CN104361062B - A kind of recommendation method and device of related information - Google Patents
A kind of recommendation method and device of related information Download PDFInfo
- Publication number
- CN104361062B CN104361062B CN201410610726.7A CN201410610726A CN104361062B CN 104361062 B CN104361062 B CN 104361062B CN 201410610726 A CN201410610726 A CN 201410610726A CN 104361062 B CN104361062 B CN 104361062B
- Authority
- CN
- China
- Prior art keywords
- node
- information
- hops
- information node
- user
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
-
- 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
- G06F16/951—Indexing; Web crawling techniques
Abstract
The invention provides a kind of recommendation method and device of related information, wherein, this method includes：Obtain m the first weighted values corresponding with the m second-level message node that first order information node is associated, wherein, the m the first weighted values redirect behavior by the click of user and got, and the first order information node is the currently used information node of the user, and m is positive integer；It is the related information that the user determines to recommend according to the m the first weighted values and default n recommendation information node, wherein, n is positive integer.The click that the embodiment of the present invention with reference to user by m the first weighted values redirects behavior, avoid the content focus issues to user's recommended node information by the content of information node, the uploader of information node or author, the type of information node etc., because the acquisition modes of m the first weighted values are to redirect behavior with reference to the click of user, realize by big data thought and to recommend related information to user.
Description
Technical field
The present invention relates to the recommendation method and device of field of computer technology, more particularly to a kind of related information.
Background technology
In computer technology and internet arena, one is made from search needs, to display carrying, then to potential
The closed-loop circulation experience of needarousal, the recommendation of related information is then the committed step for exciting potential demand, and it can be maximized
Ground shortens user's step-length and screening cost, simultaneously, moreover it is possible to substantially flow is deposited in specific product system, so that structure
Build user's viscosity.
When being associated the recommendation of information, so that the information of recommendation is specially video as an example, prior art passes through based on interior
When the association of appearance is to user's recommendation video, mainly from video, it is considered to the similarity of video information, including：Video
The information such as title (title), video presentation, video uploader, video scoring, video click volume.Inventor has found, based on content
The way of recommendation, the associated video of recommendation is all close in content, therefore the problem of can produce Content aggregation, it is impossible to excite use
The point of interest at family.
The content of the invention
Embodiments of the invention provide a kind of recommendation method and device of related information, it is to avoid by information node
Hold, the uploader of information node or author, the type of information node etc. to user's recommended node information content focus issues,
Excite the point of interest of user.
To reach above-mentioned purpose, embodiments of the invention are adopted the following technical scheme that：
A kind of recommendation method of related information, this method includes：
M the first weighted values corresponding with the m second-level message node that first order information node is associated are obtained, its
In, the m the first weighted values redirect behavior by the click of user and got, and the first order information node is the user
Currently used information node, m is positive integer；
It is that the user determines that recommends associates according to the m the first weighted values and default n recommendation information node
Information, wherein, n is positive integer.
A kind of recommendation apparatus of related information, the device includes：
First acquisition module, for acquisition m corresponding with the m second-level message node that first order information node is associated
Individual first weighted value, wherein, the m the first weighted values redirect behavior by the click of user and got, the first order letter
It is the currently used information node of the user to cease node, and m is positive integer；
Recommending module, for being that the user is true according to the m the first weighted values and default n recommendation information node
Surely the related information recommended, wherein, n is positive integer.
The recommendation method and device of related information provided in an embodiment of the present invention, with reference to use by m the first weighted values
The click at family redirects behavior, it is to avoid pass through the content of information node, the uploader of information node or author, information node
Content focus issues from type etc. to user's recommended node information, because the acquisition modes of m the first weighted values are to refer to user
Click redirect behavior, realize by big data thought to user recommend related information.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention one is provided.
Fig. 2 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention two is provided.
Fig. 3 is the schematic diagram of the related information digraph of embodiment illustrated in fig. 2.
Fig. 4 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention three is provided.
Fig. 5 is the schematic diagram of the related information digraph of embodiment illustrated in fig. 4.
Fig. 6 is the structural representation of the recommendation apparatus for the related information that the embodiment of the present invention four is provided.
Fig. 7 is the structural representation of the recommendation apparatus for the related information that the embodiment of the present invention five is provided.
Embodiment
The recommendation method and device to related information of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings.
Embodiment one：
Fig. 1 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention one is provided；As shown in figure 1, this
Inventive embodiments comprise the following steps：
Step 101, m first power corresponding with the m second-level message node that first order information node is associated is obtained
Weight values, wherein, m the first weighted values redirect behavior by the click of user and got, and first order information node is that user is current
The information node used, m is positive integer；
Step 102, it is that user determines that recommends associates according to m the first weighted values and default n recommendation information node
Information, wherein, n is positive integer.
In embodiments of the present invention, first order information node is A, m second-level message node be B1, B2 ..., Bm, A,
B1, B2 ..., the incidence relation between Bm represent that weight is bigger by weight, show that correlation degree is stronger.It can combine and use
The click at family redirects behavior and gets weighted value.
The recommendation method of related information provided in an embodiment of the present invention, the point of user is with reference to by m the first weighted values
Hit the behavior of redirecting, it is to avoid pass through the content of information node, the uploader of information node or author, the type of information node etc.
To the content focus issues of user's recommended node information, because the acquisition modes of m the first weighted values are the clicks with reference to user
Redirect behavior, realize by big data thought to user recommend related information.
Embodiment two：
Fig. 2 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention two is provided, and Fig. 3 is shown in Fig. 2
The schematic diagram of the related information digraph of embodiment；The embodiment of the present invention is more than or waited with the number m of second-level message node
It is illustrative exemplified by the number n for the information node recommended to user, as shown in Fig. 2 the embodiment of the present invention is included such as
Lower step：
Step 201, click number of hops of the user in first order information node is obtained, wherein, clicking on number of hops includes
First number of hops and the second number of hops.
Step 202, m associated with first order information node are obtained according to the first number of hops and the second number of hops
M the first weighted values of second-level message node.
Step 203, m the first weighted values are ranked up, obtain ranking results.
Step 204, n related information node is recommended from m second-level message node for user according to ranking results.
Illustrated with reference to Fig. 3, wherein, first order information node is A information nodes, and second-level message node is
B1, B2 ..., Bm information nodes, relative to the third level information node C1, C2 of A information nodes, further, C1 information nodes
For the second-level message node of B1 information nodes, C2 information nodes are the second-level message node of B2 information nodes.In the present invention
In embodiment, with A information nodes, B1, B2 ..., Bm information nodes, C1 information nodes, C2 information nodes be specially video film
Exemplified by it is illustrative.
As shown in figure 3, have A, B1, B2 ..., multiple video films such as Bm, C1, C2, A, B1, B2 ..., between Bm, C1, C2
Associated by oriented arrow, incidence relation each other can represent that weight is bigger by weight, show association
Degree is stronger.In step 201, the click of user redirect behavior can be by being got in history log, by obtaining user
Click redirect behavior, count click number of hops of the user in A information nodes, wherein, click on number of hops include first
Number of hops and the second number of hops；The click of user, which redirects behavior, mainly two kinds of sources, and one is the active point by user
Hit and redirect, the active is clicked on and the redirected adopting consecutive click chemical reaction A videos that are user in the time interval of a setting and B2 (or B3 ...,
Bm behavior), can obtain the first number of hops, another is pushed away by the correlation of video by counting the number of times actively redirected
Recommend redirecting for progress, belonging to has redirecting for guiding, i.e. for when seeing A videos, the video of associated recommendation is B1, B2 ..., Bm,
Then by count by A videos jump to B1, B2 ..., Bm click number of hops can obtain the second number of hops.
In step 202., equation can be passed throughCalculate m first power
Weight values, wherein, N (A, B) represent from first order information node A to B1, B2 ..., Bm the first number of hops, N (A, *) represent from
First order information node A to B1, B2 ..., the sum of the first number of hops of the common m second-level message nodes of Bm, M (A, B) represent
From first order information node A to B1, B2 ..., Bm the second number of hops, M (A, *) represent from first order information node A to B1,
B2 ..., the sum of the second number of hops of the common m second-level message nodes of Bm, α and β are weight proportion, in one embodiment, α
+ β=1.
Specifically, A and B1, B2 ..., the weight weight between Bm1(A,B1)、weight1(A,B2)、…、
weight1(A,Bm).In order that A information nodes and B1, B2 ..., the calculating of weight between Bm information nodes it is more accurate, and also
Can the similarity calculating method based on term vector, obtain A information nodes respectively with B1, B2 ..., the mark between Bm information nodes
Topic and the similarity weight of descriptionc(A,B1)、weightc(A,B2)、…、weightc(A,Bm).By by the first weighted value
Be weighted and be added with above-mentioned similarity, thus obtain A and B1, B2 ..., more accurate weighted value between Bm, i.e.,：
Weight (A, B1)=γ1*weight1(A,B1)+γ2*weightC(A, B1),
Weight (A, B2)=γ1*weight1(A,B2)+γ2*weightC(A, B2),
…
Weight (A, Bm)=γ1*weight1(A,Bm)+γ2*weightC(A,Bm)。
Wherein, γ1With γ2For weight coefficient, γ1+γ2=1.Further degree, can using the more accurate weighted value as
First weighted value participates in follow-up recommendation process, so that the information node recommended is closer to user's.
In step 203 and step 204, due to m the first weighted values represent B1, B2 ..., Bm information nodes and A information
Correlation degree between node, weighted value is bigger to represent that relevance is stronger, therefore by weight1(A,B1)、weight1(A,
B2)、…、weight1(A, Bm) is descending to be ranked up, and obtains ranking results, for example, m=5, n=3 from 5, it is necessary to regard
In frequency to user recommend 3 videos, ranking results be B4, B2, B3, B5, B1, according to ranking results from B1, B2 ..., B5 believe
It is user's recommendation B4, B2, B3 information node to cease in node, and the relevance of B4, B2, B3 information node and A information nodes is stronger.
The recommendation method of related information provided in an embodiment of the present invention, is jumped by user in the click of first order information node
Turn number of times and recommend the information node associated with first order information node to user, when hopping several due to click and user behavior
It is associated, therefore the embodiment of the present invention is avoided by the content of information node, the uploader of information node or author, information
The problem of content recommendation of the generations such as the type of node is focused on, the click number of hops for passing through user due to weighted value acquisition modes
The association between information node can be produced, realize by big data thought to user recommend related information；In addition, for
Through participating in the second-level message node of weighted value calculating, information node recommendation process thereafter, without participating in again
The calculating of weighted value, it is only necessary to calculate the weighted value of the information node newly increased, therefore improve information node recommendation
Real-time, substantially reduces calculating cycle.
Above-mentioned illustrative so that information node is specially video film as an example, those skilled in the art can also manage
Solution, information node can also be keyword, and such as user inputs " Liu Dehua " in a search engine, then according to above-mentioned hair
Bright embodiment, can also recommend " schoolmate, Guo Fucheng " etc. to user.
Embodiment three：
Fig. 4 is the schematic flow sheet of the recommendation method for the related information that the embodiment of the present invention three is provided, and Fig. 5 is shown in Fig. 4
The schematic diagram of the related information digraph of embodiment；The embodiment of the present invention is less than to user with the number m of second-level message node
It is illustrative exemplified by the number n of the information node of recommendation, as shown in figure 4, the embodiment of the present invention comprises the following steps：
Step 301, click number of hops of the user in first order information node is obtained, wherein, clicking on number of hops includes
First number of hops and the second number of hops.
Step 302, m associated with first order information node are obtained according to the first number of hops and the second number of hops
M the first weighted values of second-level message node.
Step 303, obtain the r third level information node associated with m second-level message node, and calculate r individual the
R the second weighted values between three-level information node and m second-level message node, r is positive integer；
Step 304, according to m the first weighted values and r the second weighted values obtain m second-level message nodes with it is respective right
Multiple recommendations between the multiple third level information nodes answered；
Step 305, it is defined as n-m third level information node of user's recommendation according to multiple recommendations.
Step 306, m second-level message node and n-m third level information node are recommended into user.
Illustrated with reference to Fig. 5, wherein, first order information node is A information nodes, and second-level message node is
B1, B2 information node, relative to the third level information node C1, C2, C3, D1, D2 of A information nodes, further, C1, C2, C3
Information node is the second-level message node of B1 information nodes, and D1, D2 information node are the second-level message section of B2 information nodes
Point.In embodiments of the present invention, carried out so that A information nodes, C1, C2, C3, D1, D2 information node are specially video film as an example
Exemplary illustration.
As shown in figure 3, there is multiple video films such as A, B1, B2, C1, C2, C3, D1, D2, A, B1, B2, C1, C2, C3, D1,
Associated between D2 by oriented arrow, incidence relation each other can represent that weight is bigger, table by weight
Bright correlation degree is stronger.
The process that implements of step 301 and step 302 may be referred to step 201 and step in above-described embodiment two
202 description, will not be repeated here.And calculating 2 obtained the first weighted values by the step is respectively：weight1(A,
B1)、weight1(A,B2)。
In step 303, as shown in Figure 5, it is necessary to recommend 3 videos, now, m=2, n=to user from 5 videos
5, m<N, second-level message node B1, B2 not enough meet recommended requirements, now, are recommending the base of B1, B2 information node to user
On plinth, in addition it is also necessary to traverse node B1, B2 associated nodes (as shown in figure 5, r=5), the information node of high weight is therefrom selected,
I.e. from second-level message node B1 third level information node C1, C2, C3 and second-level message node B2 third level information
Node D1, D2 recommend the information node of 3 high weights to user again.
Calculate 5 the second weights between 5 third level information node C1, C2, C3, D1, D2 and B1, B2 information nodes
Value, the calculating of the second weighted value may refer to the computational methods of the weighted value in above-described embodiment two, will not be repeated here.Calculate
5 obtained the second weighted value respectively weight2(B1,C1)、weight2(B1,C2)、weight2(B1,C3)、weight2
(B2,D1)、weight2(B2,D2)。
In step 304, can calculate multiple recommendations by following equation includes：
recvalue=δ * weight1*weight2, wherein, δ is decay factor, weight1Represent first order information node with
First weighted value of second-level message node, weight2Represent the second power of second-level message node and third level information node
Weight values.
Specifically, in embodiments of the present invention, 5 recommendations are specific as follows：
recvalue1=δ * weight1(A,B1)*weight2(B1, C1),
recvalue2=δ * weight1(A,B1)*weight2(B1, C2),
recvalue3=δ * weight1(A,B1)*weight2(B1, C3),
recvalue4=δ * weight1(A,B2)*weight2(B2, D1),
recvalue5=δ * weight1(A,B2)*weight2(B2,D2)。
Wherein, δ is decay factor.
In step 305 and step 306, the larger information node of weighted value can be selected from multiple recommendations, for example,
In above-mentioned 5 recommendations, if recommendation recvalue5、recvalue3、recvalue2Before coming, then by the corresponding information of recommendation
Node D2, C3, C2 recommend user together with information node B1, B2.
When user is clicked on by first order information node jumps to recommended third level information node D2, C3, C2,
Information node A is shown as in digraph shown in Fig. 5 and information node D2, C3, C2 set up directed edge, and above-mentioned calculating is obtained
Recommendation be used as the initial weight on directed edge.
In another embodiment, if the number n recommended is 10, now above-mentioned second-level message node is 2,
Third level information node is 5, and whole information nodes is also not enough to recommend 10 information nodes to user, in the case,
Remaining difference (10-2-5=3) produces recommended node by the way of recommending at random.
The recommendation method of related information provided in an embodiment of the present invention, is jumped by user in the click of first order information node
Turn number of times and recommend the information node associated with first order information node to user, when hopping several due to click and user behavior
It is associated, therefore the embodiment of the present invention is avoided by the content of information node, the uploader of information node or author, information
The problem of content recommendation of the generations such as the type of node is focused on, the click number of hops for passing through user due to weighted value acquisition modes
The association between information node can be produced, realize by big data thought to user recommend related information；In addition, for
Through participating in the second-level message node of weighted value calculating, information node recommendation process thereafter, without participating in again
The calculating of weighted value, it is only necessary to calculate the weighted value of the information node newly increased, therefore improve information node recommendation
Real-time, substantially reduces calculating cycle.
On the basis of above-described embodiment two and embodiment three, the embodiment of the present invention also comprises the following steps：
First, the title and the similarity of description of first order information node and m second-level message node are calculated；Secondly,
The first initial weighted value between first order information node and m second-level message node is determined according to similarity.
In one embodiment, in the digraph represented by above-mentioned Fig. 3 or Fig. 5, can also with the mode of graph model come
The weight on each side for the relation between the information node recommended set up in model, digraph can pass through user's
Click behavior obtains to calculate.On this basis, by upper random walk (random walk) method of figure, and then each is adjusted
Weight on side, and get second-level message node more associated with first order information node.
When in one embodiment, due to the click behavior or the less click behavior of user without user, Ke Yixian
Initial digraph, thus, the side in initial digraph are produced by the title between information node and the similarity of description
Weight be the relevance degree by video to calculate.Because the quantity of video on network is big, therefore can not be by appointing
Calculating between two videos of meaning calculates to get the weight on initial side.In one embodiment, simhash can be passed through
Method, first finds out the candidate association information node of each information node (for example, video), then passes through the similarity based on term vector
Computational methods, calculate the title (title) of two information nodes and the similarity of description, so as to obtain the side of initial digraph
Weight.
In another embodiment, if information node be an isolated node point (that is, not with any other information
There is incidence relation in node), because in the digraph shown in Fig. 3 or Fig. 5, isolated point does not have out-degree, therefore can not be isolated
Node does correlation recommendation, therefore, in one embodiment, can activate isolated node by way of randomly generating video.Such as
The video that fruit randomly generates, before next time is random, is clicked on by user and redirects, then can pass through because the click of user redirects behavior
The embodiment of the present invention calculates the weight for obtaining video association, if the video randomly generated was not all clicked on by user, then
When next round updates, the recommendation video of association is randomly generated again for the isolated node.
Above-mentioned illustrative so that information node is specially video film as an example, those skilled in the art can also manage
Solution, information node can also be keyword, and such as user inputs " Liu Dehua " in a search engine, then according to above-mentioned hair
Bright embodiment, can also recommend " schoolmate, Guo Fucheng " etc. to user.
Example IV：
Fig. 6 is the structural representation of the recommendation apparatus for the related information that the embodiment of the present invention four is provided；As shown in fig. 6, this
Inventive embodiments include：
First acquisition module 41, the m second-level message node for obtaining with first order information node is associated is corresponding
M the first weighted values, wherein, the first order information node is the currently used information node of user, and m is positive integer；
Recommending module 42, for being the user according to the m the first weighted values and default n recommendation information node
It is determined that the related information recommended, wherein, n is positive integer.
Detailed description and advantageous effects in the embodiment of the present invention refer to above-described embodiment one, no longer go to live in the household of one's in-laws on getting married herein
State.
Embodiment five：
Fig. 7 is the structural representation of the recommendation apparatus for the related information that the embodiment of the present invention five is provided；As shown in fig. 7,
On the basis of above-described embodiment four, the first acquisition module 41 in the embodiment of the present invention includes：
First acquisition unit 411, for obtaining click number of hops of the user in the first order information node, the point
Hitting number of hops includes the first number of hops and the second number of hops；
Second acquisition unit 412, for according to first number of hops and second number of hops obtain with it is described
M the first weighted values of m associated second-level message node of first order information node.
In one embodiment, second acquisition unit 412 calculates m first weighted value by following equation and included：
Wherein, N (a, b) is represented from the first order information node A
To the first number of hops of the information node B in the m second-level message node, N (a, *) is represented from the first order information
Node A is to the sum of the first number of hops of the m second-level message node, and M (a, b) represented from the first order information node
A to described information node B the second number of hops, M (a, *) is represented from the first order information node A to the m second level
The sum of second number of hops of information node, α and β are weight proportion.
In one embodiment, if the m is more than or equal to the n, recommending module 42 includes：
First sequencing unit 421, for the m the first weighted values to be ranked up, obtains ranking results；
First recommendation unit 422, for from the m second-level message node being the use according to the ranking results
Recommend n related information node in family.
In another embodiment, if the m is less than the n, recommending module 42 includes：
3rd acquiring unit 423, for obtaining the r third level information associated with the m second-level message node
Node, r is positive integer；
4th acquiring unit 424, for according to the m the first weighted values obtain the m second-level message nodes and
Multiple recommendations between each self-corresponding multiple third level information nodes；
Second recommendation unit 425, for being defined as the n-m information section that the user recommends according to the multiple recommendation
Point.
In one embodiment, the 4th acquiring unit 424 calculates the multiple recommendation by following equation and included：
recvalue=δ * weight1*weight2,
Wherein, δ is decay factor, weight1Represent the first order information node and the second-level message node
First weighted value, weight2Represent the second weighted value of the second-level message node and the third level information node.
In one embodiment, described device also includes：
Computing module 43, for calculate the first order information node and the m second-level message node title and
The similarity of description；
Second acquisition module 44, for determining the first order information node and the m second according to the similarity
The first initial weighted value between level information node.
Detailed description and advantageous effects in the embodiment of the present invention refer to above-described embodiment two and embodiment three,
This is repeated no more.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, should all be contained
Cover within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (10)
1. a kind of recommendation method of related information, it is characterised in that methods described includes：
Obtain user first order information node click number of hops, the click number of hops include the first number of hops and
Second number of hops, the first order information node is the currently used information node of the user, first number of hops
Behavior is redirected by the active click of the user to obtain, second number of hops is carried out recommendation information by the user
Click redirect behavior and obtain；
M associated with the first order information node are obtained according to first number of hops and second number of hops
M the first weighted values of second-level message node, m is positive integer, wherein, m first weight is calculated by following equation
Value：
Wherein, N (A, B) is represented from the first order information node A to institute
The first number of hops of the information node B in m second-level message node is stated, N (A, *) is represented from the first order information node
A to the first number of hops of the m second-level message node sum, M (A, B) represent from the first order information node A to
Described information node B the second number of hops, M (A, *) represents to believe from the first order information node A to the m second level
The sum of the second number of hops of node is ceased, α and β are weight proportion；
It is the related information that the user determines to recommend according to the m the first weighted values and default n recommendation information node,
Wherein, n is positive integer.
2. according to the method described in claim 1, it is characterised in that if the m is more than or equal to the n, the basis
The m the first weighted values are that the step of user determines the related information recommended is wrapped with default n recommendation information node
Include：
The m the first weighted values are ranked up, ranking results are obtained；
It is that the user recommends n related information node from the m second-level message node according to the ranking results.
It is described according to the m the 3. according to the method described in claim 1, it is characterised in that if the m is less than the n
One weighted value is that the step of user determines the related information recommended includes with default n recommendation information node：
The r third level information node associated with the m second-level message node is obtained, and calculates the r third level
R the second weighted values between information node and the m second-level message node, r is positive integer；
According to the m the first weighted values and the r the second weighted values obtain the m second-level message nodes with it is respective right
Multiple recommendations between the multiple third level information nodes answered；
The n-m information node that the user recommends is defined as according to the multiple recommendation.
4. method according to claim 3, it is characterised in that described according to the m the first weighted values and the r the
Two weighted values obtain multiple recommendations between the m second-level message node and each self-corresponding multiple third level information nodes
In the step of value, the multiple recommendation is calculated by following equation：
recvalue=δ * weight1*weight2,
Wherein, δ is decay factor, weight1Represent the first power of the first order information node and the second-level message node
Weight values, weight2Represent the second weighted value of the second-level message node and the third level information node.
5. according to any described methods of claim 1-4, it is characterised in that methods described also includes：
Calculate the title and the similarity of description of the first order information node and the m second-level message node；
Initial the between the first order information node and the m second-level message node is determined according to the similarity
One weighted value.
6. a kind of recommendation apparatus of related information, it is characterised in that described device includes：
First acquisition module, including：
First acquisition unit, for obtaining click number of hops of the user in first order information node, the click number of hops
Including the first number of hops and the second number of hops, the first order information node is the currently used information section of the user
Point, first number of hops redirects behavior by the active click of the user and obtained, and second number of hops passes through institute
State the click that user carries out to recommendation information and redirect behavior and obtain；
Second acquisition unit, believes for being obtained according to first number of hops and second number of hops with the first order
M the first weighted values of m associated second-level message node of node are ceased, m is positive integer；
Wherein, the second acquisition unit calculates m first weighted value by following equation：
Wherein, N (A, B) is represented from the first order information node A to institute
The first number of hops of the information node B in m second-level message node is stated, N (A, *) is represented from the first order information node
A to the first number of hops of the m second-level message node sum, M (A, B) represent from the first order information node A to
Described information node B the second number of hops, M (A, *) represents to believe from the first order information node A to the m second level
The sum of the second number of hops of node is ceased, α and β are weight proportion；
Recommending module, for being that the user determines to push away according to the m the first weighted values and default n recommendation information node
The related information recommended, wherein, n is positive integer.
7. device according to claim 6, it is characterised in that if the m is more than or equal to the n, the recommendation
Module includes：
First sequencing unit, for the m the first weighted values to be ranked up, obtains ranking results；
First recommendation unit, for being that the user recommends n from the m second-level message node according to the ranking results
Individual related information node.
8. device according to claim 6, it is characterised in that if the m is less than the n, the recommending module includes：
3rd acquiring unit, for obtaining the r third level information node associated with the m second-level message node, and
R the second weighted values between the r third level information node and the m second-level message node are calculated, r is just whole
Number；
4th acquiring unit, for obtaining described m second with the r the second weighted values according to the m the first weighted values
Multiple recommendations between level information node and each self-corresponding multiple third level information nodes；
Second recommendation unit, for being defined as the n-m information node that the user recommends according to the multiple recommendation.
9. device according to claim 8, it is characterised in that the 4th acquiring unit calculates described by following equation
Multiple recommendations：
recvalue=δ * weight1*weight2,
Wherein, δ is decay factor, weight1Represent the first power of the first order information node and the second-level message node
Weight values, weight2Represent the second weighted value of the second-level message node and the third level information node.
10. according to any described devices of claim 6-9, it is characterised in that described device also includes：
Computing module, for calculating title and description of the first order information node with the m second-level message node
Similarity；
Second acquisition module, for determining the first order information node and the m second-level message according to the similarity
The first initial weighted value between node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410610726.7A CN104361062B (en) | 2014-11-03 | 2014-11-03 | A kind of recommendation method and device of related information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410610726.7A CN104361062B (en) | 2014-11-03 | 2014-11-03 | A kind of recommendation method and device of related information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104361062A CN104361062A (en) | 2015-02-18 |
| CN104361062B true CN104361062B (en) | 2017-10-31 |
Family
ID=52528323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410610726.7A Active CN104361062B (en) | 2014-11-03 | 2014-11-03 | A kind of recommendation method and device of related information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104361062B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989154B (en) * | 2015-03-03 | 2020-07-14 | 华为技术有限公司 | Similarity measurement method and equipment |
| CN105488154A (en) * | 2015-11-28 | 2016-04-13 | 小米科技有限责任公司 | Theme application recommendation method and device |
| CN105867897A (en) * | 2015-12-07 | 2016-08-17 | 乐视网信息技术（北京）股份有限公司 | Page redirection analysis method and apparatus |
| CN105630766B (en) * | 2015-12-22 | 2018-11-06 | 北京奇虎科技有限公司 | Correlation calculations method and apparatus between more news |
| US10217025B2 (en) | 2015-12-22 | 2019-02-26 | Beijing Qihoo Technology Company Limited | Method and apparatus for determining relevance between news and for calculating relevance among multiple pieces of news |
| CN106919582A (en) * | 2015-12-24 | 2017-07-04 | 阿里巴巴集团控股有限公司 | The association of network articles and related information statistical method and device |
| CN106686063A (en) * | 2016-12-07 | 2017-05-17 | 乐视控股（北京）有限公司 | Information recommendation method and apparatus, and electronic device |
| CN108737902B (en) * | 2018-05-22 | 2021-03-02 | 北京奇艺世纪科技有限公司 | Associated video updating method and device and electronic equipment |
| CN117194459A (en) * | 2023-09-22 | 2023-12-08 | 天翼爱音乐文化科技有限公司 | Operation and maintenance knowledge base updating method, system, device and medium based on operation and maintenance event |
| CN117058061B (en) * | 2023-10-12 | 2024-01-30 | 广东工业大学 | Multi-focus image fusion method and related device based on target detection |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101079062A (en) * | 2007-06-22 | 2007-11-28 | 腾讯科技（深圳）有限公司 | Web page importance evaluation method and system |
| CN101520878A (en) * | 2009-04-03 | 2009-09-02 | 华为技术有限公司 | Method, device and system for pushing advertisements to users |
| CN102054004A (en) * | 2009-11-04 | 2011-05-11 | 清华大学 | Webpage recommendation method and device adopting same |
| CN103530339A (en) * | 2013-10-08 | 2014-01-22 | 北京百度网讯科技有限公司 | Mobile application information push method and device |
| CN104123321A (en) * | 2013-04-28 | 2014-10-29 | 百度在线网络技术（北京）有限公司 | Method and device for determining recommended pictures |
-
2014
- 2014-11-03 CN CN201410610726.7A patent/CN104361062B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101079062A (en) * | 2007-06-22 | 2007-11-28 | 腾讯科技（深圳）有限公司 | Web page importance evaluation method and system |
| CN101520878A (en) * | 2009-04-03 | 2009-09-02 | 华为技术有限公司 | Method, device and system for pushing advertisements to users |
| CN102054004A (en) * | 2009-11-04 | 2011-05-11 | 清华大学 | Webpage recommendation method and device adopting same |
| CN104123321A (en) * | 2013-04-28 | 2014-10-29 | 百度在线网络技术（北京）有限公司 | Method and device for determining recommended pictures |
| CN103530339A (en) * | 2013-10-08 | 2014-01-22 | 北京百度网讯科技有限公司 | Mobile application information push method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104361062A (en) | 2015-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104361062B (en) | A kind of recommendation method and device of related information | |
| CN103514304B (en) | Project recommendation method and device | |
| CN104462573B (en) | A kind of video search result methods of exhibiting and device | |
| JP5432243B2 (en) | Send and react to media object queries | |
| US8589434B2 (en) | Recommendations based on topic clusters | |
| CN104331459B (en) | A kind of network resource recommended method and device based on on-line study | |
| CN105005582B (en) | The recommendation method and device of multimedia messages | |
| CN104317835B (en) | The new user of video terminal recommends method | |
| CN103678329B (en) | Recommend method and device | |
| Carullo et al. | A triadic closure and homophily-based recommendation system for online social networks | |
| TW201717071A (en) | Recommendation method and device | |
| CN109408734A (en) | A kind of collaborative filtering recommending method of fuse information Entropy conformability degree and dynamic trust | |
| CN108287864A (en) | A kind of interest group division methods, device, medium and computing device | |
| US11556851B2 (en) | Establishing a communication session between client terminals of users of a social network selected using a machine learning model | |
| Boukhris et al. | A novel personalized academic venue hybrid recommender | |
| Tiroshi et al. | Domain ranking for cross domain collaborative filtering | |
| CN110209946A (en) | Based on social and community Products Show method, system and storage medium | |
| US20150248720A1 (en) | Recommendation engine | |
| Park et al. | Fast Collaborative Filtering with a k-nearest neighbor graph | |
| US8825698B1 (en) | Showing prominent users for information retrieval requests | |
| US9514194B1 (en) | Website duration performance based on category durations | |
| US8745074B1 (en) | Method and system for evaluating content via a computer network | |
| CN109241120A (en) | A kind of user's recommended method and device | |
| Christensen et al. | A hybrid approach for group profiling in recommender systems | |
| US20130085867A1 (en) | Niche Keyword Recommendation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |