website of various regions
Educate policy；
2) behavior label, attribute tags, third party's label are stamped for teacher, the teacher if registration is the area A constructs one
A " people teaches version chemistry " label, this label have gone out policy requirements derived from the area A and have taught version Chemistry Teaching Material using people；
3) when group is recommended, based on " people teaches version chemistry ", this third party's label thinks that the area A group teacher recommends about people
The courseware that religion version Chemistry Teaching Material is prepared lessons.
This solves greatly Xi Wo institute and does not know how to recommend most suitable courseware to teacher after new teacher's registration very much, allows
Teacher starts to get started.
Summarize the main process of this programme: acquisition behavioral data and business datum, according to the demographics in business datum
Essential information and article essential information go crawler to acquire third party's data；Behavior-based control data, business datum, third party's data point
Not Gou Jian user and article behavior label, attribute tags, third party's label；Finally group is carried out based on third party's label to push away
It recommends, alleviates cold start-up problem.
Further, the business datum includes demographics basic information data, article basic information data and affairs
Record data；The demographics basic information data and article basic information data determine the acquisition theme of third party's data.
In a preferred embodiment of the invention, the themes of third party's data by demographics basic information data and
Article basic information data determines, since network data is various, it is obviously impossible to obtain whole network datas, so need to select
Suitable third party's data are selected to obtain.In this regard, passing through the demographics basic information number in business datum in the present embodiment
According to the acquisition theme for determining third party's data with article basic information data, third party's data are selectively obtained, solve this
It is various for network data in scheme to be difficult to the problem of effectively obtaining.
Further, third party's data divide the acquisition of theme by web crawlers and save the open letter on internet
Cease data.
In a preferred embodiment of the invention, third party's data divide the acquisition and preservation of theme by web crawlers
Public information data on internet grab most emerging information needed on internet since web crawlers has as soon as possible,
The advantages of discovery page as early as possible is repeated and is eliminated, so divide the acquisition of theme by web crawlers in the present embodiment and save mutual
Public information data in networking are convenient for subsequent label engineering, solve asking for the acquisition process of network data in this programme
Topic.
In one specific embodiment of data collection steps of the present invention,
Business datum: about information and transaction journal of the user in system, such as user's gender, cell-phone number, area
When demographics basic information, such as user put what article transaction journal bought；
Behavioral data: about operation note of the user in system, when clicking what article, page browsing such as user
Long, menu clicks sequence etc.；
Third party's data: the data source introduced to solve cold start-up, suitable for answering for third party's data can be collected
With scene, and the data that third party to be acquired will depend on the demand of label engineering.
Behavioral data is that user in system has long period of operation that could generate.Business datum is relative quiescent, includes population
Statistical basis information (or article basic information) and two class of transaction journal.If demographic information has registration when user's registration
Can use, if but transaction journal user there is no affairs to there will not be data；This comprehensive analysis, determines, can solve
The data being certainly cold-started are exactly demographics basic information and third party's data in business datum.
To solve the problems, such as cold start-up that group is recommended, emphasis will utilize the demographics basic information and the in business datum
Tripartite's data.Third party's data divide Focused crawler and preservation, are used for subsequent label engineering.The acquisition crawler of third party's data is certainly
Disclosed information on internet, such as the publication of statistical yearbook, white paper, various reports, policy.It is crawled under theme and theme
Data, depending on demographics essential information and article basic information.Here it to be stressed, either theme is still
Label is all group's concept, is all comprising multiple users below the same theme or the same label.
Further, the label engineering step, comprising: according to transaction journal data and behavioral data be respectively user and
Article generates behavior label；Attribute tags are generated for user according to demographics basic information data；According to article basic information
Data are that article generates attribute tags；It is respectively that user and article generate third party's label according to third party's data.
In a preferred embodiment of the invention, give a kind of specific label and generate scheme, due to according to data into
Row mark is signed with the way of many maturations, or statistics or cluster, but introduces a variety of data in face of this programme, needs to generate more
Kind label, the difficulty that label can be brought to generate provide a kind of specific label generation scheme to this present embodiment, have cleared three classes
The relationship of data output label generates different labels according to different data, solves the problems, such as that the label in this programme generates.
In one specific embodiment of label engineering step of the present invention,
Label engineering includes the label building of two objects, user and article.For labelling with user, pass through behavior number
According to, business datum (demographics basic information and transaction journal), third party's data, it is each that each angle of user's panorama can be constructed
The preference of aspect.Third party's label will inherently alleviate cold start-up and excavate, be behavior label without in the case where, can be with
Secondary attribute label is used for subsequent recommendation.Therefore, the third party's data for generating third party's label are according to demographic information
Or the attribute of article essential information acquires.
The generation of user and article tag have three classes:
1) behavior label is extracted by behavioral data and Transaction Information；
2) attribute tags are extracted by demographic information and article essential information；
3) theme according to set by demographic information or article essential information goes to crawl third party's data for extracting
Third party's label.
If it is new user or new article, then there is no behavior label.Because new user and new article only have attribute tags, this
Cold start-up problem can be brought, therefore this programme introduces third party's data, third party's label of building secondary attribute label is for delaying
Cold start-up problem is solved, also improves the accuracy rate that group is recommended to a certain extent.
Third party's label constructs example:
System user Xiao Ming is new registration, and the age is 20 years old in the demographic information of registration, by being based on age master
The third party's data crawled are inscribed, all kinds of labels for belonging to 20 years old this group can be stamped for Xiao Ming, reading preference such as books is
Kungfu novel.
Illustrate again, a label is a community concept, and it is exactly a group that multiple users, which belong to the same label,.
Further, group's recommendation step, comprising:
When there are common labels in the label of user each in group to be recommended, and there are corresponding for the common label
Article is then recommended according to the sequence of common label to group to be recommended.
In a preferred embodiment of the invention, a kind of recommendation side, group specifically based on third party's label is given
Method, by selecting common tag present in the label of each user in group to be recommended (common label) as suitable mark
Label select corresponding article according to this suitable label, recommend the user in group, look after the inclined of each user
It is good, it is ensured that the article of recommendation is suitble to each user in group, is completed by label to user preference each in group
Treatment, and since article and user all have label, it is determined that it is same can to correspond to searching for the label of suitable user
The label of article then recommends the article of the label, it is ensured that recommendation meets user preference.This embodiment gives a kind of tools
The group recommending method based on third party's label of body is solved and how to be carried out in case of a cold start by third party's label
The problem of meeting group's recommendation of each user preference in the masses.
One specific example of the present embodiment is, it is assumed that group G1: user A (label 1,2,3), B (label 2,3,4), C
(label 2,5,6)；
1) common tag (common label) of user in group is counted, G1 has common tag (common label) 2,3；
2) there is common tag (common label) for G1, sort according to number of users, common tag (common mark
Label) order is: 2 (3 people), 3 (2 people)；
3) judge the corresponding article of common tag (common label), recommend according to tag sorting.
If the article of corresponding label 2 has the article of I1, corresponding label 3 to have I2, then the article sequence of a G1 group is recommended to be
I1、I2。
Further, group's recommendation step, further includes:
When in the label of user each in group to be recommended be not present common label or the common label there is no pair
The article answered then finds maximum similar group according to the label of the user of group to be recommended, instructs using maximum similar group as sample
Practice output model, is recommended using the output model that training obtains to group to be recommended.
In a preferred embodiment of the invention, give when there is no altogether in the label of user each in group to be recommended
Group recommending method when corresponding article is not present in same label or the common label, solves in previous embodiment and needs
There is the problem of common tag (common label) and corresponding article could be recommended, that there is no common tags is (common
Label) or when corresponding article, select group similar with group to be recommended as sample, by sample training output model, then
It is gone to predict the article that group to be measured needs with the output model that training is completed, group to be measured be recommended, solving is not having
In the case of common tag (common label) or corresponding article the problem of carrying out group's recommendation by label.
Further, in the group of the similar group of the maximum each user be user in group to be recommended maximum phase
Like user, each user in group to be recommended is the maximum similar users of each user in the group of maximum similar group,
The maximum similar users are greater than threshold value with the ratio of user's label having the same by compared with.
In a preferred embodiment of the invention, a kind of side for selecting group similar with group to be measured sample is given
Method selects maximum similar group by maximum similar users, it is ensured that the similarity of group to be measured and sample group ensures
Training effect.Solves the problems, such as samples selection when trained output model.It is of the present invention maximum similar users be with by than
Have the ratio of same label greater than the user of threshold value compared with user, the threshold value is preset relatively threshold value, guarantees the mark of user
Label are covered by most of.In the present embodiment, selecting threshold value is 60 percent, when there are user's first have label A, B, C,
The maximum similar users of user's first to be selected require the user with two labels in label A, B, C, so, Yong Huyi
It is the maximum similar users of user's first with label A, B, D.Since each user is wait push away in the group of maximum similar group
The maximum similar users of user in group are recommended, each user in group to be recommended is each in the group of maximum similar group
The maximum similar users of user indicate that user's first, second in group one to be recommended, third, fourth all have maximum in group two
Similar users, and user in group two, penta, oneself, heptan in group one also all with maximum similar users, so group two
It is the similar group of maximum of group one.
One specific example of the present embodiment is, it is assumed that group G2: user D (label 6,7,8), E (label 9,10,11),
1) D and E does not have common tag (common label) in G2；
2) maximum similar group's sample is found for G2, maximum is similar two o'clock standard, and first is each user in group
Label to cover major part, be two at least 6,7,8 three labels as the maximum of D is similar；Second is maximum similarity group
The covered user of group is greater than the user in G2, such as at least to there is the use comprising D, E maximum similar (meeting first standard)
Family；
Meet if group G3:F (label 6,7,12), G (label 10,11,12), and group G4:F (label 6,7,12),
H (label 10,12,13) is not just met, because H is with E without maximum similar, that is to say, that do not include D and E maximum phase in G2 in G4
As user, only comprising the maximum similar user of D；
3) there is maximum similar group's sample, so that it may the more disaggregated models of supervision be had according to sample training, then according to mould
Type predicts the scoring for the article recommended for G2, this is the recommended method of standard.
In one specific embodiment of group's recommendation step of the present invention, as shown in Figure 2, comprising the following steps:
S1, judges whether group user has common tag (common label), if it is entrance, then step S2, if not
It is then to enter step S4；
S2, common tag (common label) sort according to number of users, judge whether common tag (common label) has
Corresponding article, if it is, S3 is entered step, if it is not, then entering step S4；
Corresponding article is recommended group according to tag sorting, terminates recommendation step by S3；
S4 finds maximum similar group, using maximum similar group as sample training output model, is obtained using training defeated
Model is recommended to group to be recommended out, terminates recommendation step.
As shown in figure 3, the present invention also provides a kind of group's recommendation apparatus, comprising: data acquisition unit, for obtaining business
Data, behavioral data and third party's data；Label engineering unit, for according to business datum, behavioral data and third party's data
Respectively user and article labels, and the label includes behavior label, attribute tags and third party's label；Group recommends single
Member, the label for each user in the label and group to be recommended according to article are recommended to group to be recommended.
Further, the business datum includes demographics basic information data, article basic information data and affairs
Record data；The demographics basic information data and article basic information data determine the acquisition theme of third party's data.
Further, third party's data divide the acquisition of theme by web crawlers and save the open letter on internet
Cease data.
Further, the label engineering unit, comprising: behavior tag generation module, according to transaction journal data and row
It is respectively that user and article generate behavior label for data；User property tag generation module, according to demographics basic information
Attribute tags are generated for user in data；Goods attribute tag generation module is that article generates category according to article basic information data
Property label；Third party's tag generation module is respectively that user and article generate third party's label according to third party's data.
Further, group's recommendation unit, comprising: first group's recommending module, for when each in group to be recommended
There are common labels in the label of a user, and there are corresponding articles for the common label, then according to common label
Sequence is recommended to group to be recommended.
Further, group's recommendation unit, further includes: second group's recommending module, for when in group to be recommended
There is no common labels or the common label, and corresponding article is not present in the label of each user, then according to group to be recommended
The label of the user of group finds maximum similar group, using maximum similar group as sample training output model, is obtained using training
Output model recommended to group to be recommended.
Further, in the group of the similar group of the maximum each user be user in group to be recommended maximum phase
Like user, each user in group to be recommended is the maximum similar users of each user in the group of maximum similar group,
The maximum similar users are greater than threshold value with the ratio of user's label having the same by compared with.
A kind of embodiment of group's recommendation apparatus of the present invention, can be realized recommendation side, group described in any of the above-described embodiment
All processes of method, the technical effect of effect and the realization of modules, unit in device respectively with above-described embodiment institute
The technical effect of effect and the realization of the group recommending method stated corresponds to identical, and which is not described herein again.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer readable storage medium includes
The computer program of storage；Wherein, where the computer program controls the computer readable storage medium at runtime
Equipment executes group described in any of the above-described embodiment and recommends recognition methods.
It is shown in Figure 4 the embodiment of the invention also provides a kind of server, it is a kind of server provided by the invention
The structural block diagram of one preferred embodiment, the terminal device include processor 10, memory 20 and are stored in the storage
In device 20 and it is configured as the computer program executed by the processor 10, the processor 10 is executing the computer journey
Group recommending method described in any of the above-described embodiment is realized when sequence.
Preferably, the computer program can be divided into one or more module/units (such as computer program 1,
Computer program 2), one or more of module/units are stored in the memory 20, and
It is executed by the processor 10, to complete the present invention.One or more of module/units, which can be, can complete specific function
Series of computation machine program instruction section, the instruction segment is for describing execution of the computer program in the terminal device
Process.
The processor 10 can be central processing unit (Central Processing Unit, CPU), can also be
Other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc., general processor can be microprocessor or the processor 10 is also possible to any conventional place
Device is managed, the processor 10 is the control centre of the terminal device, utilizes terminal device described in various interfaces and connection
Various pieces.
The memory 20 mainly includes program storage area and data storage area, wherein program storage area can store operation
Application program needed for system, at least one function etc., data storage area can store related data etc..In addition, the memory
20 can be high-speed random access memory, can also be nonvolatile memory, such as plug-in type hard disk, intelligent memory card
(Smart Media Card, SMC), secure digital (Secure Digital, SD) card and flash card (Flash Card) etc.,
Or the memory 20 is also possible to other volatile solid-state parts.
It should be noted that above-mentioned terminal device may include, but it is not limited only to, processor, memory, those skilled in the art
Member does not constitute the restriction to terminal device it is appreciated that Fig. 4 structural block diagram is only the example of above-mentioned terminal device, can be with
Including perhaps combining certain components or different components than illustrating more or fewer components.
A kind of group recommending method, device, storage medium and server disclosed by the invention, difference capturing service data,
Behavioral data and third party's data according to the collected data label to user and article, for the user in same group, root
Suitable label is selected according to the label of user each in group, according to the corresponding article of the label recommendations of the user of selection, is realized
The group for taking into account each user preference in group is recommended.The present invention labels to user and article, is not rely on single number
According to can also be labelled according to business datum and third party's data to user, then according to label even if behavioral data does not have
Situation recommends user, and not simple dependent behavior data recommend user in group, will not be in default of behavior
Data are difficult to be recommended in cold start-up, solve the problems, such as to carry out group's recommendation in cold start-up.The population
Group recommended method introduces the reference of third party's data, according to business datum, behavioral data and third party's data to user's mark
Label, the label include behavior label, attribute tags and third party's label, are marked by behavior label, attribute tags and third party
Label can portray user and article with comprehensive various dimensions, when carrying out user's recommendation, analyze the label of each user in group
Suitable label is selected, the preference of each user can be looked after, be can choose out and met in group according to suitable label
The article of each user preferences, and recommended.The present invention generates label by introducing third party's data, passes through tag expression group
The hobby of each user in group selects suitable label according to the label of user each in group, according to the mark of the user of selection
Label recommend corresponding article, realize that the group for taking into account each user preference in group is recommended, solve current recommendation side, group
Method is difficult to the problem of carrying out group's recommendation in cold start-up.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (10)
1. a kind of group recommending method characterized by comprising
Data collection steps obtain business datum, behavioral data and third party's data；
Label engineering step is respectively that user and article label according to business datum, behavioral data and third party's data, described
Label includes behavior label, attribute tags and third party's label；
Group's recommendation step is carried out according to the label of each user in the label of article and group to be recommended to group to be recommended
Recommend.
2. a kind of group recommending method according to claim 1, which is characterized in that the business datum includes demographics
Basic information data, article basic information data and transaction journal data；The demographics basic information data and article base
The acquisition theme of plinth information data decision third party's data.
3. a kind of group recommending method according to claim 2, which is characterized in that third party's data are climbed by network
Worm divides the acquisition of theme and saves the public information data on internet.
4. a kind of group recommending method according to claim 2, which is characterized in that the label engineering step, comprising:
It is respectively that user and article generate behavior label according to transaction journal data and behavioral data；
Attribute tags are generated for user according to demographics basic information data；
It is that article generates attribute tags according to article basic information data；
It is respectively that user and article generate third party's label according to third party's data.
5. a kind of group recommending method according to claim 1, which is characterized in that group's recommendation step, comprising:
When there are common labels in the label of user each in group to be recommended, and there are corresponding objects for the common label
Product are then recommended according to the sequence of common label to group to be recommended.
6. a kind of group recommending method according to claim 5, which is characterized in that group's recommendation step, further includes:
When common label or the common label are not present in the label of user each in group to be recommended, there is no corresponding
Article then finds maximum similar group according to the label of the user of group to be recommended, defeated as sample training using maximum similar group
Model out is recommended using the output model that training obtains to group to be recommended.
7. a kind of group recommending method according to claim 6, which is characterized in that in the group of the similar group of the maximum
Each user is the maximum similar users of user in group to be recommended, and each user in group to be recommended is maximum similar
The maximum similar users of each user in the group of group, maximum similar users and user's label having the same by compared with
Ratio be greater than threshold value.
8. a kind of group's recommendation apparatus characterized by comprising
Data acquisition unit, for obtaining business datum, behavioral data and third party's data；
Label engineering unit, for being respectively that user and article label according to business datum, behavioral data and third party's data,
The label includes behavior label, attribute tags and third party's label；
Group's recommendation unit, the label for each user in the label and group to be recommended according to article is to group to be recommended
Recommended.
9. a kind of computer readable storage medium, which is characterized in that the computer readable storage medium includes the calculating of storage
Machine program；Wherein, the equipment where the computer program controls the computer readable storage medium at runtime executes such as
The described in any item group recommending methods of claim 1-7.
10. a kind of server recommended for group, which is characterized in that including processor, memory and be stored in described deposit
In reservoir and it is configured as the computer program executed by the processor, the processor is when executing the computer program
Realize such as the described in any item group recommending methods of claim 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910113590.1A CN109903127A (en) | 2019-02-14 | 2019-02-14 | A kind of group recommending method, device, storage medium and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910113590.1A CN109903127A (en) | 2019-02-14 | 2019-02-14 | A kind of group recommending method, device, storage medium and server |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109903127A true CN109903127A (en) | 2019-06-18 |
Family
ID=66944874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910113590.1A Pending CN109903127A (en) | 2019-02-14 | 2019-02-14 | A kind of group recommending method, device, storage medium and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109903127A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717097A (en) * | 2019-09-06 | 2020-01-21 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN111176505A (en) * | 2019-12-20 | 2020-05-19 | 上海淇玥信息技术有限公司 | Page display method and device based on third-party task linkage and electronic equipment |
| CN111914188A (en) * | 2020-08-19 | 2020-11-10 | 英华达（上海）科技有限公司 | Method, system, device and storage medium for selecting recommendation target user |
| CN112016003A (en) * | 2020-08-19 | 2020-12-01 | 重庆邮电大学 | Social crowd user tag mining and similar user recommending method based on CNN |
| CN112395261A (en) * | 2019-08-16 | 2021-02-23 | 中国移动通信集团浙江有限公司 | Service recommendation method and device, computing equipment and computer storage medium |
| CN112507241A (en) * | 2020-12-31 | 2021-03-16 | 北京精准沟通传媒科技股份有限公司 | Vehicle type recommendation method, system, medium and electronic device for obtaining vehicle type recommendation list |
| CN112907295A (en) * | 2021-03-19 | 2021-06-04 | 恩亿科(北京)数据科技有限公司 | Similar population expansion method and device based on computing advertisement background |
| CN113408482A (en) * | 2021-07-13 | 2021-09-17 | 杭州联吉技术有限公司 | Training sample generation method and device |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130346234A1 (en) * | 2012-06-26 | 2013-12-26 | Medio Systems, Inc. | Recommendations system |
| CN103731738A (en) * | 2014-01-23 | 2014-04-16 | 哈尔滨理工大学 | Video recommendation method and device based on user group behavioral analysis |
| CN104090888A (en) * | 2013-12-10 | 2014-10-08 | 深圳市腾讯计算机系统有限公司 | Method and device for analyzing user behavior data |
| WO2015062623A1 (en) * | 2013-10-28 | 2015-05-07 | Telefonaktiebolaget L M Ericsson (Publ) | Group recommendation for selection of service items |
| CN104750856A (en) * | 2015-04-16 | 2015-07-01 | 天天艾米（北京）网络科技有限公司 | System and method for multi-dimensional synergic recommendation |
| CN105260390A (en) * | 2015-09-11 | 2016-01-20 | 合肥工业大学 | Group-oriented project recommendation method based on joint probability matrix decomposition |
| CN105637555A (en) * | 2013-10-15 | 2016-06-01 | 谷歌公司 | Social where next suggestion |
| CN105718579A (en) * | 2016-01-22 | 2016-06-29 | 浙江大学 | Information push method based on internet-surfing log mining and user activity recognition |
| US20170004562A1 (en) * | 2013-02-14 | 2017-01-05 | Wine Ring, Inc. | Recommendation system based on group profiles of personal taste |
| CN107862532A (en) * | 2016-09-22 | 2018-03-30 | 腾讯科技（深圳）有限公司 | A kind of user characteristics extracting method and relevant apparatus |
| CN109284443A (en) * | 2018-11-28 | 2019-01-29 | 四川亨通网智科技有限公司 | A kind of tourism recommended method and system based on crawler technology |
-
2019
- 2019-02-14 CN CN201910113590.1A patent/CN109903127A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130346234A1 (en) * | 2012-06-26 | 2013-12-26 | Medio Systems, Inc. | Recommendations system |
| US20170004562A1 (en) * | 2013-02-14 | 2017-01-05 | Wine Ring, Inc. | Recommendation system based on group profiles of personal taste |
| CN105637555A (en) * | 2013-10-15 | 2016-06-01 | 谷歌公司 | Social where next suggestion |
| WO2015062623A1 (en) * | 2013-10-28 | 2015-05-07 | Telefonaktiebolaget L M Ericsson (Publ) | Group recommendation for selection of service items |
| CN104090888A (en) * | 2013-12-10 | 2014-10-08 | 深圳市腾讯计算机系统有限公司 | Method and device for analyzing user behavior data |
| CN103731738A (en) * | 2014-01-23 | 2014-04-16 | 哈尔滨理工大学 | Video recommendation method and device based on user group behavioral analysis |
| CN104750856A (en) * | 2015-04-16 | 2015-07-01 | 天天艾米（北京）网络科技有限公司 | System and method for multi-dimensional synergic recommendation |
| CN105260390A (en) * | 2015-09-11 | 2016-01-20 | 合肥工业大学 | Group-oriented project recommendation method based on joint probability matrix decomposition |
| CN105718579A (en) * | 2016-01-22 | 2016-06-29 | 浙江大学 | Information push method based on internet-surfing log mining and user activity recognition |
| CN107862532A (en) * | 2016-09-22 | 2018-03-30 | 腾讯科技（深圳）有限公司 | A kind of user characteristics extracting method and relevant apparatus |
| CN109284443A (en) * | 2018-11-28 | 2019-01-29 | 四川亨通网智科技有限公司 | A kind of tourism recommended method and system based on crawler technology |
Non-Patent Citations (1)
| Title |
|---|
| 吴云昌等: "群组推荐分析与研究综述", 《电信科学》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112395261A (en) * | 2019-08-16 | 2021-02-23 | 中国移动通信集团浙江有限公司 | Service recommendation method and device, computing equipment and computer storage medium |
| CN110717097A (en) * | 2019-09-06 | 2020-01-21 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN111176505A (en) * | 2019-12-20 | 2020-05-19 | 上海淇玥信息技术有限公司 | Page display method and device based on third-party task linkage and electronic equipment |
| CN111176505B (en) * | 2019-12-20 | 2021-04-13 | 上海淇玥信息技术有限公司 | Page display method and device based on third-party task linkage and electronic equipment |
| CN111914188A (en) * | 2020-08-19 | 2020-11-10 | 英华达（上海）科技有限公司 | Method, system, device and storage medium for selecting recommendation target user |
| CN112016003A (en) * | 2020-08-19 | 2020-12-01 | 重庆邮电大学 | Social crowd user tag mining and similar user recommending method based on CNN |
| CN112507241A (en) * | 2020-12-31 | 2021-03-16 | 北京精准沟通传媒科技股份有限公司 | Vehicle type recommendation method, system, medium and electronic device for obtaining vehicle type recommendation list |
| CN112907295A (en) * | 2021-03-19 | 2021-06-04 | 恩亿科(北京)数据科技有限公司 | Similar population expansion method and device based on computing advertisement background |
| CN113408482A (en) * | 2021-07-13 | 2021-09-17 | 杭州联吉技术有限公司 | Training sample generation method and device |
| CN113408482B (en) * | 2021-07-13 | 2023-10-10 | 杭州联吉技术有限公司 | Training sample generation method and generation device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109903127A (en) | A kind of group recommending method, device, storage medium and server | |
| TWI408560B (en) | A method, system and apparatus for recommending items or people of potential interest to users in a computer-based network | |
| CN109934619A (en) | User's portrait tag modeling method, apparatus, electronic equipment and readable storage medium storing program for executing | |
| Chen et al. | Mining students' learning patterns and performance in Web-based instruction: a cognitive style approach | |
| CN111898031B (en) | Method and device for obtaining user portrait | |
| EP2008199A2 (en) | System and method of segmenting and tagging entities based on profile matching using a multi-media survey | |
| CN106326441A (en) | Information recommendation method and device | |
| CN108959323B (en) | Video classification method and device | |
| CN111177559B (en) | Text travel service recommendation method and device, electronic equipment and storage medium | |
| TWI705411B (en) | Method and device for identifying users with social business characteristics | |
| CN112100221B (en) | Information recommendation method and device, recommendation server and storage medium | |
| CN114936301B (en) | Intelligent household building material data management method, device, equipment and storage medium | |
| CN113806630A (en) | Attention-based multi-view feature fusion cross-domain recommendation method and device | |
| CN113139141A (en) | User label extension labeling method, device, equipment and storage medium | |
| CN111191133B (en) | Service search processing method, device and equipment | |
| CN116823410B (en) | Data processing method, object processing method, recommending method and computing device | |
| CN110069686A (en) | User behavior analysis method, apparatus, computer installation and storage medium | |
| CN109829033A (en) | Method for exhibiting data and terminal device | |
| CN112446214A (en) | Method, device and equipment for generating advertisement keywords and storage medium | |
| WO2020135420A1 (en) | Method and apparatus for classifying users | |
| US7716209B1 (en) | Automated advertisement publisher identification and selection | |
| JP3547338B2 (en) | Information retrieval method and device | |
| Lee et al. | Qtag: tagging as a means of rating, opinion-expressing, sharing and visualizing | |
| CN113327132A (en) | Multimedia recommendation method, device, equipment and storage medium | |
| CN113704623A (en) | Data recommendation method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190618 |
|
| RJ01 | Rejection of invention patent application after publication |