CN105335363A - Object pushing method and system - Google Patents
Object pushing method and system Download PDFInfo
- Publication number
- CN105335363A CN105335363A CN201410231465.8A CN201410231465A CN105335363A CN 105335363 A CN105335363 A CN 105335363A CN 201410231465 A CN201410231465 A CN 201410231465A CN 105335363 A CN105335363 A CN 105335363A
- Authority
- CN
- China
- Prior art keywords
- candidate target
- weighted value
- described candidate
- targets
- candidate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
Embodiments of the present invention provide an object pushing method and system. The method comprises: obtaining a candidate object, wherein the candidate object comprises an input object of a client, a matching object obtained by performing retrieval according to the input object of the client or a recommended object obtained according to a click history record of the client; obtaining a weight absorption rate of the candidate object, wherein the weight absorption rate of the candidate object is a parameter value that is in an inversely proportional relationship with the number of other candidate objects that are directly associated with the candidate object, and the other candidate objects that are directly associated with the candidate object are other candidate objects whose correlations with the candidate object are greater than a preset correlation threshold; according to the weight absorption rate of the candidate object, obtaining a first weight value of the candidate object; and according to the first weight value and the candidate object, obtaining a to-be-pushed target object. The technical scheme provided by the embodiments of the present invention can improve the accuracy of information retrieval and the retrieval efficiency.
Description
[technical field]
The present invention relates to technical field of information retrieval, particularly relate to a kind of Object Push method and system.
[background technology]
PageRank algorithm is the proprietary algorithm of Google, for weighing the significance level of particular webpage relative to other webpages in search engine index, then sorts to webpage according to the significance level of webpage.Personalized PageRank, when calculating the weighted value of webpage, can consider concrete user, or assessing the weighted value of other webpages for a webpage, thus can meet the demand of a lot of practical application.
But, PageRank algorithm application is when search engine retrieving webpage, there is following defect: the weighted value of self can be passed to the webpage of other classifications by other webpage of same class, be equivalent to search engine retrieve in multiple different classes of webpage, therefore the accuracy of result for retrieval is lower and recall precision is lower.
[summary of the invention]
In view of this, embodiments provide a kind of Object Push method and system, the accuracy and the recall precision that improve information retrieval can be realized.
First aspect, embodiments provides a kind of Object Push method, comprising:
Obtain candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client;
Obtain the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target;
According to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
According to described first weighted value and described candidate target, obtain destination object to be pushed.
In the first possible implementation of first aspect, the described weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target, comprising:
Second weighted value of described candidate target and the screening threshold value preset are compared; Described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation;
If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, and stops transmitting to make the 3rd weighted value of described candidate target between the candidate target of association.
In conjunction with the first possible implementation of first aspect or first aspect, in the implementation that the second of first aspect is possible, the described weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target, comprising:
According to the number with other candidate targets of described candidate target direct correlation, and utilize following formula, obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; Described
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number of other candidate targets of described candidate target i direct correlation.
In conjunction with first aspect or the first possible implementation of first aspect or the possible implementation of the second of first aspect, in the third possible implementation of first aspect, described according to described first weighted value and described candidate target, obtain destination object to be pushed, comprise: described first weighted value and the weight threshold preset are compared, described first weighted value is greater than the described candidate target of described weight threshold as destination object described to be pushed.
In conjunction with the first or the second or the third possible implementation of first aspect or first aspect, in the 4th kind of possible implementation of first aspect, described method also comprises:
According to the order that described first weighted value is descending, described destination object is sorted, to obtain ranking results;
Push described ranking results.
Second aspect, embodiments provides a kind of Object Push system, comprising:
Object acquisition unit, for obtaining candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client;
First processing unit, for obtaining the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target;
Second processing unit, for the weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target;
Object screening unit, for according to described first weighted value and described candidate target, obtains destination object to be pushed.
In the first possible implementation of second aspect, described second processing unit specifically for:
Second weighted value of described candidate target and the screening threshold value preset are compared; Described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation;
If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, and stops transmitting to make the 3rd weighted value of described candidate target between the candidate target of association.
In conjunction with the first possible implementation of second aspect or second aspect, in the implementation that the second of second aspect is possible, described second processing unit specifically for:
According to the number of other candidate targets of described candidate target direct correlation, and utilize following formula, obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; Described
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number of other candidate targets of described candidate target i direct correlation.
In conjunction with second aspect or the first possible implementation of second aspect or the possible implementation of the second of second aspect, in the third possible implementation of second aspect, described object screening unit specifically for: described first weighted value and the weight threshold preset are compared, described first weighted value are greater than the described candidate target of described weight threshold as described destination object.
In conjunction with the first or the second or the third possible implementation of second aspect or second aspect, in the 4th kind of possible implementation of second aspect, described system also comprises: object output unit, for according to the descending order of described first weighted value, described destination object is sorted, to obtain ranking results; Push described ranking results.
As can be seen from the above technical solutions, the embodiment of the present invention has following beneficial effect:
In the embodiment of the present invention, the number of the weight absorptivity of candidate target and other candidate targets of described candidate target direct correlation is inversely prroportional relationship, therefore, according to the principle that the number of other candidate targets of the candidate target direct correlation at edge in incidence relation is less, the weight absorptivity of the candidate target at edge is larger, therefore, it is possible to ensure that weighted value is when arriving the edge candidate target of certain classification to a certain extent, only less weighted value is passed to the candidate target of other classifications, can ensure that weighted value transmits between other different candidate target of same class, like this, even if the candidate target of certain classification delivers weighted value to other classifications, the candidate target of other classifications also can be less because of the weighted value obtained, and can not by as destination object to be pushed, thus can ensure to a certain extent to retrieve in the generic inside of input object, to obtain destination object to be pushed, and then the cluster of information retrieval can be realized, accuracy and the recall precision of result for retrieval can be improved.
[accompanying drawing explanation]
In order to be illustrated more clearly in the technical scheme of the embodiment of the present invention, be briefly described to the accompanying drawing used required in embodiment below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
Fig. 1 is the schematic flow sheet of the Object Push method that the embodiment of the present invention provides;
Fig. 2 (a) ~ Fig. 2 (c) is the distribution schematic diagram of the weighted value belonging to other candidate target of same class in the embodiment of the present invention;
Fig. 3 is the distribution schematic diagram of the candidate target that the embodiment of the present invention provides;
Fig. 4 is the transfer probability schematic diagram of weight between the candidate target that provides of the embodiment of the present invention;
Fig. 5 is the distribution schematic diagram of the result for retrieval of prior art;
Fig. 6 is the distribution schematic diagram of the result for retrieval that the embodiment of the present invention provides;
Fig. 7 (a) ~ Fig. 7 (d) is the first schematic diagram of the image retrieval that the embodiment of the present invention provides;
Fig. 8 (a) ~ Fig. 8 (c) is the second schematic diagram of the image retrieval that the embodiment of the present invention provides;
Fig. 9 (a) ~ Fig. 9 (c) is the second schematic diagram of the image retrieval that the embodiment of the present invention provides;
Figure 10 (a) ~ Figure 10 (d) is the schematic diagram of result for retrieval under the different weighted average density that provide of the embodiment of the present invention;
Figure 11 is the functional block diagram of the Object Push system that the embodiment of the present invention provides;
Figure 12 is the structural representation of the Object Push system that the embodiment of the present invention provides.
[embodiment]
Technical scheme for a better understanding of the present invention, is described in detail the embodiment of the present invention below in conjunction with accompanying drawing.
Should be clear and definite, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making other embodiments all obtained under creative work prerequisite, belong to the scope of protection of the invention.
Retrieve in mass data (as webpage and picture), obtain and push in the process of result for retrieval, data volume in database is often larger, but the sub-fraction data in the just mass data relevant to retrieval, other such as relevant to webpage webpages are the sub-fraction of whole internet data, if therefore only can consider the sub-fraction data in mass data in the process of retrieval, will greatly improve the efficiency of retrieval and pushed information.
The embodiment of the present invention provides a kind of Object Push method, please refer to Fig. 1, the schematic flow sheet of its Object Push method provided for the embodiment of the present invention, and as shown in the figure, the method comprises the following steps:
Step 101, obtain candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client.
Concrete, client obtains candidate target, and described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client.
Preferably, described client can comprise browser, application program or image retrieval program etc.
Preferably, the input object of described client comprises the pixel that keyword or described client that described client inputs input.
Preferably, can retrieve in a database according to described input object, to obtain the match objects of described input object.Described match objects can comprise webpage, application message, Item Information or pixel.Wherein, described match objects comprises and other objects of described input object direct correlation or other objects with described input object indirect association; Be other objects being greater than default dependent thresholds with the degree of correlation of described input object with other objects of described input object direct correlation, be by other objects with described input object direct correlation with other objects of described input object indirect association, the object associated with described input object.Such as, input object A, match objects B and match objects C, wherein, input object A and match objects B direct correlation, match objects B and match objects C direct correlation, then match objects C and input object A indirect association.
Preferably, described recommended comprises Item Information or application message.
Illustrate, first, within nearest a period of time browser historical record, obtain the webpage that in nearest a period of time, user clicks in a browser; Then add up the number of clicks of each webpage, and according to the descending order of number of clicks, the webpage that user clicks is sorted; Finally, by least one forward for rank in ranking results webpage, as the recommended that this browser is corresponding.The incidence relation between webpage and webpage is preserved in database, as webpage A points to webpage B etc., then according to the webpage that browser is corresponding, retrieve in a database, the webpage with this webpage direct correlation can be obtained, and then the webpage associated by webpage of acquisition and this direct correlation, finally can obtain direct correlation webpage and the indirect association webpage of the webpage corresponding with browser.
Illustrate, user, when mobile terminal has just opened the client such as Taobao, Amazon, can push object information to user, as recommended certain article, then according to historical record corresponding to client, can determine the article that nearest a period of time of user is bought; Then add up the number of clicks of each article, and sort according to the order that number of clicks is descending, the article that user clicks are sorted; Finally, by least one forward for rank in ranking results article, as the recommended that this client is corresponding.Incidence relation between article and article is preserved in database, as article A and article B is bought simultaneously, article A and article B attribute similarity/mutually equal, then according to the Item Information that client is corresponding, retrieve in a database, the article with this article direct correlation can be obtained, and then the article associated by article of acquisition and this direct correlation, finally can obtain direct correlation article and the indirect association article of the article corresponding with client.
It should be noted that, the mode of above-mentioned acquisition candidate target, can obtain some input objects, improves the diversified degree of input object, realizes input distribution, ensures the diversity of result for retrieval.
Step 102, obtains the weight absorptivity of described candidate target, and the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target.
Concrete, in the embodiment of the present invention, also need for candidate target each at least one candidate target calculates corresponding first weighted value, and before calculating the first weighted value, need the weight absorptivity first determining each candidate target.
Wherein, the weight absorptivity of described candidate target refers to the number of other candidate targets of described candidate target direct correlation is the parameter value of inversely prroportional relationship.
Preferably, following formula can be utilized to obtain weight absorptivity:
In this formula, described α represents default parameter value; Described d (i) represents the number with other candidate targets of described candidate target i direct correlation; Can find out according to this formula, larger with the value of number d (i) of other candidate targets of described candidate target direct correlation, the value of weight absorptivity is less, less with the value of number d (i) of other candidate targets of described candidate target direct correlation, the value of weight absorptivity is larger, therefore, the number of the weight absorptivity of described candidate target and other candidate targets of described candidate target direct correlation is inversely prroportional relationship.
Please refer to Fig. 2 (a), it is for belonging to the distribution schematic diagram of the weighted value of other candidate target of same class in the embodiment of the present invention, as shown in Fig. 2 (a), a candidate target can be associated with other candidate targets, the relational structure shown in Fig. 2 (a) can be formed between candidate target, in the embodiment of the present invention, the number of the weight absorptivity of described candidate target and other candidate targets of described candidate target direct correlation is inversely prroportional relationship, as shown in Fig. 2 (a), the number of other candidate targets of candidate target 1 direct correlation is 3, i.e. candidate target 2, candidate target 3 and candidate target 4 respectively with candidate target 1 direct correlation, in like manner, the number of other candidate targets of candidate target 2 direct correlation is 2, namely candidate target 5 and candidate target 6 respectively with candidate target 2 direct correlation.Wherein, described relational structure can comprise the incidence relation between described candidate target i and other candidate targets; Preferably, this relational structure can show as reticulate texture or tree etc., does not repeat them here.
Please refer to Fig. 3, the distribution schematic diagram of its candidate target provided for the embodiment of the present invention, as shown in the figure, for candidate target, number the closer to other candidate targets of the candidate target direct correlation at relational structure center is larger, and the number the closer to other candidate targets of the candidate target direct correlation at relational structure edge is less.
Step 103, according to the weight absorptivity of described candidate target, obtains the first weighted value of described candidate target.
Concrete, each candidate target can have a weighted value inputted, and is called the second weighted value, and described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation, as shown in Fig. 2 (a), input object, the second weighted value as both candidate nodes 1 can be pre-assigned, and affiliated partner, as candidate target 2, second weighted value of candidate target 3 grade can be the weighted value that other candidate targets associated with self pass to oneself, each candidate target can the second weighted value of absorption portion, then remaining second weighted value is passed to other candidate targets with self direct correlation by candidate target, other candidate targets again absorption portion pass to self the second weighted value, then remaining second weighted value is passed to other candidate targets with self direct correlation again, by that analogy, that is, second weighted value can be received for each candidate target, then obtain the first weighted value of candidate target, finally the part in the second weighted value except the first weighted value be passed to other candidate targets of direct correlation.
As shown in Fig. 2 (a) ~ Fig. 2 (c), if the second weighted value to be regarded as the input flow rate of candidate target, so this input flow rate flows along the incidence relation between candidate target, as shown in Fig. 2 (a), candidate target 1 has input flow rate, as filled the circle of shade, as shown in Fig. 2 (b), candidate target 1 absorbs the input flow rate of part, then the input flow rate of remainder is passed to the candidate target 2 with self direct correlation, candidate target 3 and candidate target 4, as shown in Fig. 2 (c), the input flow rate that candidate target 2 absorption portion obtains, then remaining input flow rate is passed to again and the candidate target 5 of self direct correlation and candidate target 6, operation and the candidate target 2 of candidate target 3 and candidate target 4 are similar, here repeat no more.
In the embodiment of the present invention, whenever there being candidate target to obtain the second weighted value, just the second weighted value of candidate target and the screening threshold value preset are compared; If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target; If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, stop between the candidate target of association to make the 3rd weighted value of described candidate target transmitting, that is, input flow rate will stop transmitting between the candidate target of association.Here, by arranging screening threshold value, can prevent weighted value from unrestrictedly transmitting in relational structure, thus can reduce the calculated amount of weighted value, such as, described screening threshold value can be 0.001.In addition, if object set comprises at least two candidate targets, when the first weighted value of calculated candidate object, can the first weighted value of simultaneously calculated candidate object, also can the first weighted value of calculated candidate object successively.
In the embodiment of the present invention, the second weighted value of the affiliated partner in candidate target passes to this affiliated partner with other candidate targets of affiliated partner direct correlation, and the second weighted value of input object in candidate target is pre-assigned.If there are at least two input objects, because the second weighted value of input object is pre-assigned, the method of then distributing the second weighted value for each input object can be mean allocation, namely the second weighted value of each input object is identical, also the second weighted value can be distributed according to the rank of input object, second weighted value of the input object that rank is forward is comparatively large, and the second weighted value of the input object ranked behind is less, does not specifically limit in the embodiment of the present invention.
Preferably, according to the weight absorptivity of described candidate target, the method obtaining the first weighted value of described candidate target can comprise:
According to the number with other candidate targets of described candidate target direct correlation, and utilize following formula (1), obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation,
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number with other candidate targets of described candidate target i direct correlation.Wherein, basic weighted value S (i) of the described candidate target i obtained in advance can be understood as this candidate target i self weighted value before acquisition second weighted value r (i), like this, after acquisition second weighted value, basic weighted value S (i) current to self is needed to upgrade, to obtain S (i) '.
Be understandable that, the number of the weight absorptivity of described candidate target and other candidate targets of described candidate target direct correlation is inversely prroportional relationship, that is: the second weighted value is looked as a whole, the numerical value of number d (i) of other candidate targets of described candidate target i direct correlation is larger, the weight absorptivity of candidate target i is less, thus the second weighted value that candidate target i absorbs is less, the second weighted value being left not absorbed by candidate target i is larger, be left not passed to other candidate targets of direct correlation by candidate target i by the second weighted value that candidate target i absorbs, therefore to pass to the second weighted value of other candidate targets of direct correlation larger for candidate target i, in like manner, the numerical value of number d (i) of other candidate targets of described candidate target i direct correlation is less, the weight absorptivity of candidate target i is larger, thus the second weighted value that candidate target i absorbs is larger, the second weighted value being left not absorbed by candidate target i is less, remaining will do not passed to other candidate targets of direct correlation by candidate target i by the second weighted value that candidate target i absorbs, therefore to pass to the second weighted value of other candidate targets of direct correlation less for candidate target i.As shown in Figure 3, for candidate target, the closer to the weight absorptivity of the candidate target at relational structure center
less, the closer to the weight absorptivity of the candidate target at relational structure edge
larger, therefore, ensure that from the process of the extrorse candidate target of the candidate target at relational structure center, the weight absorptivity of candidate target i is increasing, residue second weighted value that the candidate target at edge transmits can be fewer, and belong to other candidate target of same class and form a relational structure, thus can ensure that weighted value is retained in each candidate target of current class as far as possible, once weighted value arrives the candidate target at relational structure edge, also can be larger because of the weight absorptivity of the candidate target at edge, make the weighted value of the candidate target being delivered to other classifications very little, therefore, ensure that weighted value can be retained on the candidate target of current class to a certain extent, the classification of retrieval can be greatly reduced, improve accuracy and the recall precision of result for retrieval.
Please refer to Fig. 4, the transfer probability schematic diagram of weight between its candidate target provided for the embodiment of the present invention, as shown in the figure, for candidate target k, candidate target i and candidate target j, weighted value is transferred to by candidate target k the probability of happening that candidate target i transfers to candidate target j again and is met following second order Markov chain model:
T, second weighted value is positioned at candidate target k, second weighted value of candidate target k reserve part, and at t+1 moment candidate target k, remaining second weighted value is passed to candidate target i, then, candidate target i retains the part in remaining second weighted value again, and at t+2 moment candidate target i, the remainder in remaining second weighted value is passed to candidate target j, the probability of happening of the transmission event of above-mentioned second weighted value can as shown in above-mentioned second order Markov chain model, as i=j and i=k time, represent to only have candidate target k, i.e. t, t+1 moment and t+2 moment, second weighted value all rests on candidate target k all the time, therefore, the probability of happening of the transmission event of above-mentioned second weighted value is 1, as i ≠ j and i=k time, t and t+1 moment, the second weighted value rests on candidate target k, therefore, second weighted value can not be passed to candidate target j by candidate target k, and therefore in the t+2 moment, candidate target k is 0 to the probability of happening of the event of the second weighted value of candidate target j transmitting portions, as i ≠ k, no matter as i=j or i ≠ j, in the t+1 moment, the second weighted value can be passed to candidate target i by candidate target k, and in the t+2 moment, the probability of happening that the second weighted value can be passed to the event of candidate target j by candidate target i is P
ij.
Preferably, utilize following formula, obtain candidate target i pass to weighted value r (j) of other candidate targets j of this candidate target i direct correlation ':
r(j)′＝r(j)+r(i)w(i,j)/(α+d(i))
Wherein, described r (j) represents the basic weighted value of candidate target j; Described r (i) represents second weighted value of described candidate target i; Described α represents default parameter value; Described d (i) represents the number with other candidate targets of described candidate target i direct correlation.
It should be noted that, in the candidate target of a classification, if relational structure G=is (V, W), V represents candidate target set, W=w (i, j) represent the degree of association matrix of candidate target i and candidate target j, this degree of association matrix w (i, j) is for representing that candidate target i and candidate target j's associates tight ness rating, if candidate target i and candidate target j direct correlation, then w (i, j) equals 1, if candidate target i and candidate target j is not direct correlation, w (i, j) equals 0.The d (i) of candidate target i equals the cumulative sum of w (i, j).
Or according to the weight absorptivity of described candidate target, utilize following formula (2), the method obtaining the first weighted value of described candidate target can comprise:
S(i)′＝[α(α×I+L)
-1r
0]
i
Wherein, described α represents default parameter value, described r
0represent described candidate target i self basic weighted value before acquisition second weighted value, described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; I representation unit matrix; L=D-W, L represents the Laplacian Matrix of relational structure, wherein, D=diag (d (1) ..., d (n)), wherein, D=diag () represents diagonal matrix, and d (i) represents the number with other candidate targets of described candidate target i direct correlation, i=1,2 ..., n; N represents the total number of candidate target.
It should be noted that, in formula (2), [α (α × I+L)
-1r
0] represent the matrix that the first weighted value of all candidate targets forms, [α (α × I+L)
-1r
0]
ijust represent the first weighted value of i-th candidate target in this matrix.
Formula (2) is applicable to the situation of candidate target negligible amounts in relational structure, if candidate target quantity is more in relational structure, also utilizes formula (2) to calculate, computational throughput will be caused larger; If therefore in relational structure, candidate target quantity is more, can the first weighted value of Selection utilization formula (1) calculated candidate object.
Step 104, according to described first weighted value and described candidate target, obtains destination object to be pushed.
Concrete, after the first weighted value obtaining candidate target, described first weighted value and the weight threshold preset is needed to compare, described first weighted value is greater than the described candidate target of described weight threshold as destination object described to be pushed, screens out the candidate target that the first weighted value is less than or equal to weight threshold.
Optionally, according to the descending order of described first weighted value, described destination object can be sorted, to obtain ranking results; Push described ranking results.
Please refer to Fig. 5, it is the distribution schematic diagram of the result for retrieval of prior art, as shown in the figure, point indicated by circle 1 represents input object, all points represent the result for retrieval of output, i.e. above-mentioned destination object, but, point indicated by circle 2 is the result for retrieval the closest with input object incidence relation, is the result for retrieval that user needs most; In circle 2, circle 3 and circle 4, weighted value the closer to the point of centre is higher, weighted value the closer to the point at edge is lower, and, in all result for retrieval, wherein more lean on the incidence relation between the point of right-hand component and input object smaller, if as result for retrieval, will the accuracy of result for retrieval be reduced.
Please refer to Fig. 6, the distribution schematic diagram of its result for retrieval provided for the embodiment of the present invention, as shown in the figure, points all in Fig. 6 represents the result for retrieval of output, i.e. above-mentioned destination object, transverse axis in Fig. 6 represents the number of candidate target, and the vertical pivot as candidate target 1 ~ candidate target 300, candidate target 300 ~ candidate target 600, candidate target 600 to candidate target 900, Fig. 6 represents the size of the second weighted value that candidate target absorbs.As shown in Figure 6, the second weighted value that candidate target 1 ~ candidate target 300 absorbs is larger, represent that candidate target ~ candidate target 300 is and input object relation candidate target the most closely, therefore, can find out that candidate target 100 ~ candidate target 300 absorbs the second most weighted values, make the second weighted value exporting to other classifications fewer, the search method that therefore embodiment of the present invention provides can embody the Clustering Retrieval of candidate target.
In addition, please refer to Fig. 7 (a) ~ Fig. 7 (d), first schematic diagram of its image retrieval provided for the embodiment of the present invention, Fig. 7 (a) represents original image, Fig. 7 (b) represents input object, coupling retrieval can be carried out in all pixels in Fig. 7 (a) according to input object, input object is the some scattered pixel in bird pattern, the schematic diagram of candidate target when Fig. 7 (c) represents that weighted value is balanced in relational structure, probably can find out the profile of result for retrieval, Fig. 7 (d) represents the final destination object needing to push.
Please refer to Fig. 8 (a) ~ Fig. 8 (c), second schematic diagram of its image retrieval provided for the embodiment of the present invention, Fig. 8 (a) represents original image, wherein cross represents input object, retrieval can mate in Fig. 8 (a) according to this input object, input object is a pixel, the schematic diagram of candidate target when Fig. 8 (b) represents that weighted value is balanced in relational structure, probably can find out the profile of result for retrieval, Fig. 8 (c) represents destination object.
Please refer to Fig. 9 (a) ~ Fig. 9 (c), second schematic diagram of its image retrieval provided for the embodiment of the present invention, Fig. 9 (a) represents original image, Fig. 9 (b) represents input object, retrieval can mate in Fig. 9 (a) according to this input object, the quantity of input object is more, and cause image ratio fuzzyyer, Fig. 9 (c) represents destination object.Please refer to Figure 10 (a) ~ Figure 10 (d), the schematic diagram of result for retrieval under its different weighted average density provided for the embodiment of the present invention, Figure 10 (a) ~ Figure 10 (d) can be good at the robustness of the technical scheme embodying the embodiment of the present invention, using 1 of weighted mean times as threshold value, candidate target is screened, obtain the destination object shown in Figure 10 (a), using 1.5 of weighted mean times as threshold value, candidate target is screened, obtain the destination object shown in Figure 10 (b), using 2 of weighted mean times as threshold value, candidate target is screened, obtain the destination object shown in Figure 10 (c), using 2.5 of weighted mean times as threshold value, candidate target is screened, obtain the destination object shown in Figure 10 (d), here, although the input of retrieval is blurred picture, but still can obtain clearly target image, and along with the continuous change of threshold value large, the target image obtained is more and more clear, the quantity of the destination object obtained is fewer and feweri, the weighted value of the pixel (pixel in Figure 10 (d)) of the target image inside obtained significantly is greater than weighted mean.
The embodiment of the present invention provides the device embodiment realizing each step and method in said method embodiment further.
Please refer to Figure 11, the functional block diagram of its Object Push system provided for the embodiment of the present invention.As shown in the figure, this system comprises:
Object retrieval unit 11, for obtaining candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client;
First processing unit 12, for obtaining the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target;
Second processing unit 13, for the weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target;
Object screening unit 14, for according to described first weighted value and described candidate target, obtains destination object to be pushed.
Preferably, described second processing unit 13 specifically for:
Second weighted value of described candidate target and the screening threshold value preset are compared; Described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation;
If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, and stops transmitting to make the 3rd weighted value of described candidate target between the candidate target of association.
Preferably, described second processing unit 13 specifically for:
According to the number of other candidate targets of described candidate target direct correlation, and utilize following formula, obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; Described
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number of other candidate targets of described candidate target i direct correlation.
Preferably, described object screening unit 14, specifically for: described first weighted value and the weight threshold preset are compared, described first weighted value is greater than the described candidate target of described weight threshold as described destination object.
Preferably, described system also comprises:
Object output unit 15, for according to the descending order of described first weighted value, sorts to described destination object, to obtain ranking results; Push described ranking results.
Please refer to Figure 12, the structural representation of its Object Push system provided for the embodiment of the present invention.As shown in the figure, this equipment comprises:
Storer 20, for storing one or more groups program code;
Processor 21, be coupled with storer 20, for calling the program code stored in storer 20, to perform the method shown in following Fig. 1, specifically comprise: obtain candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client; Obtain the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target; According to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target; According to described first weighted value and described candidate target, obtain destination object to be pushed.
Because the processor in the present embodiment can perform the method shown in Fig. 1, the part that the present embodiment is not described in detail, can with reference to the related description to Fig. 1.
The technical scheme of the embodiment of the present invention has following beneficial effect:
The number of the weight absorptivity of candidate target and other candidate targets of described candidate target direct correlation is inversely prroportional relationship, therefore, according to the principle that the number of other candidate targets of the candidate target direct correlation at edge in incidence relation is less, the weight absorptivity of the candidate target at edge is larger, therefore, it is possible to ensure that weighted value is when arriving the edge candidate target of certain classification to a certain extent, only less weighted value is passed to the candidate target of other classifications, can ensure that weighted value transmits between other different candidate target of same class, like this, even if the candidate target of certain classification delivers weighted value to other classifications, the candidate target of other analogies also can be less because of the weighted value obtained, and can not by as destination object to be pushed, thus can ensure to a certain extent to retrieve in the generic inside of input object, to obtain destination object to be pushed, and then the cluster of information retrieval can be realized, accuracy and the recall precision of result for retrieval can be improved.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within the scope of protection of the invention.
Claims (10)
1. an Object Push method, is characterized in that, described method comprises:
Obtain candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client;
Obtain the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target;
According to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
According to described first weighted value and described candidate target, obtain destination object to be pushed.
2. method according to claim 1, is characterized in that, the described weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target, comprising:
Second weighted value of described candidate target and the screening threshold value preset are compared; Described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation;
If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, and stops transmitting to make the 3rd weighted value of described candidate target between the candidate target of association.
3. method according to claim 1 and 2, is characterized in that, the described weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target, comprising:
According to the number with other candidate targets of described candidate target direct correlation, and utilize following formula, obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; Described
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number of other candidate targets of described candidate target i direct correlation.
4. according to the method in any one of claims 1 to 3, it is characterized in that, described according to described first weighted value and described candidate target, obtain destination object to be pushed, comprising:
Described first weighted value and the weight threshold preset are compared, described first weighted value is greater than the described candidate target of described weight threshold as destination object described to be pushed.
5. the method according to any one of Claims 1-4, is characterized in that, described method also comprises:
According to the order that described first weighted value is descending, described destination object is sorted, to obtain ranking results;
Push described ranking results.
6. an Object Push system, is characterized in that, described system comprises:
Object acquisition unit, for obtaining candidate target, described candidate target comprises the input object of client, the match objects carrying out retrieving rear acquisition according to the input object of described client or the recommended obtained according to the click historical record of described client;
First processing unit, for obtaining the weight absorptivity of described candidate target, the weight absorptivity of described candidate target is be the parameter value of inversely prroportional relationship with the number of other candidate targets of described candidate target direct correlation; Wherein, with other candidate targets of described candidate target direct correlation be other candidate targets being greater than default dependent thresholds with the degree of correlation of described candidate target;
Second processing unit, for the weight absorptivity according to described candidate target, obtains the first weighted value of described candidate target;
Object screening unit, for according to described first weighted value and described candidate target, obtains destination object to be pushed.
7. system according to claim 6, is characterized in that, described second processing unit specifically for:
Second weighted value of described candidate target and the screening threshold value preset are compared; Described second weighted value is the weighted value passing to described candidate target with other candidate targets of described candidate target direct correlation;
If the second weighted value of described candidate target is greater than described screening threshold value, according to the weight absorptivity of described candidate target, obtain the first weighted value of described candidate target;
If the second weighted value of described candidate target is less than or equal to described screening threshold value, described candidate target does not transmit the 3rd weighted value to other candidate targets of association, described 3rd weighted value equals the difference of the second weighted value that the second weighted value of described candidate target and described candidate target absorb, and stops transmitting to make the 3rd weighted value of described candidate target between the candidate target of association.
8. the system according to claim 6 or 7, is characterized in that, described second processing unit specifically for:
According to the number of other candidate targets of described candidate target direct correlation, and utilize following formula, obtain the first weighted value of described candidate target:
Wherein, first weighted value of the described candidate target i of described S (i) ' represent; The basic weighted value of the described candidate target i that the expression of described S (i) obtains in advance; Described α represents default parameter value; Described
represent described weight absorptivity, described r (i) represents second weighted value of described candidate target i, and described second weighted value is the weighted value passing to described candidate target i with other candidate targets of described candidate target i direct correlation; Described
represent the second weighted value that described candidate target i absorbs; Described d (i) represents the number of other candidate targets of described candidate target i direct correlation.
9. the system according to any one of claim 6 to 8, is characterized in that, described object screening unit specifically for:
Described first weighted value and the weight threshold preset are compared, described first weighted value is greater than the described candidate target of described weight threshold as described destination object.
10. the system according to any one of claim 6 to 9, is characterized in that, described system also comprises:
Object output unit, for according to the descending order of described first weighted value, sorts to described destination object, to obtain ranking results; Push described ranking results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410231465.8A CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410231465.8A CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105335363A true CN105335363A (en) | 2016-02-17 |
| CN105335363B CN105335363B (en) | 2018-12-07 |
Family
ID=55285907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410231465.8A Active CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105335363B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107545450A (en) * | 2016-06-27 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | A kind of Transaction Information method for pushing and device |
| CN107945079A (en) * | 2016-10-12 | 2018-04-20 | 普天信息技术有限公司 | A kind of poverty alleviation object selection method and device |
| CN108108419A (en) * | 2017-12-15 | 2018-06-01 | 百度在线网络技术（北京）有限公司 | A kind of information recommendation method, device, equipment and medium |
| CN110413751A (en) * | 2019-07-18 | 2019-11-05 | 腾讯科技（深圳）有限公司 | Drug searching method, device, terminal device and storage medium |
| CN111738754A (en) * | 2020-06-01 | 2020-10-02 | 上海晶赞融宣科技有限公司 | Object recommendation method and device, storage medium and computer equipment |
| TWI739359B (en) * | 2019-03-28 | 2021-09-11 | 南韓商韓領有限公司 | Computer-implemented system, computer-implemented method for arranging hyperlinks on a graphical user-interface and non-transitory computer-readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1996299A (en) * | 2006-12-12 | 2007-07-11 | 孙斌 | Ranking method for web page and web site |
| CN101105799A (en) * | 2007-08-02 | 2008-01-16 | 华中科技大学 | Method for evaluating file importance degree |
| CN101770521A (en) * | 2010-03-11 | 2010-07-07 | 东华大学 | Focusing relevancy ordering method for vertical search engine |
| US20110270847A1 (en) * | 2010-05-01 | 2011-11-03 | Adam Etkin | Method and system for appraising the extent to which a publication has been reviewed by means of a peer-review process |
| CN102270246A (en) * | 2011-09-08 | 2011-12-07 | 胡辉 | Method for calculating importance of web page |
| CN103559258A (en) * | 2013-11-04 | 2014-02-05 | 同济大学 | Webpage ranking method based on cloud computation |
-
2014
- 2014-05-28 CN CN201410231465.8A patent/CN105335363B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1996299A (en) * | 2006-12-12 | 2007-07-11 | 孙斌 | Ranking method for web page and web site |
| CN101105799A (en) * | 2007-08-02 | 2008-01-16 | 华中科技大学 | Method for evaluating file importance degree |
| CN101770521A (en) * | 2010-03-11 | 2010-07-07 | 东华大学 | Focusing relevancy ordering method for vertical search engine |
| US20110270847A1 (en) * | 2010-05-01 | 2011-11-03 | Adam Etkin | Method and system for appraising the extent to which a publication has been reviewed by means of a peer-review process |
| CN102270246A (en) * | 2011-09-08 | 2011-12-07 | 胡辉 | Method for calculating importance of web page |
| CN103559258A (en) * | 2013-11-04 | 2014-02-05 | 同济大学 | Webpage ranking method based on cloud computation |
Non-Patent Citations (3)
| Title |
|---|
| LARA SROUR ET AL: ""Personalized Web Page Ranking Using Trust and Similarity"", 《IEEE》 * |
| 杨晨: ""基于PageRank算法的中文垂直搜索引擎的研究与应用"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 药成刚: ""基于链接结构的中文网页排序算法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107545450A (en) * | 2016-06-27 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | A kind of Transaction Information method for pushing and device |
| CN107945079A (en) * | 2016-10-12 | 2018-04-20 | 普天信息技术有限公司 | A kind of poverty alleviation object selection method and device |
| CN108108419A (en) * | 2017-12-15 | 2018-06-01 | 百度在线网络技术（北京）有限公司 | A kind of information recommendation method, device, equipment and medium |
| CN108108419B (en) * | 2017-12-15 | 2021-01-26 | 百度在线网络技术（北京）有限公司 | Information recommendation method, device, equipment and medium |
| TWI739359B (en) * | 2019-03-28 | 2021-09-11 | 南韓商韓領有限公司 | Computer-implemented system, computer-implemented method for arranging hyperlinks on a graphical user-interface and non-transitory computer-readable storage medium |
| US11328328B2 (en) | 2019-03-28 | 2022-05-10 | Coupang Corp. | Computer-implemented method for arranging hyperlinks on a grapical user-interface |
| CN110413751A (en) * | 2019-07-18 | 2019-11-05 | 腾讯科技（深圳）有限公司 | Drug searching method, device, terminal device and storage medium |
| CN110413751B (en) * | 2019-07-18 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Medicine searching method, device, terminal equipment and storage medium |
| CN111738754A (en) * | 2020-06-01 | 2020-10-02 | 上海晶赞融宣科技有限公司 | Object recommendation method and device, storage medium and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105335363B (en) | 2018-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109902708B (en) | Recommendation model training method and related device | |
| CN105701216B (en) | A kind of information-pushing method and device | |
| CN110704739B (en) | Resource recommendation method and device and computer storage medium | |
| CN105335363A (en) | Object pushing method and system | |
| US20090276729A1 (en) | Adaptive user feedback window | |
| CN102591917B (en) | Data processing method and system and related device | |
| CN104933100B (en) | keyword recommendation method and device | |
| CN109190043A (en) | Recommended method and device, storage medium, electronic equipment and recommender system | |
| CN108320171A (en) | Hot item prediction technique, system and device | |
| CN103116582B (en) | A kind of information retrieval method and related system and device | |
| CN112307762B (en) | Search result sorting method and device, storage medium and electronic device | |
| CN110175895B (en) | Article recommendation method and device | |
| CN110458324B (en) | Method and device for calculating risk probability and computer equipment | |
| JP2021166109A (en) | Fusion sorting model training method and device, search sorting method and device, electronic device, storage medium, and program | |
| CN110020176A (en) | A kind of resource recommendation method, electronic equipment and computer readable storage medium | |
| CN105224959A (en) | The training method of order models and device | |
| CN104217031A (en) | Method and device for classifying users according to search log data of server | |
| CN104217030A (en) | Method and device for classifying users according to search log data of server | |
| CN104268142A (en) | Meta search result ranking algorithm based on rejection strategy | |
| CN108647691A (en) | A kind of image classification method based on click feature prediction | |
| CN110390056B (en) | Big data processing method, device and equipment and readable storage medium | |
| CN110706015B (en) | Feature selection method for advertisement click rate prediction | |
| CN107203558A (en) | Object recommendation method and apparatus, recommendation information treating method and apparatus | |
| CN112487283A (en) | Method and device for training model, electronic equipment and readable storage medium | |
| CN105740434B (en) | Network information methods of marking and device |
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 |