CN109166017A - Method for pushing, device, computer equipment and storage medium based on reunion class - Google Patents
Method for pushing, device, computer equipment and storage medium based on reunion class Download PDFInfo
- Publication number
- CN109166017A CN109166017A CN201811191703.1A CN201811191703A CN109166017A CN 109166017 A CN109166017 A CN 109166017A CN 201811191703 A CN201811191703 A CN 201811191703A CN 109166017 A CN109166017 A CN 109166017A
- Authority
- CN
- China
- Prior art keywords
- commodity
- vector
- user
- row vector
- group
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000003860 storage Methods 0.000 title claims abstract description 20
- 239000013598 vector Substances 0.000 claims abstract description 363
- 239000011159 matrix material Substances 0.000 claims abstract description 163
- 239000013065 commercial product Substances 0.000 claims abstract description 78
- 239000000047 product Substances 0.000 claims description 76
- 238000004590 computer program Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000012423 maintenance Methods 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 12
- 230000006870 function Effects 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 239000000203 mixture Substances 0.000 description 6
- 238000001914 filtration Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 239000006227 byproduct Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000006471 dimerization reaction Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000008676 import Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses method for pushing, device, computer equipment and storage mediums based on reunion class.This method is by carrying out DBSCAN cluster to row vector each in user-rating matrix, obtain the cluster group to classify by user, and with each one-to-one child user-rating matrix of cluster group, the row vector selected in child user-rating matrix is obtained as target user, according to the similar users group's rating matrix for obtaining target user in child user-rating matrix where target user, and commercial product recommending row vector corresponding with similar users group's rating matrix, commercial product recommending list is obtained according to commercial product recommending row vector, the commercial product recommending list is pushed into the corresponding receiving end of target user.The method achieve user-rating matrix is divided into multiple submatrixs to safeguard respectively, maintenance cost is reduced, and merchandise news push precisely can be carried out to target user according to submatrix cross.
Description
Technical field
The present invention relates to information advancing technique field more particularly to a kind of method for pushing based on reunion class, device, calculating
Machine equipment and storage medium.
Background technique
Currently, progress shopping at network is more and more frequent on online store Internet-based, these online stores pair
When user carries out commercial product recommending, usually used is that (collaborative filtering, principle are users to the proposed algorithm based on collaborative filtering
The commodity that those users with similar interests liked are liked, for example your friend likes film Harry Potter I, then will
You is recommended, this is the simplest collaborative filtering based on user).
Proposed algorithm based on collaborative filtering commonly uses user-rating matrix, and user-rating matrix indicates user to project
The scoring of (project can be understood as specific commodity), user-rating matrix horizontal axis are project, and the longitudinal axis is user, in the middle
Value is scoring of the user i to project j.As the data volume of commodity is increasing, user-rating matrix scoring of full dose is safeguarded
The cost of system can be higher and higher.
Summary of the invention
The embodiment of the invention provides a kind of method for pushing based on reunion class, device, computer equipment and storage medium,
Aim to solve the problem that the full dose user-corresponding points-scoring system of rating matrix in online store in the prior art with the increasing of commodity amount
It is more, it is more and more too fat to move, cause to full dose user-rating matrix problem difficult in maintenance.
In a first aspect, the embodiment of the invention provides a kind of method for pushing based on reunion class comprising:
By DBSCAN cluster acquired user-rating matrix is clustered, obtain at least one cluster group, and with
Each one-to-one child user-rating matrix of cluster group；
According to the corresponding target user of the row vector chosen in child user-rating matrix, the row of target user is obtained
The corresponding cluster group of vector；
In the corresponding cluster group of target user, calculates and obtain between each scoring row vector and the row vector of target user
Euclidean distance obtains the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance
Vector, to form similar users group's rating matrix；
According to the row vector that respectively scores in similar users group's rating matrix, similar users group is obtained to the comprehensive score of each commodity
Value, to form commercial product recommending row vector；And
Comprehensive grading value institute before being located at preset second rank threshold by the ranking that scores in commercial product recommending row vector is right
Commodity are answered to obtain commercial product recommending list, the commercial product recommending list is pushed into the corresponding receiving end of target user.
Second aspect, the embodiment of the invention provides a kind of driving means based on reunion class comprising:
User's cluster cell, for by DBSCAN cluster acquired user-rating matrix is clustered, obtain to
A few cluster group, and with each one-to-one child user-rating matrix of cluster group；
Judging unit is clustered, for according to the corresponding target user of row vector that is chosen in child user-rating matrix,
Obtain the corresponding cluster group of row vector of target user；
Similar users rating matrix acquiring unit, for calculating and obtaining each scoring in the corresponding cluster group of target user
Euclidean distance between row vector and the row vector of target user obtains ranking in each Euclidean distance and is located at preset first ranking
The corresponding scoring row vector of Euclidean distance before threshold value, to form similar users group's rating matrix；
Commercial product recommending row vector acquiring unit, for obtaining according to the row vector that respectively scores in similar users group's rating matrix
Similar users group is to the comprehensive grading values of each commodity, to form commercial product recommending row vector；
Information push unit, for being located at before preset second rank threshold by the ranking that scores in commercial product recommending row vector
Comprehensive grading value corresponding to commodity to obtain commercial product recommending list, it is corresponding that the commercial product recommending list is pushed into target user
Receiving end.
The third aspect, the embodiment of the present invention provide a kind of computer equipment again comprising memory, processor and storage
On the memory and the computer program that can run on the processor, the processor execute the computer program
Method for pushing based on reunion class described in the above-mentioned first aspect of Shi Shixian.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, wherein the computer can
It reads storage medium and is stored with computer program, it is above-mentioned that the computer program when being executed by a processor executes the processor
Method for pushing based on reunion class described in first aspect.
The embodiment of the invention provides a kind of method for pushing based on reunion class, device, computer equipment and storage mediums.
This method by row vector each in user-rating matrix carry out DBSCAN cluster, obtain by user classify cluster group, and with
Each one-to-one child user-rating matrix of cluster group obtains the row vector conduct selected in child user-rating matrix
Target user, according in child user-rating matrix where target user obtain target user similar users group's rating matrix,
And commercial product recommending row vector corresponding with similar users group's rating matrix, commercial product recommending column are obtained according to commercial product recommending row vector
The commercial product recommending list is pushed to the corresponding receiving end of target user by table.The method achieve draw user-rating matrix
It is divided into multiple submatrixs to be safeguarded respectively, reduces maintenance cost, and can precisely carry out to target user according to submatrix cross
Merchandise news push.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow diagram of the method for pushing provided in an embodiment of the present invention based on reunion class；
Fig. 2 is the sub-process schematic diagram of the method for pushing provided in an embodiment of the present invention based on reunion class；
Fig. 3 is another sub-process schematic diagram of the method for pushing provided in an embodiment of the present invention based on reunion class；
Fig. 4 is the schematic block diagram of the driving means provided in an embodiment of the present invention based on reunion class；
Fig. 5 is the subelement schematic block diagram of the driving means provided in an embodiment of the present invention based on reunion class；
Fig. 6 is another subelement schematic block diagram of the driving means provided in an embodiment of the present invention based on reunion class；
Fig. 7 is the schematic block diagram of computer equipment provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
It should be appreciated that ought use in this specification and in the appended claims, term " includes " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but one or more of the other feature, whole is not precluded
Body, step, operation, the presence or addition of element, component and/or its set.
It is also understood that mesh of the term used in this description of the invention merely for the sake of description specific embodiment
And be not intended to limit the present invention.As description of the invention and it is used in the attached claims, unless on
Other situations are hereafter clearly indicated, otherwise " one " of singular, "one" and "the" are intended to include plural form.
It will be further appreciated that the term "and/or" used in description of the invention and the appended claims is
Refer to any combination and all possible combinations of one or more of associated item listed, and including these combinations.
Referring to Fig. 1, Fig. 1 is the flow diagram of the method for pushing provided in an embodiment of the present invention based on reunion class, it should
Method for pushing based on reunion class is applied in management server, and this method passes through the application software that is installed in management server
It is executed, management server is the enterprise terminal for carrying out the push based on reunion class.
As shown in Figure 1, the method comprising the steps of S110~S150.
S110, acquired user-rating matrix is clustered by DBSCAN cluster, obtains at least one cluster
Group, and with each one-to-one child user-rating matrix of cluster group.
In the present embodiment, user-rating matrix indicates user to commodity (project can be understood as specific commodity)
Scoring, user-rating matrix horizontal axis are project, and the longitudinal axis is user, and value in the middle is scoring of the user i to project j.Such as with
The matrix that family-rating matrix S is 4 × 5, such as:
Wherein, the row vector of the first row indicates the commenting for commodity 1- commodity 5 respectively of user 1 in user-rating matrix S
Point, the row vector of the second row indicates that user 2 is directed to the scoring of commodity 1- commodity 5 respectively, and the row vector of the third line indicates that user 3 divides
Scoring of the safety pin to commodity 1- commodity 5, the row vector of fourth line indicate that user 4 is directed to the scoring of commodity 1- commodity 5 respectively.
Each row vector in user-rating matrix is clustered by DBSCAN Clustering Model, is realized according to user couple
Similar user is divided into same cluster group by the scoring of each commodity, and each user comments each commodity in same cluster group
It is approximate (namely smaller to point difference of the scoring of each commodity) for dividing.By the way that the user of full dose-rating matrix is carried out
DBSCAN clustering is after multiple child user-rating matrixs, only need be safeguarded i.e. respectively to each child user-rating matrix
Can, improve the efficiency of maintenance.
In one embodiment, as shown in Fig. 2, step S110 includes:
S111, using any one row vector in user-rating matrix as initial cluster center；
S112, points are included according to preset minimum, obtains the spacing between initial cluster center in preset scanning
Row vector within radius, using as initial clustering group；
S113, using row vector each in initial clustering group as cluster centre, obtain in user-rating matrix and in cluster
The row vector that the direct density of the heart is reachable, density is reachable or density is connected, using as cluster group adjusted.
In the present embodiment, for the clearer detailed process for understanding DBSCAN cluster, below in DBSCAN cluster
Related Feature Words are introduced.
Eps indicates sweep radius；
MinPts indicates minimum comprising points；
ξ neighborhood, indicates centered on given object, the region within the scope of the sweep radius for giving object；
Kernel object, if indicating, object number included in the ξ neighborhood of given object includes more than or equal to minimum
Points, then the given object is referred to as kernel object；
Direct density is reachable, indicates for sample set D, if sample point q, in the ξ neighborhood of p, and p is kernel object,
So object q is reachable from the direct density of object p；
Density is reachable, indicates to give a string of sample point p for sample set D1、p2、……、pnIf p1=q and pn=p,
If object piFrom pi-1Direct density is reachable, then object q is reachable from object p density；
Density is connected, and indicates that there are the point o in sample set D, if object o to object p and object q are that density can
It reaches, then p with q density is connected.
DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is
One more representational density-based algorithms.
DBSCAN needs two parameters: sweep radius (eps) and minimum include points (minPts).Optional one is not interviewed
It asks that the point of (unvisited) starts, finds out all points nearby with its distance within eps (including eps).
If quantity >=the minPts nearby put, current point and its one cluster of point formation nearby, and starting point is marked
It is denoted as and has accessed (visited).Then recurrence handles all in the cluster be not labeled as having accessed in the same way
(visited) point, to be extended to cluster.If quantity < the minPts nearby put, which is temporarily labeled to be used as and make an uproar
Sound point.If cluster is fully extended, i.e., all the points in cluster are marked as having accessed, and then go processing not with same algorithm
Accessed point.
The corresponding target user of row vector that S120, basis are chosen in child user-rating matrix, obtains target user
The corresponding cluster group of row vector.
In the present embodiment, in the corresponding child user-rating matrix of one of cluster group in multiple cluster groups, choosing
In a row vector as object vector after, while cluster group belonging to the row vector can be obtained, can quickly judge mesh in this way
Mark the similar users of user.
S130, in the corresponding cluster group of target user, calculate obtain it is each scoring row vector and target user row vector
Between Euclidean distance, the Euclidean distance obtained before ranking is located at preset first rank threshold in each Euclidean distance is corresponding
Score row vector, to form similar users group's rating matrix.
It in the present embodiment, can be in order to judge user similar with target user in cluster group belonging to target user
The Euclidean distance respectively to score between row vector and the row vector of target user in cluster group is obtained, each Euclidean distance is arranged in descending order
After sequence, the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance is obtained
Vector, to form similar users group's rating matrix.11 such as are set by the first rank threshold, then obtains ranking in each Euclidean distance
The corresponding scoring row vector of 1-10 Euclidean distances forms similar users group's rating matrix with this 10 row vectors.It obtains
After the similar users of target user, it can be calculated by row vector selected on a small quantity, without calling the user-of full dose
Rating matrix reduces the calculation amount in operational process.
In one embodiment, include: in step S130
Obtain the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance
Vector is arranged according to scoring row vector sequencing of row serial number in corresponding child user-rating matrix, is obtained similar
User group rating matrix.
In the present embodiment, occur according to each scoring row vector each row in corresponding child user-rating matrix successive
Sequentially, the scoring row vectors of each similar users is sequentially obtained, aforesaid way can accurately obtain the corresponding scoring of each similar users
Similar users group's rating matrix of row vector composition, convenient for subsequent calculating user to the comprehensive grading value of each commodity.
S140, according to the row vector that respectively scores in similar users group's rating matrix, obtain similar users group to the comprehensive of each commodity
Score value is closed, to form commercial product recommending row vector.
In the present embodiment, after obtaining the similar users of target user to the scoring of each commodity, each quotient can be directed to
Product calculate similar users to its comprehensive grading value.Euclidean in comprehensively considering target user and similar users between each user away from
From and similar users to the score values of each commodity, can operation obtain commercial product recommending row vector, with commercial product recommending row vector work
For the foundation of commercial product recommending.
In one embodiment, as shown in figure 3, step S140 includes:
S141, according to respectively scoring row vector in similar users group's rating matrix respectively between the row vector of target user
Euclidean distance, to form similar users group's Euclidean distance row vector；
S142, it is multiplied to obtain similar users with similar users group's rating matrix according to similar users group's Euclidean distance row vector
Group is to the comprehensive grading values of each commodity, to form commercial product recommending row vector.
In the present embodiment, after such as carrying out DBSCAN cluster by user-rating matrix S, the first cluster group and the are obtained
Dimerization monoid, wherein the first cluster group includes the scoring row vector of user 1 and user 2, the second cluster group includes user 3 and user
4 scoring row vector.At this time first the corresponding child user-rating matrix of group is clustered by the scoring row vector group of user 1 and user 2
At the corresponding child user-rating matrix of the second cluster group is made of the scoring row vector of user 3 and user 4.
If having selected user 1 as target user, the corresponding cluster group of the scoring row vector of user 1 is the first cluster
Group further includes the scoring row vector of user 2 in addition to the scoring vector including user 1 in the first cluster group.At this point, similar use
Family group's rating matrix is [0 413 2], the scoring row vector of similar users group rating matrix [0 413 2] and target user
Euclidean distance between [1 315 2] isBy similar users group's Euclidean distance row vectorWith the row vector of mark user
[1 315 2] it is multiplied and obtainsSimilar users group has been obtained to the synthesis of each commodity
The commercial product recommending row vector of score value composition.
It is commented when the corresponding scoring row vector of each user in the cluster group according to belonging to target user and with target user
Commercial product recommending row vector is calculated in the Euclidean distance of branch's vector, arranges in commercial product recommending row vector the scoring of each commodity
The forward commodity in position can be used as one of the component of commercial product recommending list, be pushed away in this way by what commercial product recommending row vector obtained
Hobby of the commodity due to having fully considered approximated user is recommended, therefore can accurately reflect the hobby of target user.
S150, the comprehensive grading value being located at before preset second rank threshold by the ranking that scores in commercial product recommending row vector
Corresponding commodity push to the corresponding receiving end of target user to obtain commercial product recommending list, by the commercial product recommending list.
In the present embodiment, in calculating obtained commercial product recommending row vector, you can learn that gathering belonging to target user
Each user is located at before the second rank threshold (such as the comprehensive score of each commodity with the ranking that scores in the comprehensive score in monoid
The second rank threshold, which is arranged, can be used as the recommendation items of commercial product recommending list for commodity corresponding to scoring 4).
For example, before scoring is located in the corresponding commercial product recommending row vector of similar users (user 2) of target user (user 1)
3 commodity are commodity 2, commodity 4 and commodity 5 respectively, and above-mentioned 3 commodity are pushed to target as items list at this time and are used
Family.
By the above-mentioned calculating based on Euclidean distance, the commodity of similar users can be liked and recommend quotient as to target user
The principal element considered when product can more reasonably carry out commercial product recommending.
In one embodiment, before step S110 further include:
History merchandise news set is obtained, by word frequency-inverse document frequency model to the history merchandise news collection
Each history merchandise news carries out key word information extraction in conjunction, and it is crucial to obtain commodity corresponding with each history merchandise news
Set of words；
Pass through the corresponding term vector of commodity keyword each in each commodity keyword set of Word2Vec model acquisition；
Obtain the average value of term vector corresponding to each commodity keyword in each commodity keyword set, with obtain with it is each
The corresponding statistical vector of commodity keyword set；
The corresponding statistical vector of commodity keyword set is clustered by DBSCAN Clustering Model, obtains at least one
Commercial articles clustering cluster；
If in user-rating matrix including blank value, according to the corresponding product name of the blank value, with acquisition and commodity
The corresponding statistical vector of title；
Obtain the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to；
According to the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to, obtain corresponding with the blank value
Product name similar product name accordingly, using as similar commodity result；
According to the corresponding row vector of the blank value, obtain corresponding with each product name in the similar commodity result
Scoring；
It is weighted and averaged according to scoring corresponding with product name each in the similar commodity result, obtains the sky
The corresponding commodity weighted scoring of white value, is updated to corresponding commodity weighted scoring for blank value.
In the present embodiment, in the user oriented interface UI provided in management server, user can choose a variety of
One of commodity are a variety of and bought.It is to be stored with history merchandise news set, history commodity in the management server
Each history merchandise news includes by product name and item property in information aggregate, and wherein the attribute of commodity includes the valence of commodity
Lattice, label, brand and function etc..
When by TF-IDF model, (i.e. term frequency-inverse document frequency indicates word
Frequently-inverse document frequency model) keyword letter is carried out to each history merchandise news in the history merchandise news set
Breath extracts, and the simplification of each history merchandise news can be expressed as corresponding commodity keyword set.
Word frequency-inverse document frequency model is a kind of common weighting technique for information retrieval and data mining.TF
Mean that word frequency (Term Frequency), IDF mean inverse document frequency (Inverse Document
Frequency).TF-IDF is a kind of statistical method, to assess a words in a file set or a corpus
The significance level of a copy of it file.The importance of words is with the directly proportional increase of number that it occurs hereof, but simultaneously
Can be inversely proportional decline with the frequency that it occurs in corpus.
For example, there is the basketball of a Spalding brand XX model YY member, there are also the merchandise news abundant such as the place of production, size,
But after TF-IDF model carries out keyword abstraction, the commodity keyword set finally obtained is combined into " basketball+Spalding+XX
Model ".In this way, each history merchandise news is reduced to corresponding commodity keyword set, can be convenient for being converted into term vector.
It, can after converting commodity keyword set corresponding with each history merchandise news for history merchandise news set again
By Word2Vec model, (Word2Vec is a kind of mould for learning semantic knowledge in unsupervised mode from a large amount of corpus of text
Type) it converts commodity keyword set to and each one-to-one term vector of history merchandise news.
Such as in corpus, basketball, Spalding, XX model one vector of each correspondence, only one value is 1 in vector,
Remaining is all 0, can convert the corresponding vector input Word2Vec model of above- mentioned information to the successive value of low dimensional, that is,
Dense vector, and wherein the word of similar import will be mapped to that similar position in vector space.
The term vector of each commodity keyword in obtaining commodity keyword set, then take the word of each commodity keyword to
Measure statistical vector of the average value as the commodity.It is at this time that history merchandise news each in history merchandise news set is equal
Be converted into corresponding statistical vector, later will by DBSCAN Clustering Model to the corresponding statistical vector of commodity keyword set into
Row cluster, can be obtained at least one commercial articles clustering cluster.
Since it is determined may know that is after the row vector that the blank value is belonged in initial user-rating matrix
Which user is blank value for the commodity scoring of which commodity, first obtains the corresponding product name of the blank value at this time
Know statistical vector corresponding to the product name.Then judge the commercial articles clustering cluster that the statistical vector is belonged to, can obtain
The similar product name of other commodity in the commercial articles clustering cluster, using the similar commodity as the corresponding product name of the blank value
As a result.In the row vector belonged in initial user-rating matrix by the blank value, it would know that the user for similar
The scoring of each similar product name in commodity result.Finally it is weighted according to the user for the scoring of each similar product name
It is average, obtain the corresponding commodity weighted scoring of the blank value.
In one embodiment, basis scoring corresponding with product name each in the similar commodity result is added
Weight average obtains the corresponding commodity weighted scoring of the blank value, comprising:
Using the corresponding statistical vector of product name each in the similar commodity result as statistical vector group, by the sky
The corresponding statistical vector of white value corresponding goods title obtains each in the statistical vector group as commodity to be predicted scoring vector
The distance between statistical vector and commodity to be predicted scoring vector, to obtain vector distance set；
By the corresponding scoring of product name each in the similar commodity result multiplied by corresponding vector in vector distance set
Distance is simultaneously summed, and commodity weighting overall score is obtained；
By commodity weighting overall score divided by the sum of each vector distance in vector distance set, it is corresponding to obtain the blank value
Commodity weighted scoring.
In the present embodiment, if by between each statistical vector in the statistical vector group and commodity to be predicted scoring vector
Distance be denoted as dck, by the corresponding user of row vector where the blank value to each product name in the similar commodity result
Corresponding scoring is denoted as Sic, the corresponding commodity weighted scoring of the blank value is calculated by following formula:
Wherein, ScorekIndicate that the commodity weighted scoring of the blank value corresponding goods k, m are the similar commodity result
In similar commodity c total number.
For example, user 1 is blank value for the scoring of commodity 2, and obtaining the corresponding similar commodity result of commodity 2 is commodity
4 and commodity 5, and user 1 is respectively 3 and 4 for the scoring of commodity 4 and commodity 5, the corresponding statistical vector of commodity 4 and commodity 2 are right
The distance between statistical vector answered is 0.5, between the corresponding statistical vector of commodity 5 statistical vector corresponding with commodity 2 away from
From being 1, then:
Score2=(0.5*3+1*4)/(0.5+1)=11/3；
At this point, by the above-mentioned Score being calculated2As the corresponding commodity weighted scoring of the blank value.
It, can effective completion user-by the prediction technique based on content by scoring commodity lacking in new user
Rating matrix avoids the problem of being cold-started in recommendation process.
The method achieve user-rating matrix is divided into multiple submatrixs to safeguard respectively, maintenance cost is reduced,
And merchandise news push precisely can be carried out to target user according to submatrix cross.
The embodiment of the present invention also provides a kind of driving means based on reunion class, is somebody's turn to do the driving means based on reunion class and is used for
Execute any embodiment of the aforementioned method for pushing based on reunion class.Specifically, referring to Fig. 4, Fig. 4 is that the embodiment of the present invention mentions
The schematic block diagram of the driving means based on reunion class supplied.The driving means 100 based on reunion class can be configured at management
In server.
As shown in figure 4, the driving means 100 based on reunion class include user's cluster cell 110, cluster judging unit 120,
Similar users rating matrix acquiring unit 130, commercial product recommending row vector acquiring unit 140 and information push unit 150.
User's cluster cell 110 is obtained for being clustered by DBSCAN cluster to acquired user-rating matrix
To at least one cluster group, and with each one-to-one child user-rating matrix of cluster group.
In the present embodiment, user-rating matrix indicates user to commodity (project can be understood as specific commodity)
Scoring, user-rating matrix horizontal axis are project, and the longitudinal axis is user, and value in the middle is scoring of the user i to project j.Such as with
The matrix that family-rating matrix S is 4 × 5, such as:
Wherein, the row vector of the first row indicates the commenting for commodity 1- commodity 5 respectively of user 1 in user-rating matrix S
Point, the row vector of the second row indicates that user 2 is directed to the scoring of commodity 1- commodity 5 respectively, and the row vector of the third line indicates that user 3 divides
Scoring of the safety pin to commodity 1- commodity 5, the row vector of fourth line indicate that user 4 is directed to the scoring of commodity 1- commodity 5 respectively.
Each row vector in user-rating matrix is clustered by DBSCAN Clustering Model, is realized according to user couple
Similar user is divided into same cluster group by the scoring of each commodity, and each user comments each commodity in same cluster group
It is approximate (namely smaller to point difference of the scoring of each commodity) for dividing.By the way that the user of full dose-rating matrix is carried out
DBSCAN clustering is after multiple child user-rating matrixs, only need be safeguarded i.e. respectively to each child user-rating matrix
Can, improve the efficiency of maintenance.
In one embodiment, as shown in figure 5, user's cluster cell 110 includes:
Initial center acquiring unit 111, for using any one row vector in user-rating matrix as in initial clustering
The heart；
Initial clustering group acquiring unit 112, for according to preset minimum comprising points, obtain with initial cluster center it
Between row vector of the spacing within preset sweep radius, using as initial clustering group；
Group's adjustment unit 113 is clustered, for obtaining user-using row vector each in initial clustering group as cluster centre
In rating matrix with the row vector that the direct density of cluster centre is reachable, density is reachable or density is connected, using as adjusted poly-
Monoid.
DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is
One more representational density-based algorithms.
DBSCAN needs two parameters: sweep radius (eps) and minimum include points (minPts).Optional one is not interviewed
It asks that the point of (unvisited) starts, finds out all points nearby with its distance within eps (including eps).
If quantity >=the minPts nearby put, current point and its one cluster of point formation nearby, and starting point is marked
It is denoted as and has accessed (visited).Then recurrence handles all in the cluster be not labeled as having accessed in the same way
(visited) point, to be extended to cluster.If quantity < the minPts nearby put, which is temporarily labeled to be used as and make an uproar
Sound point.If cluster is fully extended, i.e., all the points in cluster are marked as having accessed, and then go processing not with same algorithm
Accessed point.
Judging unit 120 is clustered, for using according to the corresponding target of the row vector chosen in child user-rating matrix
Family obtains the corresponding cluster group of row vector of target user.
In the present embodiment, in the corresponding child user-rating matrix of one of cluster group in multiple cluster groups, choosing
In a row vector as object vector after, while cluster group belonging to the row vector can be obtained, can quickly judge mesh in this way
Mark the similar users of user.
Similar users rating matrix acquiring unit 130 is respectively commented in the corresponding cluster group of target user, calculating to obtain
Euclidean distance between branch's vector and the row vector of target user obtains ranking in each Euclidean distance and is located at preset first row
The corresponding scoring row vector of Euclidean distance before name threshold value, to form similar users group's rating matrix.
It in the present embodiment, can be in order to judge user similar with target user in cluster group belonging to target user
The Euclidean distance respectively to score between row vector and the row vector of target user in cluster group is obtained, each Euclidean distance is arranged in descending order
After sequence, the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance is obtained
Vector, to form similar users group's rating matrix.11 such as are set by the first rank threshold, then obtains ranking in each Euclidean distance
The corresponding scoring row vector of 1-10 Euclidean distances forms similar users group's rating matrix with this 10 row vectors.It obtains
After the similar users of target user, it can be calculated by row vector selected on a small quantity, without calling the user-of full dose
Rating matrix reduces the calculation amount in operational process.
In one embodiment, it obtains ranking in each Euclidean distance and is located at the Euclidean distance before preset first rank threshold
Corresponding scoring row vector, to form similar users group's rating matrix, comprising:
Obtain the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance
Vector is arranged according to scoring row vector sequencing of row serial number in corresponding child user-rating matrix, is obtained similar
User group rating matrix.
In the present embodiment, occur according to each scoring row vector each row in corresponding child user-rating matrix successive
Sequentially, the scoring row vectors of each similar users is sequentially obtained, aforesaid way can accurately obtain the corresponding scoring of each similar users
Similar users group's rating matrix of row vector composition, convenient for subsequent calculating user to the comprehensive grading value of each commodity.
Commercial product recommending row vector acquiring unit 140, for obtaining according to the row vector that respectively scores in similar users group's rating matrix
Take similar users group to the comprehensive grading value of each commodity, to form commercial product recommending row vector.
In the present embodiment, after obtaining the similar users of target user to the scoring of each commodity, each quotient can be directed to
Product calculate similar users to its comprehensive grading value.Euclidean in comprehensively considering target user and similar users between each user away from
From and similar users to the score values of each commodity, can operation obtain commercial product recommending row vector, with commercial product recommending row vector work
For the foundation of commercial product recommending.
In one embodiment, as shown in fig. 6, commercial product recommending row vector acquiring unit 140 includes:
Euclidean distance row vector acquiring unit 141, for according to the row vector point that respectively scores in similar users group's rating matrix
Euclidean distance not between the row vector of target user, to form similar users group's Euclidean distance row vector；
Comprehensive grading value computing unit 142, for being commented according to similar users group's Euclidean distance row vector with similar users group
Sub-matrix is multiplied to obtain similar users group to the comprehensive grading value of each commodity, to form commercial product recommending row vector.
In the present embodiment, after such as carrying out DBSCAN cluster by user-rating matrix S, the first cluster group and the are obtained
Dimerization monoid, wherein the first cluster group includes the scoring row vector of user 1 and user 2, the second cluster group includes user 3 and user
4 scoring row vector.At this time first the corresponding child user-rating matrix of group is clustered by the scoring row vector group of user 1 and user 2
At the corresponding child user-rating matrix of the second cluster group is made of the scoring row vector of user 3 and user 4.
If having selected user 1 as target user, the corresponding cluster group of the scoring row vector of user 1 is the first cluster
Group further includes the scoring row vector of user 2 in addition to the scoring vector including user 1 in the first cluster group.At this point, similar use
Family group's rating matrix is [0 413 2], the scoring row vector of similar users group rating matrix [0 413 2] and target user
Euclidean distance between [1 315 2] isBy similar users group's Euclidean distance row vectorWith mark user row to
Amount [1 315 2], which is multiplied, to be obtainedSimilar users group has been obtained to the comprehensive of each commodity
Close the commercial product recommending row vector of score value composition.
It is commented when the corresponding scoring row vector of each user in the cluster group according to belonging to target user and with target user
Commercial product recommending row vector is calculated in the Euclidean distance of branch's vector, arranges in commercial product recommending row vector the scoring of each commodity
The forward commodity in position can be used as one of the component of commercial product recommending list, be pushed away in this way by what commercial product recommending row vector obtained
Hobby of the commodity due to having fully considered approximated user is recommended, therefore can accurately reflect the hobby of target user.
Information push unit 150, for being located at preset second rank threshold by the ranking that scores in commercial product recommending row vector
Commodity corresponding to comprehensive grading value before push to target user to obtain commercial product recommending list, by the commercial product recommending list
Corresponding receiving end.
In the present embodiment, in calculating obtained commercial product recommending row vector, you can learn that gathering belonging to target user
Each user is located at before the second rank threshold (such as the comprehensive score of each commodity with the ranking that scores in the comprehensive score in monoid
The second rank threshold, which is arranged, can be used as the recommendation items of commercial product recommending list for commodity corresponding to scoring 4).
For example, before scoring is located in the corresponding commercial product recommending row vector of similar users (user 2) of target user (user 1)
3 commodity are commodity 2, commodity 4 and commodity 5 respectively, and above-mentioned 3 commodity are pushed to target as items list at this time and are used
Family.
By the above-mentioned calculating based on Euclidean distance, the commodity of similar users can be liked and recommend quotient as to target user
The principal element considered when product can more reasonably carry out commercial product recommending.
In one embodiment, based on the driving means 100 of reunion class, further includes:
History keyword set of words acquiring unit is referred to for obtaining history merchandise news set by the inverse text frequency of word frequency-
Exponential model carries out key word information extraction to each history merchandise news in the history merchandise news set, obtain with it is each
The corresponding commodity keyword set of history merchandise news；
Term vector conversion unit, it is crucial for obtaining each commodity in each commodity keyword set by Word2Vec model
The corresponding term vector of word；
Statistical vector acquiring unit, for obtaining term vector corresponding to each commodity keyword in each commodity keyword set
Average value, to obtain statistical vector corresponding with each commodity keyword set；
Commercial articles clustering unit, for being carried out by DBSCAN Clustering Model to the corresponding statistical vector of commodity keyword set
Cluster, obtains at least one commercial articles clustering cluster；
Vector acquiring unit to be predicted, if for including blank value in user-rating matrix, it is corresponding according to the blank value
Product name, to obtain corresponding with product name statistical vector；
Commercial articles clustering cluster belongs to judging unit, poly- for obtaining the commodity that statistical vector corresponding with product name is belonged to
Class cluster；
Similar commodity result acquiring unit, the commercial articles clustering for being belonged to according to statistical vector corresponding with product name
Cluster obtains the corresponding similar product name of corresponding with blank value product name, using as similar commodity result；
Similar commodity scoring acquiring unit, for according to the corresponding row vector of the blank value, acquisition and the similar quotient
The corresponding scoring of each product name in product result；
Commodity weighted scoring acquiring unit, for according to each product name is corresponding in the similar commodity result comments
Divide and be weighted and averaged, obtains the corresponding commodity weighted scoring of the blank value, blank value is updated to corresponding commodity and is added
Power scoring.
In the present embodiment, in the user oriented interface UI provided in management server, user can choose a variety of
One of commodity are a variety of and bought.It is to be stored with history merchandise news set, history commodity in the management server
Each history merchandise news includes by product name and item property in information aggregate, and wherein the attribute of commodity includes the valence of commodity
Lattice, label, brand and function etc..
When by TF-IDF model, (i.e. term frequency-inverse document frequency indicates word
Frequently-inverse document frequency model) keyword letter is carried out to each history merchandise news in the history merchandise news set
Breath extracts, and the simplification of each history merchandise news can be expressed as corresponding commodity keyword set.
Word frequency-inverse document frequency model is a kind of common weighting technique for information retrieval and data mining.TF
Mean that word frequency (Term Frequency), IDF mean inverse document frequency (Inverse Document
Frequency).TF-IDF is a kind of statistical method, to assess a words in a file set or a corpus
The significance level of a copy of it file.The importance of words is with the directly proportional increase of number that it occurs hereof, but simultaneously
Can be inversely proportional decline with the frequency that it occurs in corpus.
For example, there is the basketball of a Spalding brand XX model YY member, there are also the merchandise news abundant such as the place of production, size,
But after TF-IDF model carries out keyword abstraction, the commodity keyword set finally obtained is combined into " basketball+Spalding+XX
Model ".In this way, each history merchandise news is reduced to corresponding commodity keyword set, can be convenient for being converted into term vector.
It, can after converting commodity keyword set corresponding with each history merchandise news for history merchandise news set again
By Word2Vec model, (Word2Vec is a kind of mould for learning semantic knowledge in unsupervised mode from a large amount of corpus of text
Type) it converts commodity keyword set to and each one-to-one term vector of history merchandise news.
Such as in corpus, basketball, Spalding, XX model one vector of each correspondence, only one value is 1 in vector,
Remaining is all 0, can convert the corresponding vector input Word2Vec model of above- mentioned information to the successive value of low dimensional, that is,
Dense vector, and wherein the word of similar import will be mapped to that similar position in vector space.
The term vector of each commodity keyword in obtaining commodity keyword set, then take the word of each commodity keyword to
Measure statistical vector of the average value as the commodity.It is at this time that history merchandise news each in history merchandise news set is equal
Be converted into corresponding statistical vector, later will by DBSCAN Clustering Model to the corresponding statistical vector of commodity keyword set into
Row cluster, can be obtained at least one commercial articles clustering cluster.
Since it is determined may know that is after the row vector that the blank value is belonged in initial user-rating matrix
Which user is blank value for the commodity scoring of which commodity, first obtains the corresponding product name of the blank value at this time
Know statistical vector corresponding to the product name.Then judge the commercial articles clustering cluster that the statistical vector is belonged to, can obtain
The similar product name of other commodity in the commercial articles clustering cluster, using the similar commodity as the corresponding product name of the blank value
As a result.In the row vector belonged in initial user-rating matrix by the blank value, it would know that the user for similar
The scoring of each similar product name in commodity result.Finally it is weighted according to the user for the scoring of each similar product name
It is average, obtain the corresponding commodity weighted scoring of the blank value.
In one embodiment, commodity weighted scoring acquiring unit, comprising:
Vector distance set acquiring unit, for by each product name in the similar commodity result it is corresponding count to
Amount be used as statistical vector group, using the corresponding statistical vector of blank value corresponding goods title as commodity to be predicted score to
Amount obtains the distance between each statistical vector and commodity to be predicted scoring vector in the statistical vector group, to obtain vector
Distance set；
Commodity weight overall score acquiring unit, for multiplying the corresponding scoring of product name each in the similar commodity result
It with corresponding vector distance in vector distance set and sums, obtains commodity weighting overall score；
Average mark acquiring unit, for by commodity weighting overall score divided by the sum of each vector distance in vector distance set,
Obtain the corresponding commodity weighted scoring of the blank value.
In the present embodiment, if by between each statistical vector in the statistical vector group and commodity to be predicted scoring vector
Distance be denoted as dck, by the corresponding user of row vector where the blank value to each product name in the similar commodity result
Corresponding scoring is denoted as Sic, the corresponding commodity weighted scoring of the blank value is calculated by following formula:
Wherein, ScorekIndicate that the commodity weighted scoring of the blank value corresponding goods k, m are the similar commodity result
In similar commodity c total number.
For example, user 1 is blank value for the scoring of commodity 2, and obtaining the corresponding similar commodity result of commodity 2 is commodity
4 and commodity 5, and user 1 is respectively 3 and 4 for the scoring of commodity 4 and commodity 5, the corresponding statistical vector of commodity 4 and commodity 2 are right
The distance between statistical vector answered is 0.5, between the corresponding statistical vector of commodity 5 statistical vector corresponding with commodity 2 away from
From being 1, then:
Score2=(0.5*3+1*4)/(0.5+1)=11/3；
At this point, by the above-mentioned Score being calculated2As the corresponding commodity weighted scoring of the blank value.
It, can effective completion user-by the prediction technique based on content by scoring commodity lacking in new user
Rating matrix avoids the problem of being cold-started in recommendation process.
The arrangement achieves user-rating matrix is divided into multiple submatrixs to safeguard respectively, maintenance cost is reduced,
And merchandise news push precisely can be carried out to target user according to submatrix cross.
The above-mentioned driving means based on reunion class can be implemented as the form of computer program, which can be
It is run in computer equipment as shown in Figure 7.
Referring to Fig. 7, Fig. 7 is the schematic block diagram of computer equipment provided in an embodiment of the present invention.
Refering to Fig. 7, which includes processor 502, memory and the net connected by system bus 501
Network interface 505, wherein memory may include non-volatile memory medium 503 and built-in storage 504.
The non-volatile memory medium 503 can storage program area 5031 and computer program 5032.The computer program
5032 are performed, and processor 502 may make to execute the method for pushing based on reunion class.
The processor 502 supports the operation of entire computer equipment 500 for providing calculating and control ability.
The built-in storage 504 provides environment for the operation of the computer program 5032 in non-volatile memory medium 503, should
When computer program 5032 is executed by processor 502, processor 502 may make to execute the method for pushing based on reunion class.
The network interface 505 is for carrying out network communication, such as the transmission of offer data information.Those skilled in the art can
To understand, structure shown in Fig. 7, only the block diagram of part-structure relevant to the present invention program, is not constituted to this hair
The restriction for the computer equipment 500 that bright scheme is applied thereon, specific computer equipment 500 may include than as shown in the figure
More or fewer components perhaps combine certain components or with different component layouts.
Wherein, the processor 502 is for running computer program 5032 stored in memory, to realize following function
Can: by DBSCAN cluster acquired user-rating matrix is clustered, obtain at least one cluster group, and with it is each
Cluster the one-to-one child user-rating matrix of group；According to the corresponding mesh of the row vector chosen in child user-rating matrix
User is marked, the corresponding cluster group of row vector of target user is obtained；In the corresponding cluster group of target user, calculates to obtain and respectively comment
Euclidean distance between branch's vector and the row vector of target user obtains ranking in each Euclidean distance and is located at preset first row
The corresponding scoring row vector of Euclidean distance before name threshold value, to form similar users group's rating matrix；According to similar users group
Respectively score row vector in rating matrix, obtains similar users group to the comprehensive grading values of each commodity, with form commercial product recommending row to
Amount；And as scoring corresponding to the comprehensive grading value that ranking is located at before preset second rank threshold in commercial product recommending row vector
Commodity push to the corresponding receiving end of target user to obtain commercial product recommending list, by the commercial product recommending list.
In one embodiment, processor 502 gathers user-rating matrix in described clustered by DBSCAN of execution
Class performs the following operations when obtaining the step of at least one cluster group: any one row vector in user-rating matrix is made
For initial cluster center；Include points according to preset minimum, the spacing obtained between initial cluster center is swept preset
The row vector within radius is retouched, using as initial clustering group；Using row vector each in initial clustering group as cluster centre, obtain
In user-rating matrix with the row vector that the direct density of cluster centre is reachable, density is reachable or density is connected, as after adjustment
Cluster group.
In one embodiment, processor 502 execute it is described according to the row vector that respectively scores in similar users group's rating matrix,
Similar users group is obtained to the comprehensive grading values of each commodity, when step to form commercial product recommending row vector, performed the following operations:
According to Euclidean distance of the row vector respectively between the row vector of target user that respectively score in similar users group's rating matrix, with group
At similar users group's Euclidean distance row vector；According to similar users group's Euclidean distance row vector and similar users group's rating matrix phase
It is multiplied to arrive similar users group to the comprehensive grading value of each commodity, to form commercial product recommending row vector.
In one embodiment, the ranking in executing each Euclidean distance of acquisition of processor 502 is located at preset first row
Euclidean distance corresponding scoring row vector before name threshold value when step to form similar users group's rating matrix, executes such as
Lower operation: obtain the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance to
Amount is arranged according to scoring row vector sequencing of row serial number in corresponding child user-rating matrix, obtains similar use
Family group's rating matrix.
In one embodiment, processor 502 execute it is described by DBSCAN cluster to acquired user-rating matrix
Clustered, obtain at least one cluster group, and the step of child user-rating matrix one-to-one with each cluster group it
Before, it also performs the following operations: history merchandise news set is obtained, by word frequency-inverse document frequency model to the history
Each history merchandise news carries out key word information extraction in merchandise news set, obtains corresponding with each history merchandise news
Commodity keyword set；Pass through the corresponding word of commodity keyword each in each commodity keyword set of Word2Vec model acquisition
Vector；The average value of term vector corresponding to each commodity keyword in each commodity keyword set is obtained, to obtain and each quotient
The corresponding statistical vector of product keyword set；By DBSCAN Clustering Model to the corresponding statistical vector of commodity keyword set into
Row cluster, obtains at least one commercial articles clustering cluster；It is corresponding according to the blank value if in user-rating matrix including blank value
Product name, to obtain corresponding with product name statistical vector；Statistical vector corresponding with product name is obtained to be belonged to
Commercial articles clustering cluster；According to the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to, obtain and the blank value
Corresponding product name similar product name accordingly, using as similar commodity result；According to the corresponding row of the blank value to
Amount obtains scoring corresponding with product name each in the similar commodity result；According to every to the similar commodity result
The corresponding scoring of one product name is weighted and averaged, and the corresponding commodity weighted scoring of the blank value is obtained, by blank value
It is updated to corresponding commodity weighted scoring.
In one embodiment, processor 502 is executing the basis and each product name in the similar commodity result
Corresponding scoring is weighted and averaged, and when obtaining the step of the corresponding commodity weighted scoring of the blank value, is performed the following operations:
It is using the corresponding statistical vector of product name each in the similar commodity result as statistical vector group, the blank value is corresponding
The corresponding statistical vector of product name obtains each statistical vector in the statistical vector group as commodity to be predicted scoring vector
The distance between commodity to be predicted scoring vector, to obtain vector distance set；By each quotient in the similar commodity result
The name of an article claims corresponding scoring multiplied by corresponding vector distance in vector distance set and sums, and obtains commodity weighting overall score；By quotient
Product weight overall score divided by the sum of each vector distance in vector distance set, obtain the corresponding commodity weighting of the blank value and comment
Point.
It will be understood by those skilled in the art that the embodiment of computer equipment shown in Fig. 7 is not constituted to computer
The restriction of equipment specific composition, in other embodiments, computer equipment may include components more more or fewer than diagram, or
Person combines certain components or different component layouts.For example, in some embodiments, computer equipment can only include depositing
Reservoir and processor, in such embodiments, the structure and function of memory and processor are consistent with embodiment illustrated in fig. 7,
Details are not described herein.
It should be appreciated that in embodiments of the present invention, processor 502 can be central processing unit (Central
ProcessingUnit, CPU), which can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic
Device, discrete gate or transistor logic, discrete hardware components etc..Wherein, general processor can be microprocessor or
Person's processor is also possible to any conventional processor etc..
Computer readable storage medium is provided in another embodiment of the invention.The computer readable storage medium can be with
For non-volatile computer readable storage medium.The computer-readable recording medium storage has computer program, wherein calculating
Machine program performs the steps of when being executed by processor gathers acquired user-rating matrix by DBSCAN cluster
Class, obtain at least one cluster group, and with each one-to-one child user-rating matrix of cluster group；It is commented according in child user-
The corresponding target user of the row vector chosen in sub-matrix obtains the corresponding cluster group of row vector of target user；In target
In the corresponding cluster group of user, the Euclidean distance obtained between each scoring row vector and the row vector of target user is calculated, is obtained
Ranking is located at the corresponding scoring row vector of Euclidean distance before preset first rank threshold in each Euclidean distance, to form phase
Like user group rating matrix；According to the row vector that respectively scores in similar users group's rating matrix, similar users group is obtained to each commodity
Comprehensive grading value, to form commercial product recommending row vector；And it is located at preset the by the ranking that scores in commercial product recommending row vector
Commodity corresponding to comprehensive grading value before two rank thresholds are pushed the commercial product recommending list with obtaining commercial product recommending list
To the corresponding receiving end of target user.
In one embodiment, described that user-rating matrix is clustered by DBSCAN cluster, it is poly- to obtain at least one
Monoid, comprising: using any one row vector in user-rating matrix as initial cluster center；Include according to preset minimum
Points obtain row vector of the spacing within preset sweep radius between initial cluster center, using as initial clustering
Group；Using row vector each in initial clustering group as cluster centre, obtain in user-rating matrix with the direct density of cluster centre
The row vector that reachable, density is reachable or density is connected, using as cluster group adjusted.
In one embodiment, described according to the row vector that respectively scores in similar users group's rating matrix, obtain similar users group
To the comprehensive grading value of each commodity, to form commercial product recommending row vector, comprising: respectively score according in similar users group's rating matrix
The row vector Euclidean distance between the row vector of target user respectively, to form similar users group's Euclidean distance row vector；Root
It is multiplied to obtain similar users group to the comprehensive of each commodity with similar users group's rating matrix according to similar users group's Euclidean distance row vector
Score value is closed, to form commercial product recommending row vector.
In one embodiment, European before ranking is located at preset first rank threshold in each Euclidean distance of acquisition
Apart from corresponding scoring row vector, to form similar users group's rating matrix, comprising: obtain ranking in each Euclidean distance and be located in advance
If the first rank threshold before the corresponding scoring row vector of Euclidean distance, according to scoring row vector in corresponding child user-
The sequencing of row serial number is arranged in rating matrix, obtains similar users group's rating matrix.
In one embodiment, it is described by DBSCAN cluster acquired user-rating matrix is clustered, obtain to
A few cluster group, and with before each one-to-one child user-rating matrix of cluster group, further includes: acquisition history commodity
Information aggregate, by word frequency-inverse document frequency model to each history merchandise news in the history merchandise news set
Key word information extraction is carried out, commodity keyword set corresponding with each history merchandise news is obtained；Pass through Word2Vec
Model obtains the corresponding term vector of each commodity keyword in each commodity keyword set；It obtains in each commodity keyword set
The average value of term vector corresponding to each commodity keyword, to obtain statistical vector corresponding with each commodity keyword set；It is logical
It crosses DBSCAN Clustering Model to cluster the corresponding statistical vector of commodity keyword set, obtains at least one commercial articles clustering
Cluster；If in user-rating matrix including blank value, according to the corresponding product name of the blank value, with acquisition and product name
Corresponding statistical vector；Obtain the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to；According to product name
The commercial articles clustering cluster that corresponding statistical vector is belonged to obtains the corresponding similar commodity of corresponding with blank value product name
Title, using as similar commodity result；According to the corresponding row vector of the blank value, obtain every to the similar commodity result
The corresponding scoring of one product name；It is weighted according to scoring corresponding with product name each in the similar commodity result flat
, the corresponding commodity weighted scoring of the blank value is obtained, blank value is updated to corresponding commodity weighted scoring.
In one embodiment, basis scoring corresponding with product name each in the similar commodity result is added
Weight average obtains the corresponding commodity weighted scoring of the blank value, comprising: by each product name in the similar commodity result
Corresponding statistical vector is as statistical vector group, using the corresponding statistical vector of blank value corresponding goods title as to be predicted
Commodity scoring vector obtains the distance between each statistical vector and commodity to be predicted scoring vector in the statistical vector group,
To obtain vector distance set；By the corresponding scoring of product name each in the similar commodity result multiplied by vector distance set
Corresponding vector distance is simultaneously summed, and commodity weighting overall score is obtained；By commodity weighting overall score divided by each in vector distance set
The sum of vector distance obtains the corresponding commodity weighted scoring of the blank value.
It is apparent to those skilled in the art that for convenience of description and succinctly, foregoing description is set
The specific work process of standby, device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
Those of ordinary skill in the art may be aware that unit described in conjunction with the examples disclosed in the embodiments of the present disclosure and algorithm
Step can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and software
Interchangeability generally describes each exemplary composition and step according to function in the above description.These functions are studied carefully
Unexpectedly the specific application and design constraint depending on technical solution are implemented in hardware or software.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
The scope of the present invention.
In several embodiments provided by the present invention, it should be understood that disclosed unit and method, it can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only logical function partition, there may be another division manner in actual implementation, can also will be with the same function
Unit set is at a unit, such as multiple units or components can be combined or can be integrated into another system or some
Feature can be ignored, or not execute.In addition, shown or discussed mutual coupling, direct-coupling or communication connection can
Be through some interfaces, the indirect coupling or communication connection of device or unit, be also possible to electricity, mechanical or other shapes
Formula connection.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.Some or all of unit therein can be selected to realize the embodiment of the present invention according to the actual needs
Purpose.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, is also possible to two or more units and is integrated in one unit.It is above-mentioned integrated
Unit both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in one storage medium.Based on this understanding, technical solution of the present invention is substantially in other words to existing
The all or part of part or the technical solution that technology contributes can be embodied in the form of software products, should
Computer software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be
Personal computer, server or network equipment etc.) execute all or part of step of each embodiment the method for the present invention
Suddenly.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), magnetic disk or
The various media that can store program code such as person's CD.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection scope subject to.
Claims (10)
1. a kind of method for pushing based on reunion class characterized by comprising
By DBSCAN cluster acquired user-rating matrix is clustered, obtain at least one cluster group, and with it is each
Cluster the one-to-one child user-rating matrix of group；
According to the corresponding target user of the row vector chosen in child user-rating matrix, the row vector of target user is obtained
Corresponding cluster group；
In the corresponding cluster group of target user, calculate European between each scoring row vector of acquisition and the row vector of target user
Distance, obtain the corresponding scoring row of Euclidean distance before ranking is located at preset first rank threshold in each Euclidean distance to
Amount, to form similar users group's rating matrix；
According to the row vector that respectively scores in similar users group's rating matrix, similar users group is obtained to the comprehensive grading value of each commodity,
To form commercial product recommending row vector；And
As the quotient corresponding to the comprehensive grading value that ranking is located at before preset second rank threshold that scores in commercial product recommending row vector
Product push to the corresponding receiving end of target user to obtain commercial product recommending list, by the commercial product recommending list.
2. the method for pushing according to claim 1 based on reunion class, which is characterized in that described to pass through DBSCAN cluster pair
User-rating matrix clusters, and obtains at least one cluster group, comprising:
Using any one row vector in user-rating matrix as initial cluster center；
Include points according to preset minimum, obtains the spacing between initial cluster center within preset sweep radius
Row vector, using as initial clustering group；
Using row vector each in initial clustering group as cluster centre, obtain directly close with cluster centre in user-rating matrix
The row vector that reachable, density is reachable or density is connected is spent, using as cluster group adjusted.
3. the method for pushing according to claim 1 based on reunion class, which is characterized in that described to be commented according to similar users group
Respectively score row vector in sub-matrix, obtains similar users group to the comprehensive grading values of each commodity, to form commercial product recommending row vector,
Include:
According to Euclidean distance of the row vector respectively between the row vector of target user that respectively score in similar users group's rating matrix,
To form similar users group's Euclidean distance row vector；
It is multiplied to obtain similar users group to each quotient with similar users group's rating matrix according to similar users group's Euclidean distance row vector
The comprehensive grading value of product, to form commercial product recommending row vector.
4. the method for pushing according to claim 1 based on reunion class, which is characterized in that described to obtain in each Euclidean distance
Ranking is located at the corresponding scoring row vector of Euclidean distance before preset first rank threshold, to form similar users group scoring
Matrix, comprising:
It obtains ranking in each Euclidean distance and is located at the corresponding scoring row vector of Euclidean distance before preset first rank threshold,
It is arranged according to scoring row vector sequencing of row serial number in corresponding child user-rating matrix, obtains similar users
Group's rating matrix.
5. the method for pushing according to claim 1 based on reunion class, which is characterized in that described to pass through DBSCAN cluster pair
Acquired user-rating matrix clusters, and obtains at least one cluster group, and use with the one-to-one son of each cluster group
Before family-rating matrix, further includes:
History merchandise news set is obtained, by word frequency-inverse document frequency model in the history merchandise news set
Each history merchandise news carries out key word information extraction, obtains commodity keyword set corresponding with each history merchandise news
It closes；
Pass through the corresponding term vector of commodity keyword each in each commodity keyword set of Word2Vec model acquisition；
The average value of term vector corresponding to each commodity keyword in each commodity keyword set is obtained, to obtain and each commodity
The corresponding statistical vector of keyword set；
The corresponding statistical vector of commodity keyword set is clustered by DBSCAN Clustering Model, obtains at least one commodity
Clustering cluster；
If in user-rating matrix including blank value, according to the corresponding product name of the blank value, with acquisition and product name
Corresponding statistical vector；
Obtain the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to；
According to the commercial articles clustering cluster that statistical vector corresponding with product name is belonged to, commodity corresponding with the blank value are obtained
Title similar product name accordingly, using as similar commodity result；
According to the corresponding row vector of the blank value, obtain and each product name is corresponding in the similar commodity result comments
Point；
It is weighted and averaged according to scoring corresponding with product name each in the similar commodity result, obtains the blank value
Blank value is updated to corresponding commodity weighted scoring by corresponding commodity weighted scoring.
6. the method for pushing according to claim 5 based on reunion class, which is characterized in that the basis and the similar quotient
Corresponding score of each product name is weighted and averaged in product result, obtains the corresponding commodity weighted scoring of the blank value,
Include:
Using the corresponding statistical vector of product name each in the similar commodity result as statistical vector group, by the blank value
The corresponding statistical vector of corresponding goods title obtains each statistics in the statistical vector group as commodity to be predicted scoring vector
The distance between vector and commodity to be predicted scoring vector, to obtain vector distance set；
By the corresponding scoring of product name each in the similar commodity result multiplied by corresponding vector distance in vector distance set
And sum, obtain commodity weighting overall score；
By commodity weighting overall score divided by the sum of each vector distance in vector distance set, the corresponding commodity of the blank value are obtained
Weighted scoring.
7. a kind of driving means based on reunion class characterized by comprising
User's cluster cell obtains at least one for clustering by DBSCAN cluster to acquired user-rating matrix
A cluster group, and with each one-to-one child user-rating matrix of cluster group；
Judging unit is clustered, for obtaining according to the corresponding target user of row vector chosen in child user-rating matrix
The corresponding cluster group of the row vector of target user；
Similar users rating matrix acquiring unit, in the corresponding cluster group of target user, calculate obtain each scoring it is capable to
Euclidean distance between amount and the row vector of target user, obtains ranking in each Euclidean distance and is located at preset first rank threshold
The corresponding scoring row vector of Euclidean distance before, to form similar users group's rating matrix；
Commercial product recommending row vector acquiring unit, for obtaining similar according to the row vector that respectively scores in similar users group's rating matrix
User group is to the comprehensive grading values of each commodity, to form commercial product recommending row vector；
Information push unit, for comprehensive before being located at preset second rank threshold by the ranking that scores in commercial product recommending row vector
Commodity corresponding to score value are closed to obtain commercial product recommending list, the commercial product recommending list pushes to target user is corresponding to be connect
Receiving end.
8. the driving means according to claim 7 based on reunion class, which is characterized in that user's cluster cell, packet
It includes:
Initial center acquiring unit, for using any one row vector in user-rating matrix as initial cluster center；
Initial clustering group's acquiring unit, for including points according to preset minimum, between obtaining between initial cluster center
Away from the row vector within preset sweep radius, using as initial clustering group；
Group's adjustment unit is clustered, for obtaining user-rating matrix using row vector each in initial clustering group as cluster centre
In with the row vector that the direct density of cluster centre is reachable, density is reachable or density is connected, using as cluster group adjusted.
9. a kind of computer equipment, including memory, processor and it is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that the processor realizes such as claim 1 to 6 when executing the computer program
Any one of described in the method for pushing based on reunion class.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
Sequence, the computer program make the processor execute such as base as claimed in any one of claims 1 to 6 when being executed by a processor
In the method for pushing of reunion class.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811191703.1A CN109166017B (en) | 2018-10-12 | 2018-10-12 | Push method and device based on re-clustering, computer equipment and storage medium |
| PCT/CN2018/125334 WO2020073534A1 (en) | 2018-10-12 | 2018-12-29 | Pushing method and apparatus based on re-clustering, and computer device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811191703.1A CN109166017B (en) | 2018-10-12 | 2018-10-12 | Push method and device based on re-clustering, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109166017A true CN109166017A (en) | 2019-01-08 |
| CN109166017B CN109166017B (en) | 2024-06-04 |
Family
ID=64878156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811191703.1A Active CN109166017B (en) | 2018-10-12 | 2018-10-12 | Push method and device based on re-clustering, computer equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109166017B (en) |
| WO (1) | WO2020073534A1 (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829761A (en) * | 2019-01-31 | 2019-05-31 | 广州视源电子科技股份有限公司 | Commodity selection method, commodity selection device, commodity selection equipment and storage medium |
| CN109903138A (en) * | 2019-02-28 | 2019-06-18 | 华中科技大学 | A kind of individual commodity recommendation method |
| CN109949070A (en) * | 2019-01-28 | 2019-06-28 | 平安科技（深圳）有限公司 | Usage rate of the user appraisal procedure, device, computer equipment and storage medium |
| CN109977299A (en) * | 2019-02-21 | 2019-07-05 | 西北大学 | A kind of proposed algorithm of convergence project temperature and expert's coefficient |
| CN110503506A (en) * | 2019-07-05 | 2019-11-26 | 平安科技（深圳）有限公司 | Item recommendation method, device and medium based on score data |
| CN110717097A (en) * | 2019-09-06 | 2020-01-21 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN111639263A (en) * | 2020-06-03 | 2020-09-08 | 小红书科技有限公司 | Note recommendation method, device and system |
| CN111881191A (en) * | 2020-08-05 | 2020-11-03 | 厦门力含信息技术服务有限公司 | Client portrait key feature mining system and method under mobile internet |
| CN111986005A (en) * | 2020-08-31 | 2020-11-24 | 上海博泰悦臻电子设备制造有限公司 | Activity recommendation method and related equipment |
| CN111984837A (en) * | 2019-05-23 | 2020-11-24 | 浙江口碑网络技术有限公司 | Commodity data processing method, device and equipment |
| CN112307309A (en) * | 2019-07-26 | 2021-02-02 | 中移(苏州)软件技术有限公司 | Information recommendation method and device and storage medium |
| CN112307324A (en) * | 2020-03-04 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Information processing method, apparatus, device and medium |
| CN113763134A (en) * | 2021-09-22 | 2021-12-07 | 携程旅游信息技术（上海）有限公司 | Information recommendation method, system, device and storage medium |
| CN113836310A (en) * | 2021-08-04 | 2021-12-24 | 欧冶工业品股份有限公司 | Knowledge graph driven industrial product supply chain management method and system |
| CN117851650A (en) * | 2024-03-05 | 2024-04-09 | 烟台大学 | Project recommendation method, system and equipment based on user group and reinforcement learning |
| CN118096324A (en) * | 2024-04-23 | 2024-05-28 | 成都帆点创想科技有限公司 | Recommendation model training method and device |
| CN118313922A (en) * | 2024-02-22 | 2024-07-09 | 朴道征信有限公司 | Virtual article pushing method, device, electronic equipment and readable medium |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612583B (en) * | 2020-05-19 | 2024-03-19 | 江苏大学 | Personalized shopping guide system based on clustering |
| CN111859281B (en) * | 2020-05-29 | 2023-09-08 | 石化盈科信息技术有限责任公司 | Article replenishment time prediction method and device, storage medium and electronic equipment |
| CN111813905B (en) * | 2020-06-17 | 2024-05-10 | 平安科技（深圳）有限公司 | Corpus generation method, corpus generation device, computer equipment and storage medium |
| CN111950632B (en) * | 2020-08-12 | 2021-11-09 | 贝壳找房(北京)科技有限公司 | Method, apparatus, medium, and electronic device for implementing cell information clustering |
| CN112036987B (en) * | 2020-09-11 | 2024-04-02 | 杭州海康威视数字技术股份有限公司 | Method and device for determining recommended commodity |
| CN112381598B (en) * | 2020-10-26 | 2023-12-05 | 泰康保险集团股份有限公司 | Product service information pushing method and device |
| CN112633978B (en) * | 2020-12-22 | 2024-03-08 | 重庆大学 | Method, device and equipment for building graphic neural network model and method, device and equipment for recommending commodities |
| CN113822734B (en) * | 2021-01-18 | 2024-07-19 | 北京沃东天骏信息技术有限公司 | Method and device for generating information |
| CN114707067A (en) * | 2022-04-08 | 2022-07-05 | 湖南茶旅云科技有限公司 | Tourism product recommendation method and device, electronic equipment and storage medium |
| CN115018584A (en) * | 2022-06-13 | 2022-09-06 | 浙江理工大学 | Recommendation method integrating comment text subject word emotional tendency and user trust relationship |
| CN116402399B (en) * | 2023-04-14 | 2023-12-29 | 上海锦咏数据科技有限公司 | Business data processing method and system based on artificial intelligence and electronic mall |
| CN116702304B (en) * | 2023-08-08 | 2023-10-20 | 中建五局第三建设有限公司 | Method and device for grouping foundation pit design schemes based on unsupervised learning |
| CN117633165B (en) * | 2023-10-20 | 2024-05-31 | 广州天晟网络信息有限公司 | Intelligent AI customer service dialogue guiding method |
| CN117493913B (en) * | 2023-11-24 | 2024-05-14 | 济南高更食品科技有限公司 | Composite seasoning control method and system based on user feedback |
| CN117350823A (en) * | 2023-12-04 | 2024-01-05 | 北京国双科技有限公司 | Commodity information recommendation method, system, equipment and medium for electronic mall |
| CN117851464B (en) * | 2024-03-07 | 2024-05-14 | 济南道图信息科技有限公司 | Auxiliary analysis method for user behavior pattern for psychological assessment |
| CN117952726B (en) * | 2024-03-27 | 2024-07-23 | 摘星社信息科技(浙江)股份有限公司 | Personalized equity package recommendation system based on operator data analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060208185A1 (en) * | 2005-03-18 | 2006-09-21 | International Business Machines Corporation | Preparing peptide spectra for identification |
| CN103412948A (en) * | 2013-08-27 | 2013-11-27 | 北京交通大学 | Cluster-based collaborative filtering commodity recommendation method and system |
| CN106651546A (en) * | 2017-01-03 | 2017-05-10 | 重庆邮电大学 | Intelligent community oriented electronic commerce information recommendation method |
| CN107861945A (en) * | 2017-11-01 | 2018-03-30 | 平安科技（深圳）有限公司 | Finance data analysis method, application server and computer-readable recording medium |
| CN108205682A (en) * | 2016-12-19 | 2018-06-26 | 同济大学 | It is a kind of for the fusion content of personalized recommendation and the collaborative filtering method of behavior |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017023872A1 (en) * | 2015-07-31 | 2017-02-09 | RCRDCLUB Corporation | Systems and methods of providing recommendations of content items |
| CN106919699A (en) * | 2017-03-09 | 2017-07-04 | 华北电力大学 | A kind of recommendation method for personalized information towards large-scale consumer |
| CN107577786B (en) * | 2017-09-15 | 2019-09-10 | 合肥工业大学 | A kind of matrix decomposition recommendation method based on joint cluster |
| CN107944485B (en) * | 2017-11-17 | 2020-03-06 | 西安电子科技大学 | Recommendation system and method based on cluster group discovery and personalized recommendation system |
| CN108197285A (en) * | 2018-01-15 | 2018-06-22 | 腾讯科技（深圳）有限公司 | A kind of data recommendation method and device |
-
2018
- 2018-10-12 CN CN201811191703.1A patent/CN109166017B/en active Active
- 2018-12-29 WO PCT/CN2018/125334 patent/WO2020073534A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060208185A1 (en) * | 2005-03-18 | 2006-09-21 | International Business Machines Corporation | Preparing peptide spectra for identification |
| CN103412948A (en) * | 2013-08-27 | 2013-11-27 | 北京交通大学 | Cluster-based collaborative filtering commodity recommendation method and system |
| CN108205682A (en) * | 2016-12-19 | 2018-06-26 | 同济大学 | It is a kind of for the fusion content of personalized recommendation and the collaborative filtering method of behavior |
| CN106651546A (en) * | 2017-01-03 | 2017-05-10 | 重庆邮电大学 | Intelligent community oriented electronic commerce information recommendation method |
| CN107861945A (en) * | 2017-11-01 | 2018-03-30 | 平安科技（深圳）有限公司 | Finance data analysis method, application server and computer-readable recording medium |
Non-Patent Citations (1)
| Title |
|---|
| 胡朝举: "基于用户模糊聚类的个性化推荐研究", 《软件导刊》, vol. 17, no. 20, pages 31 - 34 * |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109949070A (en) * | 2019-01-28 | 2019-06-28 | 平安科技（深圳）有限公司 | Usage rate of the user appraisal procedure, device, computer equipment and storage medium |
| CN109949070B (en) * | 2019-01-28 | 2024-03-26 | 平安科技（深圳）有限公司 | User viscosity evaluation method, device, computer equipment and storage medium |
| CN109829761A (en) * | 2019-01-31 | 2019-05-31 | 广州视源电子科技股份有限公司 | Commodity selection method, commodity selection device, commodity selection equipment and storage medium |
| CN109977299A (en) * | 2019-02-21 | 2019-07-05 | 西北大学 | A kind of proposed algorithm of convergence project temperature and expert's coefficient |
| CN109977299B (en) * | 2019-02-21 | 2022-12-27 | 西北大学 | Recommendation algorithm fusing project popularity and expert coefficient |
| CN109903138A (en) * | 2019-02-28 | 2019-06-18 | 华中科技大学 | A kind of individual commodity recommendation method |
| CN111984837B (en) * | 2019-05-23 | 2021-04-23 | 浙江口碑网络技术有限公司 | Commodity data processing method, device and equipment |
| CN111984837A (en) * | 2019-05-23 | 2020-11-24 | 浙江口碑网络技术有限公司 | Commodity data processing method, device and equipment |
| CN110503506A (en) * | 2019-07-05 | 2019-11-26 | 平安科技（深圳）有限公司 | Item recommendation method, device and medium based on score data |
| CN110503506B (en) * | 2019-07-05 | 2022-07-08 | 平安科技（深圳）有限公司 | Item recommendation method, device and medium based on grading data |
| CN112307309B (en) * | 2019-07-26 | 2022-12-06 | 中移(苏州)软件技术有限公司 | Information recommendation method and device and storage medium |
| CN112307309A (en) * | 2019-07-26 | 2021-02-02 | 中移(苏州)软件技术有限公司 | Information recommendation method and device and storage medium |
| CN110717097A (en) * | 2019-09-06 | 2020-01-21 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN112307324A (en) * | 2020-03-04 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Information processing method, apparatus, device and medium |
| CN111639263A (en) * | 2020-06-03 | 2020-09-08 | 小红书科技有限公司 | Note recommendation method, device and system |
| CN111639263B (en) * | 2020-06-03 | 2023-11-24 | 小红书科技有限公司 | Note recommending method, device and system |
| CN111881191B (en) * | 2020-08-05 | 2021-06-11 | 留洋汇(厦门)金融技术服务有限公司 | Client portrait key feature mining system and method under mobile internet |
| CN111881191A (en) * | 2020-08-05 | 2020-11-03 | 厦门力含信息技术服务有限公司 | Client portrait key feature mining system and method under mobile internet |
| CN111986005A (en) * | 2020-08-31 | 2020-11-24 | 上海博泰悦臻电子设备制造有限公司 | Activity recommendation method and related equipment |
| CN113836310A (en) * | 2021-08-04 | 2021-12-24 | 欧冶工业品股份有限公司 | Knowledge graph driven industrial product supply chain management method and system |
| CN113836310B (en) * | 2021-08-04 | 2022-11-29 | 欧冶工业品股份有限公司 | Knowledge graph driven industrial product supply chain management method and system |
| CN113763134A (en) * | 2021-09-22 | 2021-12-07 | 携程旅游信息技术（上海）有限公司 | Information recommendation method, system, device and storage medium |
| CN118313922A (en) * | 2024-02-22 | 2024-07-09 | 朴道征信有限公司 | Virtual article pushing method, device, electronic equipment and readable medium |
| CN117851650A (en) * | 2024-03-05 | 2024-04-09 | 烟台大学 | Project recommendation method, system and equipment based on user group and reinforcement learning |
| CN117851650B (en) * | 2024-03-05 | 2024-05-10 | 烟台大学 | Project recommendation method, system and equipment based on user group and reinforcement learning |
| CN118096324A (en) * | 2024-04-23 | 2024-05-28 | 成都帆点创想科技有限公司 | Recommendation model training method and device |
| CN118096324B (en) * | 2024-04-23 | 2024-07-19 | 成都帆点创想科技有限公司 | Recommendation model training method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109166017B (en) | 2024-06-04 |
| WO2020073534A1 (en) | 2020-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166017A (en) | Method for pushing, device, computer equipment and storage medium based on reunion class | |
| CN109360057A (en) | Information-pushing method, device, computer equipment and storage medium | |
| CN107273438B (en) | Recommendation method, device, equipment and storage medium | |
| TWI743428B (en) | Method and device for determining target user group | |
| Corominas-Murtra et al. | On the origins of hierarchy in complex networks | |
| Shemilt et al. | Pinpointing needles in giant haystacks: use of text mining to reduce impractical screening workload in extremely large scoping reviews | |
| CN107784066A (en) | Information recommendation method, device, server and storage medium | |
| CN108334592B (en) | Personalized recommendation method based on combination of content and collaborative filtering | |
| WO2015188699A1 (en) | Item recommendation method and device | |
| Zhang et al. | Integrating topic and latent factors for scalable personalized review-based rating prediction | |
| CN110119474A (en) | Recommended models training method, the prediction technique based on recommended models and device | |
| US10496691B1 (en) | Clustering search results | |
| Selke et al. | Pushing the boundaries of crowd-enabled databases with query-driven schema expansion | |
| CN109087177A (en) | To the method, apparatus and computer readable storage medium of target user's Recommendations | |
| CN112231584B (en) | Data pushing method and device based on small sample transfer learning and computer equipment | |
| CN108665148B (en) | Electronic resource quality evaluation method and device and storage medium | |
| CN103198089A (en) | Information processing apparatus, information processing method, and program | |
| CN109165975A (en) | Label recommendation method, device, computer equipment and storage medium | |
| CN106168980A (en) | Multimedia resource recommends sort method and device | |
| CN107402961B (en) | Recommendation method and device and electronic equipment | |
| CN105468628B (en) | A kind of sort method and device | |
| CN104298787A (en) | Individual recommendation method and device based on fusion strategy | |
| CN110322318B (en) | Client grouping method, device and computer storage medium | |
| KR20190128246A (en) | Searching methods and apparatus and non-transitory computer-readable storage media | |
| CN110503506A (en) | Item recommendation method, device and medium based on score data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |