CN106570031A - Service object recommending method and device - Google Patents
Service object recommending method and device Download PDFInfo
- Publication number
- CN106570031A CN106570031A CN201510657556.2A CN201510657556A CN106570031A CN 106570031 A CN106570031 A CN 106570031A CN 201510657556 A CN201510657556 A CN 201510657556A CN 106570031 A CN106570031 A CN 106570031A
- Authority
- CN
- China
- Prior art keywords
- business object
- user profile
- similarity
- matrix
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The embodiment of the invention provides a service object recommending method and device. The method comprises the following steps: acquiring final relevant data between user information and service objects; calculating a similarity among the service objects by the final relevant data; and recommending the service objects according to the similarity among the service objects, wherein the final relevant data between the user information and the service objects are generated by the following steps: acquiring interactive behavior data between the user information and the service objects, wherein the service objects are provided with corresponding service feature data; generating initial relevant data between the user information and the service objects according to the interactive behavior data; partitioning the service objects into service object clusters by the service feature data; and generating the final relevant data between the user information and the service objects according to the service object clusters and the initial relevant data. Through adoption of the service object recommending method and device, the other problems of inaccurate recommendation and the like caused by spare data of interactive behaviors are solved.
Description
Technical field
The application is related to technical field of data processing, more particularly to a kind of recommendation side of business object
Method and a kind of recommendation apparatus of business object.
Background technology
In recent years, with the fast development of ecommerce, electric business advertisement also grows up simultaneously, by
In electric business advertisement be the main business revenue means of electric business platform, therefore the either advertisement form of electric business advertisement
Or charging mode is all presented booming situation.
Electric business advertisement has advertisement form miscellaneous, includes display advertisement and search advertisements, exhibition
Show that advertisement is also presented increasingly consequence in the proportion of electric business advertisement.For search advertisements,
Search scene user is often input into search inquiry query, that is, actively expresses wish, then exhibition
Show the advertisement related to inquiry query.But for the display advertisement of non-search, search non-
Rope scene user does not express and is intended to, it is therefore desirable to speculate the demand of user at that time, it is so-called to realize
" thinking that you are thought ".Another non-search scene is entered after user clicks certain commodity
The details page of commodity, shows and the click commodity or advertisement and its related or phase in details page
As advertisement.
At present, display advertisement or in non-search scene, personalized recommendation technology is a kind of conventional
And effective method.Wherein, in personalized recommendation technology based on user historical behavior (as click on,
Purchase etc.) collaborative filtering method prove on each big electric business platform one it is fairly simple and
Effective personalized recommendation method.But it has been found that built by the historical behavior of user
(element in matrix is user to the click of the commodity or purchase feelings to the user-commodity matrix for coming
Condition), the matrix is often a very sparse matrix, and this also results in the behaviour on the matrix
Either all there is serious deficiency on accuracy or in coverage rate in the recommendation results that obtain of work, because
This can not realize so-called " thinking that you are thought ", be that user shows and meets the wide of the demand of user at that time
Accuse.
The content of the invention
In view of the above problems, it is proposed that the embodiment of the present application so as to provide one kind overcome the problems referred to above or
A kind of a kind of recommendation method and business object of business object that person solves the above problems at least in part
Recommendation apparatus.
In order to solve the above problems, this application discloses a kind of recommendation method of business object, including:
Obtain the final related data between user profile and business object;
Using the similarity between the final correlation data calculation business object;
Recommend business object according to the similarity between the business object;
Wherein,
Final related data between the user profile and business object is generated in the following way:
Obtain the interbehavior data between user profile and business object;The business object has
Corresponding service feature data;
According to the interbehavior data, generate initial between the user profile and business object
Related data;
The business object is divided into into business object cluster using the service feature data;
According to the business object cluster and the initial related data, the user profile and industry are generated
Final related data between business object.
Preferably, it is described according to interbehavior data, generate the user profile and business object it
Between initial related data the step of include:
Create matrix;The matrix is made up of the user profile and business object, in the matrix
Including matrix element;
Judge to whether there is interbehavior data between the user profile and business object;
If so, then in the matrix by user profile matrix element corresponding with business object
It is filled to the first record identification;
If it is not, then by user profile matrix element corresponding with business object in the matrix
It is filled to the second record identification.
Preferably, the business object is divided into business object cluster by the employing service feature data
Step includes:
Obtain the corresponding service feature data of the business object;
Determine the core set of words of the business object using the service feature data;
Content similarity between business object is calculated using the core set of words;
Obtain the business object that content similarity reaches preset content similarity threshold;
The business object that the content similarity reaches preset content similarity threshold is combined as into business
Object cluster.
Preferably, the foundation business object cluster and the initial related data, generate the user
The step of final related data between information and business object, includes:
Inquiry is the matrix element of the first record identification in the matrix;The matrix element has
Corresponding user profile and business object;
Judge the business object in the business object cluster whether with it is described be the first record identification square
The corresponding business object of array element element is identical;
If so, the corresponding user profile of matrix element for the first record identification is then obtained;
In the matrix by with the user information correlation, and with the business object cluster in
Business object identical matrix element is filled to the first record identification;Matrix after the filling is rich
Rich matrix.
Preferably, the step of similarity using between final correlation data calculation business object
Including:
Determine the corresponding matrix element of any two business object in the abundant matrix;
Count the number that the corresponding matrix element of described two business objects is the first record identification;
Determine the contribution weight of each user profile;
Using the contribution weight and the corresponding matrix of described two business objects of each user profile
Element is the number of the first record identification, calculates the similarity between any two business object.
Preferably, the step of contribution weight of each user profile of determination includes:
Obtain the number that the corresponding matrix element of each user profile is the first record identification;
Adopt the corresponding matrix element of described each user profile for the number of the first record identification, obtain
Obtain the contribution weight of each user profile.
Preferably, methods described also includes:
Using the similarity between the Similarity Measure user profile between the business object;
Recommend business object according to the similarity between the user profile.
Preferably, it is similar between the Similarity Measure user profile between the employing business object
The step of spending includes:
For the business object that there are interbehavior data between any user information, by the industry
Similarity between business object organizes the user profile correspondence in the business object of default similarity threshold
User preferences set;
Calculate similar between any user profile using the user preferences set according to preset rules
Degree.
Preferably, the business object be commodity, the interbehavior data include click data,
Feedback data, transaction data and/or collection data, the service feature data of the business object data
Title and details content for commodity.
The embodiment of the present application also discloses a kind of recommendation apparatus of business object, including:
Final related data acquisition module, it is final between user profile and business object for obtaining
Related data;
Business object similarity calculation module, for using the final correlation data calculation business pair
Similarity as between;
First business object recommending module, for recommending according to the similarity between the business object
Business object;
Wherein, described device also includes:
Interbehavior data acquisition module, for obtaining interacting between user profile and business object
Behavioral data;The business object has corresponding service feature data;
Initial correlation data generation module, for according to the interbehavior data, generating the use
Initial related data between family information and business object;
Business object cluster is divided into module, for using the service feature data by the business object
It is divided into business object cluster;
Final related data generation module, for initial related to described according to the business object cluster
Data, generate the final related data between the user profile and business object.
Preferably, the initial related data generation module includes:
Matrix creates submodule, for creating matrix;The matrix is by the user profile and business
Object is constituted, and the matrix includes matrix element;
Interbehavior judging submodule, for whether judging between the user profile and business object
There are interbehavior data;If so, the first filling submodule is then called, if it is not, then calling second
Filling submodule;
First filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the first record identification;
Second filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the second record identification.
Preferably, institute's business object cluster is divided into module includes:
Service feature data acquisition submodule, for obtaining the corresponding service feature of the business object
Data;
Core set of words determination sub-module, for determining the business using the service feature data
The core set of words of object;
Content similarity calculating sub module, for using the core set of words calculate business object it
Between content similarity;
Business object acquisition submodule, for obtaining content similarity preset content similarity threshold is reached
The business object of value;
Business object cluster combines submodule, similar for the content similarity to be reached into preset content
The business object of degree threshold value is combined as business object cluster.
Preferably, the final related data generation module includes:
Matrix element inquires about submodule, for inquiring about in the matrix as the square of the first record identification
Array element element;The matrix element has corresponding user profile and business object;
Business object judging submodule, for judging the business object cluster in business object whether
With it is described be the first record identification the corresponding business object of matrix element it is identical;If so, then call
User profile acquisition submodule;
User profile acquisition submodule, for obtaining the matrix element pair for the first record identification
The user profile answered;
3rd filling submodule, in the matrix by with the user information correlation, and
The first record identification is filled to the business object identical matrix element in the business object cluster;
Matrix horn of plenty matrix after the filling.
Preferably, the business object similarity calculation module includes:
Matrix element determination sub-module, for determining the abundant matrix in any two business object
Corresponding matrix element;
First record identification number statistic submodule, it is corresponding for counting described two business objects
Matrix element is the number of the first record identification;
Contribution weight determination sub-module, for determining the contribution weight of each user profile;
Similarity Measure submodule between business object, for using each user profile
Contribution weight and the number that the corresponding matrix element of described two business objects is the first record identification,
Calculate the similarity between any two business object.
Preferably, the contribution weight determination sub-module includes:
First record identification number statistic unit, for obtaining the corresponding matrix element of each user profile
Element is the number of the first record identification;
Contribution weight obtaining unit, for adopt the corresponding matrix element of described each user profile for
The number of the first record identification, obtains the contribution weight of each user profile.
Preferably, also include:
User profile similarity calculation module, based on using the similarity between the business object
Calculate the similarity between user profile;
Second business object recommending module, for recommending according to the similarity between the user profile
Business object.
Preferably, the user preferences set obtains submodule and includes:
User preferences set obtains submodule, for interacting for existing between any user information
The business object of behavioral data, by the similarity between the business object in default similarity threshold
Business object organize the corresponding user preferences set of the user profile;
Similarity Measure submodule between user profile, for adopting the use according to preset rules
Family hobby set calculates the similarity between any user profile.
The embodiment of the present application includes advantages below:
In the embodiment of the present application, in the application scenarios for recommending business object, usual user profile pair
Answering is sparse with business object interbehavior data, therefore the initial related data for obtaining
It is often sparse, this also result in using initial related data business object recommendation results not
Accurately and the not enough problem of overlay capacity, the business of combined content collaborative filtering therefore in the embodiment of the present application
Object cluster so as to obtain more abundant final related data, is used in combination filling initial related data
In the application scenarios for recommending business object, solve that interbehavior Sparse brings is a series of
Problem, for example, operation is carried out on final related data after filling will obtain preferably recommendation effect
Really, the accuracy rate of recommendation is improved, so-called " thinking that you are thought " can be realized, be that user shows symbol
The business object of the demand of user when being fated, improves Consumer's Experience effect.
Description of the drawings
The step of Fig. 1 is a kind of recommendation embodiment of the method for business object of the application flow chart;
Fig. 2 is a kind of collaborative filtering based on Adc and the collaborative filtering based on Item of the application
The schematic diagram of difference;
Fig. 3 be the application a kind of collaborative filtering based on Adc in data matrix processing procedure show
It is intended to;
Fig. 4 is that a kind of of the application realizes the master that advertisement is recommended based on collaboration in combination with information filtering
Want flow chart;
Fig. 5 is that the another kind of the application realizes that advertisement is recommended based on collaboration in combination with information filtering
Broad flow diagram;
Fig. 6 is a kind of structured flowchart of the recommendation apparatus embodiment of business object of the application.
Specific embodiment
It is understandable to enable the above-mentioned purpose of the application, feature and advantage to become apparent from, with reference to
The drawings and specific embodiments are described in further detail to the application.
In recent years ecommerce is developed rapidly at home, occurs in that advertisement form miscellaneous,
Search advertisements can specifically be included also display advertisement.For display advertisement, how effectively
Supply and be accurately positioned the demand of active user, become the key for determining display advertisement effect.
Industry proves the orientation display advertisement by the interbehavior data based on user, can reach
The lifting (CTR and RPM aspects) of 6-8 times of business effect.Therefore how to recommend its potential for user
The advertisement of demand, meets the personalized recommendation that user was intended at that time, is that industry is wanted to solve always
Problem.Specifically, above-mentioned CTR and RPM are referred respectively to:
CTR (click-through rate, clicking rate), that is, an advertisement or commodity it is clicked
The number of times that number of times is demonstrated divided by this advertisement or commodity, be generally used to evaluation advertisement represents effect
Really;At identical conditions, clicking rate is higher, and the effect of showing advertisement is better.
RPM (Revenue Per Thousand Ad impressions, the receipts that thousand showing advertisements bring
Benefit):The situation of Profit of general advertisement platform, the index weighed as core using RPM, in stream
The fixed situation of amount, the income of RPM more high platforms is higher;RPM=1000*CTR*PPC, its
Middle PPC (Pay Per Click, pay-per-click advertisement) is each income clicked on.
At present, the conventional personalized recommendation method of some industries is based on collaborative filtering
(Collaborative Filtering, abbreviation CF) technology.Collaborative filtering be by analyze user interest,
Similar (interest) user of specified user is found in customer group, comprehensive these similar users are to certain
The evaluation of one information, forms system and the specified user is predicted the fancy grade of this information.Collaboration
Filtration can specifically include following several method, be described below middle Item and represent advertisement or business
Product, User represents user:
(1) a most frequently used class method is to be based on the collaborative filtering method of Item, that is, is passed through
The interbehavior data of User and Item obtaining the similarity between Item, if central principle is exactly
User clicks on or has interacted Item A and Item B simultaneously, then the phase between Item A and Item B
A ticket has been thrown like degree, so the phase between Item just can have finally been determined by substantial amounts of interbehavior data
Like degree.
(2) an other class is that, based on the collaborative filtering method of User, central principle just assumes that User
A is similar User to User B, then the interactive Item of User B can be directly as User A
Recommendation Item;And determine the similarity degree of User A and User B often interacting using User
Item is vectorial, that is, calculate the cosine angle of both Item vectors, intuitively says to be exactly both interactions
Common Item more both are more similar.
(3) additionally, an also class method is exactly the Item according to User interactions, its Item is obtained
Title (theme) or details in information obtain the preference word of User to represent User, and
The table of falling row chain of word-Item is set up in rear end, then generates the preference word of User according to the table of falling row chain on line,
Preference word recalls the mode of Item to represent.
However, above-mentioned a few class personalized recommendation methods all have the shortcomings that it is certain, specifically:
(1) collaborative filtering method of the first kind based on Item, depends on the interaction of substantial amounts of User
Behavioral data (the such as behavior such as click, transaction).And the interbehavior of User is often found in putting into practice
Data are extremely precious, for single User, Item lists of its interaction be often it is a small amount of,
This is also the sparse sex chromosome mosaicism of User interbehaviors of generally existing in personalized recommendation.This problem is past
Toward directly resulting in the inaccurate (arbitrarily accidental to click on the recommendation that causes jointly several times and tie of recommendation results
Fruit is often inaccurate) and overlay capacity not enough (a large amount of Item are because being total to by User and other Item
With clicking on etc., behavior is very few leads to not release similar Item) problem.
(2) the another kind of collaborative filtering method based on User, is recommended by the hobby of User
Its similar User is realizing, but often relatively more various (the such as User A of the hobby of User
Like " mobile phone " and " football " etc. simultaneously), this interactive Item that also directly results in User is
Comparison is general and various, and this also results in the incorrectness based on the collaborative filtering recommending result of User.
Such as User A judge similar (clicking on " mobile phone " because of them a large amount of identical with User B
Item), but User B do not like " football ", User A hobby " football ", by this
Mode, " football " will recommend User B, and this also directly results in the dislike of User B.
(3) and for preference word is extracted, the method for recalling Item by preference word is often present not
The problem of accuracy, predominantly:Firstth, stating Item as several words inherently has many letters
Breath lose problem, second, User is expressed as preference word information equally exist information loss ask
Topic;Lacking for this two pieces of information, often leads to recommendation results not fully up to expectations.
To overcome above weak point, the embodiment of the present application propose it is a kind of it is new based on collaboration with it is interior
Hold and filter the Generalization bounds for combining, Item is obtained in table mainly by way of information filtering
State content or promote Item set extremely similar in wish and constitute Adc (Ad cluster, advertisement
Cluster), and the similarity between Item and Item is obtained using collaborative filtering method on Adc, from
And the series of problems that the interbehavior Sparse for solving user brings.
With reference to Fig. 1, show the application a kind of business object recommendation embodiment of the method the step of flow
Cheng Tu, specifically may include steps of:
Step 101, obtains the final related data between user profile and business object;
It should be noted that the business object in the embodiment of the present application can include different business field
Concrete things, such as the corresponding advertisement of commodity or commodity etc..To make those skilled in the art more
The embodiment of the present application is understood well, in this manual, mainly using commodity and advertisement as business
A kind of example of object is illustrated.
Advertisement in the embodiment of the present application can be shown by one or more websites or platform
Money or many money Commdity advertisements, can include one or more commodity letters in the information of the Commdity advertisement
Breath, such as item property, such as commodity image, trade name, commodity price, descriptive labelling,
Parameter of marque or commodity etc..
In implementing, user profile is referred to for identifying the corresponding ID of user, or
User account etc. can navigate to the mark of a certain user.In the embodiment of the present application, user profile
Final related data between business object is referred to, based on by cooperateing with combination with information filtering
Obtain, the associated data of incidence relation between user profile and business object can be accurately reflected.
Typically can be represented using matrix or list.
It is final related between a kind of user profile and business object of the application with reference to shown in Fig. 2
The step of data genaration flow chart, in a preferred embodiment of the present application, the user profile
Final related data between business object can be generated in the following way:
Step S1, obtains the interbehavior data between user profile and business object;The business
Object has corresponding service feature data;
In implementing, the interbehavior that user is carried out on the net can produce corresponding interbehavior
Data.For example, for clicking on, feeding back, conclude the business, collecting these interbehaviors, can accordingly produce
Click data, feedback data, transaction data, collection data these interbehavior data.
Step S2, according to the interbehavior data, generate the user profile and business object it
Between initial related data;
In a preferred embodiment of the present application, step S2 can include following sub-step:
Sub-step S11, creates matrix;The matrix is made up of the user profile and business object,
The matrix includes matrix element;
Sub-step S12, judges to whether there is interbehavior number between the user profile and business object
According to;If so, sub-step S13 is then performed;If it is not, then performing sub-step S14;
Sub-step S13, by user profile matrix element corresponding with business object in the matrix
Element is filled to the first record identification;
Sub-step S14, by user profile matrix element corresponding with business object in the matrix
Element is filled to the second record identification.
In the embodiment of the present application, the history interbehavior data of user and business object are primarily based on
User-Item matrixes are built, as initial related data.Wherein, User represents user, Item tables
Show advertisement.
Assume that the first record identification is yes, the second record identification is no, it is assumed that the history interaction of user
There are N number of User, M Item in behavioral data, if User A and Item_1 had and interact
Corresponding matrix element is set to into yes, will be corresponding if Item_1 not interaction if User A
Matrix element is set to no.Certainly, other marks such as 1,0 can also be adopted in practical application to represent use
Family interacts with whether advertisement had, or can also be recorded using list, and the application is implemented
Example need not be any limitation as to this.
It should be noted that due to the history interbehavior data of user now it is less, therefore currently
The User-Item matrixes of structure, i.e., initial related data, typically one very sparse matrix,
If therefore now recommended using this sparse matrix, it is impossible to obtain preferable recommendation effect.
Therefore in the embodiment of the present application, final related data can further be obtained based on initial related data and be used
In recommendation.
Step S3, business object cluster is divided into using the service feature data by the business object;
In a preferred embodiment of the present application, step S3 can include following sub-step:
Sub-step S21, obtains the corresponding service feature data of the business object;
Sub-step S22, using the service feature data core word set of the business object is determined
Close;
Sub-step S23, using the core set of words content similarity between business object is calculated;
Sub-step S24, obtains the business object that content similarity reaches preset content similarity threshold;
Sub-step S25, by the content similarity business object of preset content similarity threshold is reached
It is combined as business object cluster.
In the embodiment of the present application mainly by way of information filtering (Content Filtering), according to
According to the information of business object itself, that is, the corresponding service feature data of business object are calculating industry
Content similarity between business object, the higher conduct of the content similarity between business object is same
Individual business object cluster.
By taking advertisement (Ad) as an example, Cempetency-based education is combined into the higher Ad of content similarity extensively
Cluster (Ad Cluster, abbreviation Adc) is accused, specifically, the main content by utilize Ad itself,
The title and details content of such as Ad, is intuitively assumed to be user and likes identical with oneself hobby content
Advertisement, the advertisement with identical content implication is got together by information filtering.Detailed process
For:
(1) title in all of Ad and details content are carried out into participle, and is determined wherein important
Word as core word, so each Ad can just be expressed as a term vector;
(2) any 2 Ad content similarities in terms of content are calculated by equation 1 below:
Wherein, the oi in formula (1), oj represent respectively i-th Ad and j-th Ad, w table
Show word, W (oi) is the core set of words in i-th Ad, wherein numeral 1 can immobilize,
Can change according to Ad is corresponding to the interaction times of user, the embodiment of the present application is not limited this
System.
Obtained after the content similarity between Ad using formula (1), so that it may according to the content phase of setting
Like degree threshold value, the Ad that content similarity exceedes the content similarity threshold value is carried out as same cluster
Process.
It should be noted that when the embodiment of the present application is implemented, above-mentioned content similarity threshold value can
To be set according to the actual requirements, the embodiment of the present application is not any limitation as to this.
Step S4, according to the business object cluster and the initial related data, generates the user
Final related data between information and business object.
In a preferred embodiment of the present application, step S4 can include following sub-step:
Sub-step S31, inquires about in the matrix as the matrix element of the first record identification;The square
Array element element has corresponding user profile and business object;
Sub-step S32, judge the business object in the business object cluster whether with it is described be the first note
The corresponding business object of matrix element of record mark is identical;If so, sub-step S33 is then performed;
Sub-step S33, obtains the corresponding user profile of matrix element for the first record identification;
Sub-step S34, in the matrix by with the user information correlation, and with the business
Business object identical matrix element in object cluster is filled to the first record identification;After the filling
Matrix horn of plenty matrix.
Continue by taking Ad as an example, in this example Ad is referred to as Item.Through information filtering
Afterwards, the content similarity between each Ad or each Item can be acquired, then basis
The content similarity threshold value of setting, using Ad of all the elements similarity in the range of certain as same
Individual Adc.
A kind of collaborative filtering based on Adc of the application with reference to shown in Fig. 2 and the association based on Item
With the schematic diagram of the difference filtered, by taking User shown in Fig. 2 and Item as an example, it can be seen that:User
Direct interaction is Item A and Item K, if the simple collaborative filtering with based on Item, can only
The similarity of ItemA and ItemK is obtained by the interbehavior of User;Implement relative to the application
For the collaborative filtering based on Adc of example, Item A are with Item a, b, c in content of text information
It is upper to there is high content similarity and belong to same Adc A, it is same for Item K, Item
K, i, j also have high content similarity in content of text information with Item K and belong to same
Individual Adc K, if both combine the interactive Item for just enriching User, therefore the collaboration on Adc
Filtration not only can obtain the similarity of Item A and Item K, it is also possible to obtain Adc A or Adc
Similarity in K between any Item, can also obtain between Adc A and Adc K between any Item
Similarity, such as Item a and Item j.
Data matrix is processed in a kind of collaborative filtering based on Adc of the application with reference to shown in Fig. 3
The schematic diagram of process, it is main in this example to illustrate using the process of Adc filled matrix elements.Fig. 3
The User-Item matrixes on the left side are the matrix for directly consisting of the Item of User history interbehaviors,
Assume that history interbehavior data have N number of User, M Item, if User is A and Item_1
Then matrix element is set to yes interaction, is otherwise no;Usually one very sparse matrix.
Assume to obtain the higher Item_1 of similarity and Item_2, Item_3 composition by information filtering
Adc, and similarity relation between the Item in this Adc is filled in User-Item matrixes,
Can see and obtain this figure of Fig. 4 the right, it is obvious that the matrix is far richer, already becomes
One abundant matrix.
It should be noted that obtaining abundant after information filtering is filled in the embodiment of the present application
User-Item matrixes after, use based on the collaborative filtering of Adc to obtain the phase between Item
Seemingly spend for recommending;Adopt SVD after filling on User-Item matrixes that can certainly be
(Singular Value Decomposition, singular value decomposition) etc. operates to obtain the phase between Item
Seemingly spend and for recommending, the embodiment of the present application is not any limitation as to this.
It is abundant using what is obtained after processing in the embodiment of the present application in the recommendation scene of actual advertisement
User-Item matrixes can obtain following benefit:
1st, in the application scenarios of display advertisement, because the interbehavior data of user are often
Sparse, so as to this common problem that the inaccurate and overlay capacity of caused recommendation results is not enough,
Therefore in the embodiment of the present application by the content information of advertisement effectively with the collaborative filtering of Behavior-based control
The collaborative filtering based on Adc for combining, so as to enrich User-Item matrixes, solves user
The sparse series of problems for bringing of interbehavior.
2nd, the abundant User-Item squares of the collaborative filtering based on Adc proposed in the embodiment of the present application
During battle array, the content similarity of text between Ad is obtained by way of information filtering, and by high content
Unknown element in the Ad filling User-Item matrixes of similarity, because the matrix compares content more
Plus it is abundant, so as to carry out the operation such as recommending will obtain more preferable effect on matrix after filling.
Step 102, using the similarity between the final correlation data calculation business object;
Step 103, according to the similarity between the business object business object is recommended.
In a preferred embodiment of the present application, described method can be with following steps:
Using the similarity between the Similarity Measure user profile between the business object;
Recommend business object according to the similarity between the user profile.
The embodiment of the present application can apply to various different recommendation scenes, such as Ad triggerings are recommended
Scene triggers the scene recommended with User.Specifically, under the scene recommended is triggered in Ad, can be with
Calculate the similarity between Ad to carry out Ad recommendations, trigger under the scene recommended, then in User
The similarity between User can be calculated to carry out Ad recommendations.
For Ad triggers the scene recommended, User is may refer on certain webpage or certain platform
When clicking on some Ad, then can now be recommended as User A and recommend to surpass with the Ad similarities clicked on
Cross certain and recommend threshold value, or sequence in the Ad of former.
For User trigger recommend scene, may refer to User A log in certain webpage or certain
During individual platform, just recommend for User A and the similarity between User A exceed certain recommendation threshold value,
Or sequence is in the association Ad of the User B of former.
In a preferred embodiment of the present application, the step 102 can include following sub-step:
Sub-step 41, determines the corresponding matrix element of any two business object in the abundant matrix;
Sub-step 42, it is the first record identification to count the corresponding matrix element of described two business objects
Number;
Sub-step 43, determines the contribution weight of each user profile;
Sub-step 44, using the contribution weight and described two business objects of each user profile
Corresponding matrix element is the number of the first record identification, is calculated between any two business object
Similarity.
In a preferred embodiment of the present application, the sub-step 43 can include following sub-step:
Sub-step 43-1, it is the first record identification to obtain the corresponding matrix element of each user profile
Number;
Sub-step 43-2, adopts the corresponding matrix element of described each user profile for the first record mark
The number of knowledge, obtains the contribution weight of each user profile.
In the embodiment of the present application, can obtain on the basis of abundant matrix by using formula (2)
Similarity between any Item, that is, realize based on the collaborative filtering on Adc.
Wherein, Ii, Ij are represented i-th and j-th Ad or Item, U (Ii) representing matrix i-th is arranged
Middle matrix element is the User of yes, and Wv is the weight of user v contributions, can specifically pass through formula (3)
To be calculated:
Wherein, N (v) represents the element number of the corresponding yes of user v places this line of matrix, because
Actually it has been generally acknowledged that the excessive User of interbehavior similar to Ad contribution similarity is less.Separately
Outward, the value of the log truth of a matter can be other numerical value such as 2 or 10, and the embodiment of the present invention is not added with to this
To limit.
Specifically, when the scene recommended is triggered into Ad, the similarity between Ad can be calculated,
Then other Ad of correlation are accordingly recommended according to the Ad of user's triggering.
Similarity meter in a preferred embodiment of the present application, between the employing business object
The step of calculating the similarity between user profile can include following sub-step:
Sub-step 51, for the business object that there are interbehavior data between any user information,
Business object of the similarity between the business object in default similarity threshold is organized into the user
The corresponding user preferences set of information;
Sub-step 52, any user profile is calculated according to preset rules using the user preferences set
Between similarity.
In the embodiment of the present application, when the scene recommended is triggered into User, user can be first directed to
The ad of history of existence interbehavior, using the similarity between the Ad and Ad for formerly obtaining, constitutes and uses
The bigger hobby Ad set in family, the hobby Ad set includes the Ad of user direct interaction itself
And the similar Ad lists of these direct interactions Ad).
Assume that the Ad for having interbehavior for User A is collected, it is then sharp on this basis
With the similarity between Ad, same set will be organized as more than the Ad of a certain similarity threshold,
With the hobby Ad set that this obtains User A, then gather to count according to hobby Ad of User A
Calculate and other User hobby Ad set be calculating the similarity between a User, subsequently can by with
Certain similarity exceedes a certain recommendation threshold value, or sequence in the happiness of the User of former between User A
Good Ad represents as the candidate hobby Ad of User A, so as to further enrich User dimensions
Recommend Ad set.
In order that those skilled in the art more fully understand the embodiment of the present application, below using specific
The process that example is recommended to illustrate the application to realize, one kind of the application with reference to shown in Fig. 4 is based on
The broad flow diagram that advertisement is recommended, by taking advertisement as an example, this Shen are realized in collaboration in combination with information filtering
Recommendation process that please be described can be divided in realization four main modulars:
(1) profit after being collected using various advertisement interbehaviors (as clicked on, feeding back, conclude the business, collect)
With the collaborative filtering method based on Item, the phase based on interbehavior between advertisement is so can be obtained by
Seemingly spend, but now lacking due to interbehavior data, and most of advertisement can not obtain similar
Advertisement;
(2) Adc (advertisement cluster) is constituted:Mainly clustered by way of information filtering, it is main
It is to build the similarity between advertisement, the higher advertisement of similarity according to the information of advertisement itself
As same Adc;
(3) collaborative filtering based on Adc:The Adc that will be constituted using advertisement content information itself
Effectively it is combined with user behavior interaction data, obtains the wider related data of overlay capacity,
Going out the similarity between advertisement according to correlation data calculation is used to recommend, and can efficiently solve user
Sparse can the not push away accurate similarity advertisement Universal Problems with causing of interbehavior;
(4) the embodiment of the present application can apply to two kinds of different scenes, be respectively that Ad triggerings are pushed away
The scene recommended triggers the scene recommended with User;For the User that non-search region be directly facing User is touched
The scene recommended is sent out, the embodiment of the present application there can also be 2 steps to operate:
A () can first for the Ad, the Ad obtained using (3rd) step of user's history interbehavior
Similarity between Ad, constitutes bigger hobby Ad of user and gathers (including user direct interaction itself
Ad and the similar Ad lists of these direct interactions Ad);
B collaborative filtering that () the embodiment of the present application tries again based on User on the basis of (a),
The hobby Ad set of User is obtained after (a), is calculated according to the hobby Ad collection of User
Similarity between User, using hobby Ad of the high User of certain User similarity as the User
Candidate hobby Ad representing, so as to further enrich the recommendation Ad set of User dimensions.This
Bright innovation essentially consists in structure Adc, and realizes, based on the collaborative filtering of Adc, making full use of simultaneously
The text message and user mutual Ad behavioral data of Ad itself are combined, is pushed away so as to solve main flow
Recommend the sparse caused problem of Technical behaviour.
Another vivid process description of the embodiment of the present application with reference to shown in Fig. 5, specific mistake
Cheng Shi:
(1) according to the cross-correlation between user and advertisement (as clicked on, feeding back, conclude the business, collect
These are operation associated) it is compressed and merges, obtain the interbehavior data between user and advertisement;
(2) the Item-based CF (associations based on Item are carried out according to the interbehavior data for obtaining
With filtration), obtain initial related data;
(3) Adc is obtained by information filtering, the initial related data separate Adc is carried out
Adc-based CF (collaborative filtering based on Adc) obtain final related data, and according to most last phase
The similarity that data are obtained between advertisement is closed, and/or, the similarity between user;Wherein, Adc be by
Advertisement in advertisement base, the title and details content according to corresponding to advertisement is clustered by advertisement
The higher advertisement of similarity is organized as advertisement cluster by mode, namely refers to Adc;
(4) when user searches for certain commodity on certain platform, or when user clicks on certain commodity
Advertisement when, Offer triggerings now can recommend related advertisement according to the similarity between advertisement;
(5) when user conducts interviews on certain platform, User triggerings, it will carry out
User-based CF (collaborative filtering based on User), push away according to the similarity between user for user
Recommend the advertisement of correlation.
In the embodiment of the present application, in the application scenarios of Recommendations, usual user is corresponding and business
Product or the interbehavior data of advertisement are sparse, therefore the initial related data for obtaining also is often
It is sparse, this also result in using initial related data commodity or advertisement recommendation results it is inaccurate
And the problem that overlay capacity is not enough, therefore the commodity of combined content collaborative filtering or wide in the embodiment of the present application
Accuse cluster to fill initial related data, so as to obtain more abundant final related data, and be used for
In Recommendations or the application scenarios of advertisement, solve that interbehavior Sparse brings is a series of
Problem, for example, operation is carried out on final related data after filling will obtain preferably recommendation effect
Really, the accuracy rate of recommendation is improved, so-called " thinking that you are thought " can be realized, be that user shows symbol
The commodity of the demand of user or advertisement when being fated, improve Consumer's Experience effect.
It should be noted that for embodiment of the method, in order to be briefly described, therefore it is all expressed as
A series of combination of actions, but those skilled in the art should know, and the embodiment of the present application is not
Limited by described sequence of movement, because according to the embodiment of the present application, some steps can be adopted
With other order or while carry out.Secondly, those skilled in the art also should know, description
Described in embodiment belong to preferred embodiment, involved action not necessarily the application
Necessary to embodiment.
With reference to Fig. 6, a kind of structural frames of the recommendation apparatus embodiment of business object of the application are shown
Figure, specifically can include such as lower module:
Final related data acquisition module 201, for obtaining between user profile and business object most
Whole related data;
Business object similarity calculation module 202, for using the final correlation data calculation business
Similarity between object, and/or, the similarity between user profile;
First business object recommending module 203, for according to the similarity between the business object,
And/or, the similarity between user profile recommends business object;
In a preferred embodiment of the present application, the business object similarity calculation module 202
Following submodule can be included:
Matrix element determination sub-module, for determining the abundant matrix in any two business object
Corresponding matrix element;
First record identification number statistic submodule, it is corresponding for counting described two business objects
Matrix element is the number of the first record identification;
Contribution weight determination sub-module, for determining the contribution weight of each user profile;
Similarity Measure submodule between business object, for using each user profile
Contribution weight and the number that the corresponding matrix element of described two business objects is the first record identification,
Calculate the similarity between any two business object.
In a preferred embodiment of the present application, the contribution weight determination sub-module can include
Such as lower unit:
First record identification number statistic unit, for obtaining the corresponding matrix element of each user profile
Element is the number of the first record identification;
Contribution weight obtaining unit, for adopt the corresponding matrix element of described each user profile for
The number of the first record identification, obtains the contribution weight of each user profile.
In a preferred embodiment of the present application, described device can also be included such as lower module:
User profile similarity calculation module, based on using the similarity between the business object
Calculate the similarity between user profile;
Second business object recommending module, for recommending according to the similarity between the user profile
Business object.
In a preferred embodiment of the present application, the user preferences set obtains submodule can be with
Including following submodule:
User preferences set obtains submodule, for interacting for existing between any user information
The business object of behavioral data, by the similarity between the business object in default similarity threshold
Business object organize the corresponding user preferences set of the user profile;
Similarity Measure submodule between user profile, for adopting the use according to preset rules
Family hobby set calculates the similarity between any user profile.
Wherein, described device can also be included such as lower module:
Interbehavior data acquisition module, for obtaining interacting between user profile and business object
Behavioral data;The business object has corresponding service feature data;
Initial correlation data generation module, for according to the interbehavior data, generating the use
Initial related data between family information and business object;
Business object cluster is divided into module, for using the service feature data by the business object
It is divided into business object cluster;
Final related data generation module, for initial related to described according to the business object cluster
Data, generate the final related data between the user profile and business object.
In a preferred embodiment of the present application, the initial related data generation module can be wrapped
Include following submodule:
Matrix creates submodule, for creating matrix;The matrix is by the user profile and business
Object is constituted, and the matrix includes matrix element;
Interbehavior judging submodule, for whether judging between the user profile and business object
There are interbehavior data;If so, the first filling submodule is then called, if it is not, then calling second
Filling submodule;
First filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the first record identification;
Second filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the second record identification.
In a preferred embodiment of the present application, institute's business object cluster be divided into module can include as
Lower submodule:
Service feature data acquisition submodule, for obtaining the corresponding service feature of the business object
Data;
Core set of words determination sub-module, for determining the business using the service feature data
The core set of words of object;
Content similarity calculating sub module, for using the core set of words calculate business object it
Between content similarity;
Business object acquisition submodule, for obtaining content similarity preset content similarity threshold is reached
The business object of value;
Business object cluster combines submodule, similar for the content similarity to be reached into preset content
The business object of degree threshold value is combined as business object cluster.
In a preferred embodiment of the present application, the final related data generation module can be wrapped
Include following submodule:
Matrix element inquires about submodule, for inquiring about in the matrix as the square of the first record identification
Array element element;The matrix element has corresponding user profile and business object;
Business object judging submodule, for judging the business object cluster in business object whether
With it is described be the first record identification the corresponding business object of matrix element it is identical;If so, then call
User profile acquisition submodule;
User profile acquisition submodule, for obtaining the matrix element pair for the first record identification
The user profile answered;
3rd filling submodule, in the matrix by with the user information correlation, and
The first record identification is filled to the business object identical matrix element in the business object cluster;
Matrix horn of plenty matrix after the filling.
In a preferred embodiment of the present application, the business object can be commodity, the friendship
Mutually behavioral data can include click data, feedback data, transaction data and/or collection data, institute
The service feature data for stating business object data can be the title and details content of commodity.
For device embodiment, due to itself and embodiment of the method basic simlarity, so description
Fairly simple, related part is illustrated referring to the part of embodiment of the method.
Each embodiment in this specification is described by the way of progressive, each embodiment emphasis
What is illustrated is all the difference with other embodiment, identical similar part between each embodiment
Mutually referring to.
Those skilled in the art it should be appreciated that the embodiment of the embodiment of the present application can be provided as method,
Device or computer program.Therefore, the embodiment of the present application can using complete hardware embodiment,
Complete software embodiment or the form with reference to the embodiment in terms of software and hardware.And, this Shen
Please embodiment can adopt and wherein include the computer of computer usable program code at one or more
It is real in usable storage medium (including but not limited to disk memory, CD-ROM, optical memory etc.)
The form of the computer program applied.
In a typical configuration, the computer equipment includes one or more processors
(CPU), input/output interface, network interface and internal memory.Internal memory potentially includes computer-readable medium
In volatile memory, the shape such as random access memory (RAM) and/or Nonvolatile memory
Formula, such as read only memory (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium
Example.Computer-readable medium includes permanent and non-permanent, removable and non-removable media
Information Store can be realized by any method or technique.Information can be computer-readable instruction,
Data structure, the module of program or other data.The example of the storage medium of computer includes, but
It is not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic random to deposit
Access to memory (DRAM), other kinds of random access memory (RAM), read only memory
(ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other in
Deposit technology, read-only optical disc read only memory (CD-ROM), digital versatile disc (DVD) or other
Optical storage, magnetic cassette tape, tape magnetic rigid disk storage other magnetic storage apparatus or it is any its
His non-transmission medium, can be used to store the information that can be accessed by a computing device.According to herein
Define, computer-readable medium does not include the computer readable media (transitory media) of non-standing,
Such as the data signal and carrier wave of modulation.
The embodiment of the present application be with reference to according to the method for the embodiment of the present application, terminal unit (system) and
The flow chart and/or block diagram of computer program is describing.It should be understood that can be by computer journey
Sequence instructs flowchart and/or each flow process and/or square frame and flow chart in block diagram
And/or the combination of the flow process in block diagram and/or square frame.These computer program instructions can be provided
To general purpose computer, special-purpose computer, Embedded Processor or other programmable data processing terminals
The processor of equipment is producing a machine so that processed by computer or other programmable datas
The instruction of the computing device of terminal unit is produced for realizing in one flow process of flow chart or multiple streams
The device of the function of specifying in one square frame of journey and/or block diagram or multiple square frames.
These computer program instructions may be alternatively stored in can be guided at computer or other programmable datas
In the computer-readable memory that reason terminal unit works in a specific way so that be stored in the calculating
Instruction in machine readable memory produces the manufacture for including command device, and the command device is realized
Specify in one flow process of flow chart or one square frame of multiple flow processs and/or block diagram or multiple square frames
Function.
These computer program instructions can also be loaded into computer or other programmable data processing terminals
On equipment so that on computer or other programmable terminal equipments perform series of operation steps with
Computer implemented process is produced, so as to what is performed on computer or other programmable terminal equipments
Instruction is provided for realizing in one square frame of one flow process of flow chart or multiple flow processs and/or block diagram
Or specify in multiple square frames function the step of.
Although having been described for the preferred embodiment of the embodiment of the present application, those skilled in the art
Once knowing basic creative concept, then other change and modification can be made to these embodiments.
So, claims are intended to be construed to include preferred embodiment and fall into the embodiment of the present application
Scope has altered and changes.
Finally, in addition it is also necessary to explanation, herein, such as first and second or the like relation
Term is used merely to make a distinction an entity or operation with another entity or operation, and not
Necessarily require either to imply and there is any this actual relation or suitable between these entities or operation
Sequence.And, term " including ", "comprising" or its any other variant are intended to nonexcludability
Comprising so that a series of process, method, article or terminal unit including key elements is not only
Including those key elements, but also including other key elements being not expressly set out, or also include for
The intrinsic key element of this process, method, article or terminal unit.In no more restrictions
In the case of, the key element limited by sentence " including ... ", it is not excluded that including the key element
Also there is other identical element in process, method, article or terminal unit.
Recommendation method above to a kind of business object provided herein and a kind of business object
Recommendation apparatus, are described in detail, specific case used herein to the principle of the application and
Embodiment is set forth, and the explanation of above example is only intended to help the side for understanding the application
Method and its core concept;Simultaneously for one of ordinary skill in the art, according to the think of of the application
Think, will change in specific embodiments and applications, in sum, this explanation
Book content should not be construed as the restriction to the application.
Claims (17)
1. a kind of recommendation method of business object, it is characterised in that include:
Obtain the final related data between user profile and business object;
Using the similarity between the final correlation data calculation business object;
Recommend business object according to the similarity between the business object;
Wherein,
Final related data between the user profile and business object is generated in the following way:
Obtain the interbehavior data between user profile and business object;The business object has
Corresponding service feature data;
According to the interbehavior data, generate initial between the user profile and business object
Related data;
The business object is divided into into business object cluster using the service feature data;
According to the business object cluster and the initial related data, the user profile and industry are generated
Final related data between business object.
2. method according to claim 1, it is characterised in that described according to interbehavior number
According to including the step of generate the initial related data between the user profile and business object:
Create matrix;The matrix is made up of the user profile and business object, in the matrix
Including matrix element;
Judge to whether there is interbehavior data between the user profile and business object;
If so, then in the matrix by user profile matrix element corresponding with business object
It is filled to the first record identification;
If it is not, then by user profile matrix element corresponding with business object in the matrix
It is filled to the second record identification.
3. method according to claim 1, it is characterised in that the employing service feature number
According to including the step of the business object is divided into into business object cluster:
Obtain the corresponding service feature data of the business object;
Determine the core set of words of the business object using the service feature data;
Content similarity between business object is calculated using the core set of words;
Obtain the business object that content similarity reaches preset content similarity threshold;
The business object that the content similarity reaches preset content similarity threshold is combined as into business
Object cluster.
4. method according to claim 1, it is characterised in that described according to business object cluster
With the initial related data, the final dependency number between the user profile and business object is generated
According to the step of include:
Inquiry is the matrix element of the first record identification in the matrix;The matrix element has
Corresponding user profile and business object;
Judge the business object in the business object cluster whether with it is described be the first record identification square
The corresponding business object of array element element is identical;
If so, the corresponding user profile of matrix element for the first record identification is then obtained;
In the matrix by with the user information correlation, and with the business object cluster in
Business object identical matrix element is filled to the first record identification;Matrix after the filling is rich
Rich matrix.
5. the method according to claim 1 or 2 or 3 or 4, it is characterised in that described to adopt
The step of with similarity between final correlation data calculation business object, includes:
Determine the corresponding matrix element of any two business object in the abundant matrix;
Count the number that the corresponding matrix element of described two business objects is the first record identification;
Determine the contribution weight of each user profile;
Using the contribution weight and the corresponding matrix of described two business objects of each user profile
Element is the number of the first record identification, calculates the similarity between any two business object.
6. method according to claim 5, it is characterised in that the determination each user's letter
The step of contribution weight of breath, includes:
Obtain the number that the corresponding matrix element of each user profile is the first record identification;
Adopt the corresponding matrix element of described each user profile for the number of the first record identification, obtain
Obtain the contribution weight of each user profile.
7. the method according to claim 1 or 2 or 3 or 4, it is characterised in that also include:
Using the similarity between the Similarity Measure user profile between the business object;
Recommend business object according to the similarity between the user profile.
8. method according to claim 7, it is characterised in that the employing business object it
Between Similarity Measure user profile between similarity the step of include:
For the business object that there are interbehavior data between any user information, by the industry
Similarity between business object organizes the user profile correspondence in the business object of default similarity threshold
User preferences set;
Calculate similar between any user profile using the user preferences set according to preset rules
Degree.
9. method according to claim 1, it is characterised in that the business object is commodity,
The interbehavior data include click data, feedback data, transaction data and/or collection data,
The service feature data of the business object data are the title and details content of commodity.
10. a kind of recommendation apparatus of business object, it is characterised in that include:
Final related data acquisition module, it is final between user profile and business object for obtaining
Related data;
Business object similarity calculation module, for using the final correlation data calculation business pair
Similarity as between;
First business object recommending module, for recommending according to the similarity between the business object
Business object;
Wherein, described device also includes:
Interbehavior data acquisition module, for obtaining interacting between user profile and business object
Behavioral data;The business object has corresponding service feature data;
Initial correlation data generation module, for according to the interbehavior data, generating the use
Initial related data between family information and business object;
Business object cluster is divided into module, for using the service feature data by the business object
It is divided into business object cluster;
Final related data generation module, for initial related to described according to the business object cluster
Data, generate the final related data between the user profile and business object.
11. devices according to claim 10, it is characterised in that the initial related data
Generation module includes:
Matrix creates submodule, for creating matrix;The matrix is by the user profile and business
Object is constituted, and the matrix includes matrix element;
Interbehavior judging submodule, for whether judging between the user profile and business object
There are interbehavior data;If so, the first filling submodule is then called, if it is not, then calling second
Filling submodule;
First filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the first record identification;
Second filling submodule, in the matrix by the user profile and business object pair
The matrix element answered is filled to the second record identification.
12. devices according to claim 10, it is characterised in that institute's business object cluster is divided into
Module includes:
Service feature data acquisition submodule, for obtaining the corresponding service feature of the business object
Data;
Core set of words determination sub-module, for determining the business using the service feature data
The core set of words of object;
Content similarity calculating sub module, for using the core set of words calculate business object it
Between content similarity;
Business object acquisition submodule, for obtaining content similarity preset content similarity threshold is reached
The business object of value;
Business object cluster combines submodule, similar for the content similarity to be reached into preset content
The business object of degree threshold value is combined as business object cluster.
13. devices according to claim 10, it is characterised in that the final related data
Generation module includes:
Matrix element inquires about submodule, for inquiring about in the matrix as the square of the first record identification
Array element element;The matrix element has corresponding user profile and business object;
Business object judging submodule, for judging the business object cluster in business object whether
With it is described be the first record identification the corresponding business object of matrix element it is identical;If so, then call
User profile acquisition submodule;
User profile acquisition submodule, for obtaining the matrix element pair for the first record identification
The user profile answered;
3rd filling submodule, in the matrix by with the user information correlation, and
The first record identification is filled to the business object identical matrix element in the business object cluster;
Matrix horn of plenty matrix after the filling.
14. devices according to claim 10 or 11 or 12 or 13, it is characterised in that
The business object similarity calculation module includes:
Matrix element determination sub-module, for determining the abundant matrix in any two business object
Corresponding matrix element;
First record identification number statistic submodule, it is corresponding for counting described two business objects
Matrix element is the number of the first record identification;
Contribution weight determination sub-module, for determining the contribution weight of each user profile;
Similarity Measure submodule between business object, for using each user profile
Contribution weight and the number that the corresponding matrix element of described two business objects is the first record identification,
Calculate the similarity between any two business object.
15. devices according to claim 14, it is characterised in that the contribution weight determines
Submodule includes:
First record identification number statistic unit, for obtaining the corresponding matrix element of each user profile
Element is the number of the first record identification;
Contribution weight obtaining unit, for adopt the corresponding matrix element of described each user profile for
The number of the first record identification, obtains the contribution weight of each user profile.
16. devices according to claim 10 or 11 or 12 or 13, it is characterised in that
Also include:
User profile similarity calculation module, based on using the similarity between the business object
Calculate the similarity between user profile;
Second business object recommending module, for recommending according to the similarity between the user profile
Business object.
17. devices according to claim 16, it is characterised in that the user preferences set
Obtaining submodule includes:
User preferences set obtains submodule, for interacting for existing between any user information
The business object of behavioral data, by the similarity between the business object in default similarity threshold
Business object organize the corresponding user preferences set of the user profile;
Similarity Measure submodule between user profile, for adopting the use according to preset rules
Family hobby set calculates the similarity between any user profile.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510657556.2A CN106570031A (en) | 2015-10-13 | 2015-10-13 | Service object recommending method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510657556.2A CN106570031A (en) | 2015-10-13 | 2015-10-13 | Service object recommending method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106570031A true CN106570031A (en) | 2017-04-19 |
Family
ID=58508672
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510657556.2A Pending CN106570031A (en) | 2015-10-13 | 2015-10-13 | Service object recommending method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106570031A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108052639A (en) * | 2017-12-21 | 2018-05-18 | 中国联合网络通信集团有限公司 | Industry user based on carrier data recommends method and device |
CN109460518A (en) * | 2018-12-07 | 2019-03-12 | 杭州东信北邮信息技术有限公司 | A kind of book recommendation method based on user website access record |
CN109636479A (en) * | 2018-12-19 | 2019-04-16 | 武汉斗鱼鱼乐网络科技有限公司 | A kind of advertisement recommended method, device, electronic equipment and storage medium |
CN109740063A (en) * | 2019-01-17 | 2019-05-10 | 北京奇艺世纪科技有限公司 | Information recalls, information cluster method, device and equipment |
CN109934673A (en) * | 2019-02-25 | 2019-06-25 | 校宝在线(杭州)科技股份有限公司 | A kind of intelligent sales lead distribution method based on recommender system |
CN110163662A (en) * | 2019-04-26 | 2019-08-23 | 阿里巴巴集团控股有限公司 | A kind of business model training method, device and equipment |
CN110766493A (en) * | 2018-07-26 | 2020-02-07 | 阿里巴巴集团控股有限公司 | Business object providing method, server, electronic device and storage medium |
CN112256958A (en) * | 2020-09-27 | 2021-01-22 | 中智关爱通(上海)科技股份有限公司 | Commodity recommendation method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102789462A (en) * | 2011-05-18 | 2012-11-21 | 阿里巴巴集团控股有限公司 | Project recommendation method and system |
CN103353880A (en) * | 2013-06-20 | 2013-10-16 | 兰州交通大学 | Data mining method adopting dissimilarity degree clustering and association |
CN104965903A (en) * | 2015-06-30 | 2015-10-07 | 北京奇艺世纪科技有限公司 | Resource recommendation method and apparatus |
-
2015
- 2015-10-13 CN CN201510657556.2A patent/CN106570031A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102789462A (en) * | 2011-05-18 | 2012-11-21 | 阿里巴巴集团控股有限公司 | Project recommendation method and system |
CN103353880A (en) * | 2013-06-20 | 2013-10-16 | 兰州交通大学 | Data mining method adopting dissimilarity degree clustering and association |
CN104965903A (en) * | 2015-06-30 | 2015-10-07 | 北京奇艺世纪科技有限公司 | Resource recommendation method and apparatus |
Non-Patent Citations (1)
Title |
---|
熊拥军等: "《数字图书馆个性化服务研究与实践 基于新型决策支持系统》", 30 September 2012 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108052639A (en) * | 2017-12-21 | 2018-05-18 | 中国联合网络通信集团有限公司 | Industry user based on carrier data recommends method and device |
CN110766493A (en) * | 2018-07-26 | 2020-02-07 | 阿里巴巴集团控股有限公司 | Business object providing method, server, electronic device and storage medium |
CN110766493B (en) * | 2018-07-26 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Service object providing method, server, electronic device, and storage medium |
CN109460518A (en) * | 2018-12-07 | 2019-03-12 | 杭州东信北邮信息技术有限公司 | A kind of book recommendation method based on user website access record |
CN109460518B (en) * | 2018-12-07 | 2020-07-24 | 杭州东信北邮信息技术有限公司 | Book recommendation method based on user website access records |
CN109636479A (en) * | 2018-12-19 | 2019-04-16 | 武汉斗鱼鱼乐网络科技有限公司 | A kind of advertisement recommended method, device, electronic equipment and storage medium |
CN109740063A (en) * | 2019-01-17 | 2019-05-10 | 北京奇艺世纪科技有限公司 | Information recalls, information cluster method, device and equipment |
CN109934673A (en) * | 2019-02-25 | 2019-06-25 | 校宝在线(杭州)科技股份有限公司 | A kind of intelligent sales lead distribution method based on recommender system |
CN110163662A (en) * | 2019-04-26 | 2019-08-23 | 阿里巴巴集团控股有限公司 | A kind of business model training method, device and equipment |
CN110163662B (en) * | 2019-04-26 | 2024-04-05 | 创新先进技术有限公司 | Service model training method, device and equipment |
CN112256958A (en) * | 2020-09-27 | 2021-01-22 | 中智关爱通(上海)科技股份有限公司 | Commodity recommendation method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106570031A (en) | Service object recommending method and device | |
US9336282B2 (en) | Systems and methods for identifying and analyzing internet users | |
US20170206276A1 (en) | Large Scale Recommendation Engine Based on User Tastes | |
US8893012B1 (en) | Visual indicator based on relative rating of content item | |
CN106503025B (en) | Application recommendation method and system | |
CN106803190A (en) | A kind of ad personalization supplying system and method | |
US20110184977A1 (en) | Recommendation method and system based on collaborative filtering | |
CN103970850B (en) | Site information recommends method and system | |
CN105975472A (en) | Method and device for recommendation | |
CN105913273A (en) | Intelligent recommendation all-terminal display method and device | |
CN104102648A (en) | User behavior data based interest recommending method and device | |
EP2304619A1 (en) | Correlated information recommendation | |
CN105718184A (en) | Data processing method and apparatus | |
CN107230098A (en) | Method and system is recommended in a kind of timesharing of business object | |
CN102063433A (en) | Method and device for recommending related items | |
CN101482884A (en) | Cooperation recommending system based on user predilection grade distribution | |
CN107292648A (en) | A kind of user behavior analysis method and device | |
Miklošík | Changes in purchasing decision-making process of consumers in the digital era | |
US10896458B2 (en) | Method, system, and computer-readable medium for product and vendor selection | |
US20170193531A1 (en) | Intelligent Digital Media Content Creator Influence Assessment | |
CN105894310A (en) | Personalized recommendation method | |
CN108629010A (en) | It is a kind of to recommend method based on the web services of theme and service combination information | |
CN105426550A (en) | Collaborative filtering tag recommendation method and system based on user quality model | |
US8997008B2 (en) | System and method for searching through a graphic user interface | |
CN106570020A (en) | Method and apparatus used for providing recommended information |
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 | ||
CB03 | Change of inventor or designer information |
Inventor after: Wu Lixia Inventor after: Zhan Yang Inventor after: Cheng Zhangling Inventor before: Wu Lixia |
|
CB03 | Change of inventor or designer information | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170419 |
|
RJ01 | Rejection of invention patent application after publication |