CN103544206B - A kind of realization method and system of personalized recommendation - Google Patents
A kind of realization method and system of personalized recommendation Download PDFInfo
- Publication number
- CN103544206B CN103544206B CN201310297189.0A CN201310297189A CN103544206B CN 103544206 B CN103544206 B CN 103544206B CN 201310297189 A CN201310297189 A CN 201310297189A CN 103544206 B CN103544206 B CN 103544206B
- Authority
- CN
- China
- Prior art keywords
- multinuclear
- different group
- user behavior
- nuclear
- data set
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 claims abstract description 98
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 claims abstract description 98
- 210000000299 nuclear matrix Anatomy 0.000 claims abstract description 98
- 230000003542 behavioural effect Effects 0.000 claims abstract description 33
- 238000005457 optimization Methods 0.000 claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 238000013507 mapping Methods 0.000 claims description 20
- 239000012141 concentrate Substances 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 206010034719 Personality change Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000004080 punching Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of realization method and system of personalized recommendation, wherein, method includes step：Original user behavioural characteristic data set is trained by different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria is set up, about subtracts nuclear space data set with the initial different group of multinuclear user behavior examination criteria of foundation, about subtracted data set；Data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria will about be subtracted to be input in nuclear space clustering algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；Classification results collection is obtained to about subtracting data set progress classification and Detection according to different group of multinuclear user behavior examination criteria of optimization and cluster category result, is integrated according to the classification results and recommends the content associated with user behavior feature classification as user.
Description
Technical field
The present invention relates to the information personalized recommendation field of smart machine, more particularly to a kind of realization side of personalized recommendation
Method and system.
Background technology
In the epoch that current network is prevailing, the development of the combination of TV and internet, particularly intelligent television with it is popular into
For a kind of trend.However, in the course that intelligent television develops, how more accurately to enter row information for user（Program or
Person's application etc.）Personalized recommendation, as intelligent television develop an important technology.Many recommendation methods can compare at present
Recommendation process is preferably completed, but still suffers from some defects：There is openness, the complexity of large-scale data processing in the data of recommendation
Property, both the diversity and accuracy of recommendation are difficult to take into account, the ambiguity of user's behavior pattern mining, to the statistics of user behavior
Inaccurately, these problems may all cause the content for recommending user not to be suitable for user, reduce the usage experience of user.
Therefore, prior art has yet to be improved and developed.
The content of the invention
In view of above-mentioned the deficiencies in the prior art, it is an object of the invention to provide a kind of implementation method of personalized recommendation and
System, it is intended to solve that the content that existing recommendation method recommended is inaccurate, it is more complicated, various to there is openness, data processing
The problems such as both property and accuracy are difficult to take into account.
Technical scheme is as follows：
A kind of implementation method of personalized recommendation, wherein, including step：
A, acquisition user behavior characteristic, original user behavioural characteristic number is obtained according to the user behavior characteristic
According to collection；
B, using different group of multinuclear nuclear matrix to original user behavioural characteristic data set carry out nuclear space mapping, obtain nuclear space
Data set；
C, by different group of multinuclear nuclear matrix original user behavioural characteristic data set is trained, sets up initial different group of multinuclear
User behavior examination criteria, about subtracts nuclear space data set with the initial different group of multinuclear user behavior examination criteria of foundation, obtains about
Subtract data set；
D, will about subtract data set, different group of multinuclear nuclear matrix, that initial different group of multinuclear user behavior examination criteria is input to core is empty
Between in clustering algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
E, according to optimization different group of multinuclear user behavior examination criteria and cluster category result classify to about subtracting data set
Detection obtains classification results collection, is integrated according to the classification results and recommended as user in associated with user behavior feature classification
Hold.
The method of described intelligent television personalized recommendation, wherein, the step B is specifically included：
B1, one group of kernel function is pre-set, calculate nuclear moment of each kernel function on original user behavioural characteristic data set
Battle array, and determine the optimal running parameter of each nuclear matrix；
B2, the optimal running parameter for combining each nuclear matrix using the interior point method of Semidefinite Programming calculate different group of multinuclear nuclear moment
The optimum combination coefficient of battle array；
B3, according to the optimum combination coefficient by each nuclear matrix carry out linear combination, obtain different group of multinuclear nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
B4, using the different group of multinuclear nuclear matrix to user behavior characteristic carry out nuclear space mapping, obtain nuclear space number
According to collection.
The method of described intelligent television personalized recommendation, wherein, the step C is specifically included：
C1, the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that initial different group of multinuclear user behavior
Examination criteria；
C2, initial different group of multinuclear user behavior examination criteria about subtracted into nuclear space data set, about subtracted data set.
The method of described intelligent television personalized recommendation, wherein, the step D is specifically included：
D1, by data object of the about subtrahend according to concentrationIt is mapped in nuclear space, is mapped with different group of multinuclear nuclear matrix
Data object afterwards is；
D2, from about subtrahend according to concentrate choose m object as quasi- initial center point, with quasi- initial center point to data objectCarry out Preliminary division；
D3, it is directed at initial center point in the classification that Preliminary division is obtained and is adjusted and obtains final initial center point；
D4, according to final initial center point again to data objectProgress divides the classification optimized as cluster
Category result；
D5, increase central point Candidate Set are replaced to initial center point, and continuous iteration updates central point until central point
No longer change, obtain optimizing different group of multinuclear user behavior examination criteria.
The method of described intelligent television personalized recommendation, wherein, in the step D3, adjust the mode of initial center point
For：Each data object in each classification is set as class center, each class center is calculated and is counted with other in respective classes
According to the distance between object sum so that be used as final initial center point apart from the minimum class center of sum.
The method of described intelligent television personalized recommendation, it is characterised in that in the step D5, central point Candidate Set
Acquisition process includes：Calculating about subtrahend, in the distance of nuclear space, central point is obtained so as to collect according to each data object of concentration
Candidate Set.
A kind of personalized recommendation realizes system, wherein, including：
Original user behavioural characteristic data set acquisition module, for obtaining user behavior characteristic, according to the user
Behavioural characteristic data obtain original user behavioural characteristic data set；
Nuclear space mapping block, it is empty for carrying out core to original user behavioural characteristic data set using different group of multinuclear nuclear matrix
Between map, obtain nuclear space data set；
Data are intensive to subtract module, for being instructed by different group of multinuclear nuclear matrix to original user behavioural characteristic data set
Practice, set up initial different group of multinuclear user behavior examination criteria, about subtracted with the initial different group of multinuclear user behavior examination criteria of foundation
Nuclear space data set, is about subtracted data set；
Optimization module, for will about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria
It is input in nuclear space clustering algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
Recommending module, for according to optimization different group of multinuclear user behavior examination criteria and cluster category result to about subtrahend evidence
Collection carries out classification and Detection and obtains classification results collection, is integrated according to the classification results and recommended and user behavior feature classification phase as user
The content of association.
Described personalized recommendation realizes system, wherein, the nuclear space mapping block specifically includes：
Nuclear matrix acquiring unit, for pre-setting one group of kernel function, calculates each kernel function special in original user behavior
The nuclear matrix on data set is levied, and determines the optimal running parameter of each nuclear matrix；
Optimum combination coefficient calculation unit, the optimal work of each nuclear matrix is combined for the interior point method using Semidefinite Programming
Parameter calculates the optimum combination coefficient of different group of multinuclear nuclear matrix；
Linear combination unit, for each nuclear matrix to be carried out into linear combination according to the optimum combination coefficient, obtains different
Group multinuclear nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
Nuclear space map unit, reflects for carrying out nuclear space to user behavior characteristic using the different group of multinuclear nuclear matrix
Penetrate, obtain nuclear space data set.
Described personalized recommendation realizes system, wherein, the intensive module that subtracts of the data includes：
Primary standard determining unit, for the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that just
Begin different group of multinuclear user behavior examination criteria；
About subtract unit, for initial different group of multinuclear user behavior examination criteria about to be subtracted into nuclear space data set, about subtracted
Data set.
Described personalized recommendation realizes system, wherein, the optimization module includes：
Data object map unit, for by data object of the about subtrahend according to concentrationIt is mapped to different group of multinuclear nuclear matrix
In nuclear space, the data object after being mapped is；
Preliminary division unit, for choosing m object as quasi- initial center point from about subtrahend according to concentrating, with it is accurate initially in
Heart point is to data objectCarry out Preliminary division；
Initial center point adjustment unit, is adjusted for being directed at initial center point in the classification that Preliminary division is obtained
To final initial center point；
Repartition unit, for according to final initial center point again to data objectProgress, which is divided, to be optimized
Classification be used as cluster category result；
Criteria optimization unit, is replaced for increasing central point Candidate Set to initial center point, during continuous iteration updates
Heart point no longer changes up to central point, obtains optimizing different group of multinuclear user behavior examination criteria.
Beneficial effect：The present invention carries out nuclear space mapping to user behavior characteristic using different group of multinuclear nuclear matrix, and
Clustered using the nuclear space clustering algorithm of optimization come the data set after to about subtracting, obtain cluster category result and optimize different
Group multinuclear user behavior examination criteria, and then classification results collection is obtained to the hobby tagsort detection of user, it is that user recommends
Corresponding content.The present invention had both improved the satisfaction that the accuracy, high efficiency and user of content recommendation are used, and can reduce
The time complexity of data training, improves recommendation efficiency.
Brief description of the drawings
Fig. 1 is a kind of flow chart of the implementation method preferred embodiment of personalized recommendation of the invention.
Fig. 2 is the design drawing of recommendation application interface in the embodiment of the present invention.
The particular flow sheet that Fig. 3 is step S102 in method shown in Fig. 1.
The particular flow sheet that Fig. 4 is step S103 in method shown in Fig. 1.
The particular flow sheet that Fig. 5 is step S104 in method shown in Fig. 1.
Fig. 6 is a kind of structured flowchart for realizing system preferred embodiment of personalized recommendation of the invention.
Fig. 7 is the concrete structure block diagram of system shown in Figure 6 center space mapping module.
Fig. 8 is the intensive concrete structure block diagram for subtracting module of data in system shown in Figure 6.
Fig. 9 is the concrete structure block diagram of optimization module in system shown in Figure 6.
Embodiment
The present invention provides a kind of realization method and system of personalized recommendation, for make the purpose of the present invention, technical scheme and
Effect is clearer, clear and definite, and the present invention is described in more detail below.It should be appreciated that specific embodiment described herein
Only to explain the present invention, it is not intended to limit the present invention.
Referring to Fig. 1, Fig. 1 is a kind of implementation method of personalized recommendation of the invention, as illustrated, its preferred embodiment bag
Include step：
S101, acquisition user behavior characteristic, obtain original user behavior special according to the user behavior characteristic
Levy data set；
S102, using different group of multinuclear nuclear matrix to original user behavioural characteristic data set carry out nuclear space mapping, obtain core
Space data sets；
S103, by different group of multinuclear nuclear matrix original user behavioural characteristic data set is trained, sets up initial different group
Multinuclear user behavior examination criteria, about subtracts nuclear space data set with the initial different group of multinuclear user behavior examination criteria of foundation, obtains
To about subtracting data set；
S104, it will about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria and be input to core
In Spatial Clustering, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
S105, according to optimization different group of multinuclear user behavior examination criteria and cluster category result divided about subtracting data set
Class detection obtains classification results collection, is integrated according to the classification results and recommended as user in associated with user behavior feature classification
Hold.
Below by taking the personalized recommendation of intelligent television as an example, above-mentioned steps are described in detail respectively.
In step S101, user enters recommendation application interface first, user behavior characteristic is obtained, so as to combine
Obtain original user behavioural characteristic data set.If first entering into recommendation application interface, then default recommendation content is shown, recommended
The layout of application interface as shown in Fig. 2 can be adjusted according to actual needs certainly.
Described user behavior characteristic includes：
1st, user uses access vestige during intelligent television；2nd, user proposes to intelligent television improvement requirement and use anti-
Present opinion；3rd, user uses the evaluation information or mark after intelligent television application software or viewing movie and television contents；4th, use
Family has access to the vestige of the content recommendation of intelligent television by other approach；5th, the label that leaves when user is using network.It is above-mentioned
Content can as user behavior characteristic main source, other users row can be obtained from other modes as needed certainly
It is characterized data.
In step s 102, a different group of multinuclear nuclear matrix is first obtained, original user behavioural characteristic data set is carried out
Nuclear space maps, in the present invention, and different group of multinuclear nuclear matrix is that multiple different nuclear matrix are carried out into the new of linear combination formation
Nuclear matrix.
Because user use intelligent television when it is random, openness the features such as, the use obtained in step S101
Family behavioural characteristic data have higher-dimension and nonlinear feature, cause preferably recognize the behavioural characteristic of user, reduce
The validity of content recommendation.And the dimension that higher-dimension and nonlinear data can be passed through mapping processing reduction data set by nuclear space
Number, while making data become linear separability.The data distribution obtained after the characteristics of having different due to each kernel function, mapping
Be also different, the embodiment of the present invention is exactly that the corresponding nuclear matrix of multiple different kernel functions is carried out into linear combination, with for
The characteristics of user behavior characteristic, improves the validity of content recommendation.
Specifically, as shown in figure 3, the step S102 can specifically be refined as following steps：
S201, one group of kernel function is pre-set, calculate core of each kernel function on original user behavioural characteristic data set
Matrix, and determine the optimal running parameter of each nuclear matrix；
In the present embodiment, the kernel function used includes gaussian kernel function, Polynomial kernel function, perceptron kernel function,
The corresponding nuclear matrix of above-mentioned kernel function is subjected to linear combination and obtains different group of multinuclear nuclear matrix.
S202, the optimal running parameter for combining each nuclear matrix using the interior point method of Semidefinite Programming calculate different group of multinuclear core
The optimum combination coefficient of matrix；
S203, according to the optimum combination coefficient by each nuclear matrix carry out linear combination, obtain different group of multinuclear nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
S204, using the different group of multinuclear nuclear matrix to user behavior characteristic carry out nuclear space mapping, obtain nuclear space
Data set, always according to theorem in Euclid space range formula, calculate the core that obtains each data in nuclear space data set in nuclear space away from
From core distance refers to distance between points of the data in nuclear space.
In step s 103, user behavior characteristic is trained by different group of multinuclear nuclear matrix and obtains one initially
Different group of multinuclear user behavior examination criteria, then about subtracts initial nuclear space using the initial different group of multinuclear user behavior examination criteria
Data set, specifically, as shown in figure 4, step S103 can be refined as following steps：
S301, the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that initial different group of multinuclear user's row
For examination criteria；
S302, initial different group of multinuclear user behavior examination criteria about subtracted into nuclear space data set, about subtracted data set.
About subtract data set and be also referred to as border support data set, what it was represented is in contingency table directrix or distance classification mark
The closer data acquisition system of directrix, these data are influenceed than larger on data tagsort result, these data can be referred to as into spy
Vector is levied, and the data about subtracted influence and little on classification results, belong to redundant data, so needing to weed out these redundancies
Data.
In step S104, by it is foregoing obtain about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user's row
It is input to for standard in nuclear space clustering algorithm, nuclear space mapping is carried out to about subtracting data set using different group of multinuclear nuclear matrix, from
And obtain optimizing nuclear space data set.As shown in figure 5, it specifically includes step：
S401, by data object of the about subtrahend according to concentrationIt is mapped in nuclear space, is reflected with different group of multinuclear nuclear matrix
Data object after penetrating is；
S402, from about subtrahend according to concentrate choose m object as quasi- initial center point, with quasi- initial center point to data pair
AsCarry out Preliminary division；
S403, it is directed at initial center point in the classification that Preliminary division is obtained and is adjusted and obtains final initial center
Point；
S404, according to final initial center point again to data objectDivide the classification conduct optimized
Cluster category result；
S405, increase central point Candidate Set are replaced to initial center point, and continuous iteration updates central point until center
Point no longer changes, and obtains optimizing different group of multinuclear user behavior examination criteria.
In the present embodiment, nuclear space clustering algorithm is that original nuclear space K-means clustering algorithms are optimized to change
Enter what is obtained.
It is mainly reflected in：
1st, in step S403, small range has been carried out to initial center point and optimized and revised, clustered in original nuclear space
In algorithm, initial center point is randomly selected, and the embodiment of the present invention is then adjusted to initial center point：Randomly select
Surely after initial center point, data object is divided, clicked through dividing alignment initial center in obtained each classification
Row adjustment, it is assumed that each data object in each classification is class center, is calculated in each class center and respective classes
The distance between other data objects sum, makes it apart from the minimum class center of sum as final initial center point.
2nd, in step S405, increase central point Candidate Set is replaced to center point, if desired replaces one of those
Central point, then around this central point（I.e. in this classification or in close classification）There can be proper candidate point,
So only need to search for these candidate points can just complete the replacement step to central point.Specifically can be by being calculated in abovementioned steps
The mode of core distance collects candidate point, i.e.,：Calculate about subtrahend according to concentration each data object nuclear space distance so that
Collection obtains central point Candidate Set（The set of candidate point）.The collection to candidate point is all first completed when updating central point every time, is waited
Reconnaissance is integrated into be continuously updated in interative computation to be increased always, with the increase of iterations, the search model of candidate point set
Enclosing can be incremented on the object of all non-central points in data set, so as to cover whole data set, improve clustering precision.
Cluster iterations is greatly reduced in nuclear space clustering algorithm after improvement, and the time complexity reduced is carried
High clustering precision.
The input and output of nuclear space clustering algorithm in the present invention are as follows：
Input：The number m of classification, different group of multinuclear nuclear matrixWith the training dataset for including n object, it is initial different
Group multinuclear user behavior examination criteria.
Output：M classification, different group of multinuclear user behavior examination criteria of optimization.
In step S105, by optimizing different group of multinuclear user behavior examination criteria and cluster category result to about subtrahend evidence
Collection carries out classification and Detection, obtains classification results collection, is integrated according to classification results and recommended as user in related to behavioural characteristic classification
Hold.
In this step, user behavior feature classification has been obtained after carrying out classification and Detection, and then can obtain classification results collection, i.e.,
The content related to user behavior feature classification.Classification results therein concentrate the recommending data included to be applied according to recommendation
It is configured the need for interface, such as when recommendation program is applied, this classification results, which is concentrated, includes program poster, program ID, section
The information such as mesh details, program broadcast source.
Application is recommended to obtain the recommending data that classification results are concentrated by downloading, being then shown to these recommending datas should
With interface, user is showed, the process of personalized recommendation is completed.
The embodiment of the present invention is by the way that by data set High Dimensional Mapping, to nuclear space, the characteristics of adding to data similarity is gathered
Collection, so that the problem of preferably resolving Deta sparseness；The embodiment of the present invention utilizes the linear combination technology of kernel function, fully
The characteristics of having played each kernel function, obtains different group of multinuclear nuclear matrix, large-scale data is clustered and about subtract processing, are obtained
More accurate cluster race and user behavior feature, improve the accuracy of recommendation results；The embodiment of the present invention passes through to higher-dimension
Data carry out nuclear space mapping, eliminate redundant data, obtain useful feature vector data（About subtract data set）, significantly drop
The low dimension and the time complexity of proposed algorithm of data, improves recommendation efficiency；The embodiment of the present invention is also recommended cluster
Algorithm is optimized, and reduces the iterations of cluster, while improving clustering precision；The present invention is directed to existing recommendation side
Excavation of the method to user behavior has certain limitation, and keeping punching to excavate using user behavior characteristic about subtracts processing,
Recommendation service is done for user under the conditions of more more accurately data messages, so as to improve users' satisfaction degree, and makes individual character
Change recommendation method more intelligent, more personalized.
Based on the above method, what the present invention also provided a kind of personalized recommendation realizes system, as shown in fig. 6, it is preferably real
Applying example includes：
Original user behavioural characteristic data set acquisition module 100, for obtaining user behavior characteristic, is used according to described
Family behavioural characteristic data obtain original user behavioural characteristic data set；
Nuclear space mapping block 200, for being carried out using different group of multinuclear nuclear matrix to original user behavioural characteristic data set
Nuclear space maps, and obtains nuclear space data set；
Data are intensive to subtract module 300, for being carried out by different group of multinuclear nuclear matrix to original user behavioural characteristic data set
Training, sets up initial different group of multinuclear user behavior examination criteria, with the initial different group of multinuclear user behavior examination criteria of foundation about
Subtract nuclear space data set, about subtracted data set；
Optimization module 400, for will about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior detection
Standard is input in nuclear space clustering algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
Recommending module 500, for according to optimization different group of multinuclear user behavior examination criteria and cluster category result to about subtracting
Data set carries out classification and Detection and obtains classification results collection, is integrated according to the classification results and recommended and user behavior feature class as user
Not Xiang Guanlian content.
Further, as shown in fig. 7, the nuclear space mapping block 200 is specifically included：
Nuclear matrix acquiring unit 210, for pre-setting one group of kernel function, calculates each kernel function in original user behavior
Nuclear matrix on characteristic data set, and determine the optimal running parameter of each nuclear matrix；
Optimum combination coefficient calculation unit 220, for combining the optimal of each nuclear matrix using the interior point method of Semidefinite Programming
Running parameter calculates the optimum combination coefficient of different group of multinuclear nuclear matrix；
Linear combination unit 230, for each nuclear matrix to be carried out into linear combination according to the optimum combination coefficient, is obtained
Different group of multinuclear nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
Nuclear space map unit 240, it is empty for carrying out core to user behavior characteristic using the different group of multinuclear nuclear matrix
Between map, obtain nuclear space data set.
Further, as shown in figure 8, the intensive module 300 that subtracts of the data includes：
Primary standard determining unit 310, for the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that
Initial different group of multinuclear user behavior examination criteria；
About subtract unit 320, for initial different group of multinuclear user behavior examination criteria about to be subtracted into nuclear space data set, obtain about
Subtract data set.
Further, as shown in figure 9, the optimization module 400 includes：
Data object map unit 410, for by data object of the about subtrahend according to concentrationReflected with different group of multinuclear nuclear matrix
It is mapped in nuclear space, the data object after being mapped is；
Preliminary division unit 420, for choosing m object as quasi- initial center point from about subtrahend according to concentration, with the beginning of standard
Beginning, central point was to data objectCarry out Preliminary division；
Initial center point adjustment unit 430, is adjusted for being directed at initial center point in the classification that Preliminary division is obtained
It is whole to obtain final initial center point；
Repartition unit 440, for according to final initial center point again to data objectDivided
The classification of optimization is used as cluster category result；
Criteria optimization unit 450, is replaced for increasing central point Candidate Set to initial center point, and continuous iteration updates
Central point no longer changes up to central point, obtains optimizing different group of multinuclear user behavior examination criteria.On above-mentioned modular unit
Ins and outs, have been described in detail, therefore repeat no more in method above.
In summary, the embodiment of the present invention is reflected using different group of multinuclear nuclear matrix to user behavior characteristic progress nuclear space
Penetrate, and clustered using the nuclear space clustering algorithm of optimization come the data set after to about subtracting, obtain clustering category result and
Optimize different group of multinuclear user behavior examination criteria, and then classification results collection obtained to the hobby tagsort detection of user, for
Corresponding content is recommended at family, and what the accuracy, high efficiency and user that the embodiment of the present invention had both improved content recommendation were used expires
Meaning degree, can reduce the time complexity of data training, improve recommendation efficiency again.
It should be appreciated that the application of the present invention is not limited to above-mentioned citing, for those of ordinary skills, can
To be improved or converted according to the above description, all these modifications and variations should all belong to the guarantor of appended claims of the present invention
Protect scope.
Claims (6)
1. a kind of implementation method of personalized recommendation, it is characterised in that including step：
A, acquisition user behavior characteristic, original user behavioural characteristic data are obtained according to the user behavior characteristic
Collection；
B, using different group of multinuclear nuclear matrix to original user behavioural characteristic data set carry out nuclear space mapping, obtain nuclear space data
Collection；
The step B is specifically included：B1, one group of kernel function is pre-set, calculate each kernel function in original user behavioural characteristic
Nuclear matrix on data set, and determine the optimal running parameter of each nuclear matrix；
B2, the optimal running parameter for combining each nuclear matrix using the interior point method of Semidefinite Programming calculate different group of multinuclear nuclear matrix
Optimum combination coefficient；
B3, according to the optimum combination coefficient by each nuclear matrix carry out linear combination, obtain different group of multinuclear nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
B4, using the different group of multinuclear nuclear matrix to user behavior characteristic carry out nuclear space mapping, obtain nuclear space data set；
C, by different group of multinuclear nuclear matrix original user behavioural characteristic data set is trained, sets up initial different group of multinuclear user
Behavioral value standard, about subtracts nuclear space data set with the initial different group of multinuclear user behavior examination criteria of foundation, obtains about subtrahend
According to collection；
D, it will about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria and be input to nuclear space and gather
In class algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
E, according to optimization different group of multinuclear user behavior examination criteria and cluster category result to about subtract data set carry out classification and Detection
Classification results collection is obtained, is integrated according to the classification results and recommends the content associated with user behavior feature classification as user；
The step D is specifically included：
D1, by data object of the about subtrahend according to concentrationIt is mapped to different group of multinuclear nuclear matrix in nuclear space, the number after being mapped
It is according to object；
D2, from about subtrahend according to concentrate choose m object as quasi- initial center point, with quasi- initial center point to data objectEnter
Row Preliminary division；
D3, it is directed at initial center point in the classification that Preliminary division is obtained and is adjusted and obtains final initial center point；
D4, according to final initial center point again to data objectThe classification optimized divide as cluster classification
As a result；
D5, increase central point Candidate Set are replaced to initial center point, continuous iteration update central point until central point no longer
Change, obtains optimizing different group of multinuclear user behavior examination criteria.
2. the implementation method of personalized recommendation according to claim 1, it is characterised in that the step C is specifically included：
C1, the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that initial different group of multinuclear user behavior detection
Standard；
C2, initial different group of multinuclear user behavior examination criteria about subtracted into nuclear space data set, about subtracted data set.
3. the implementation method of personalized recommendation according to claim 1, it is characterised in that in the step D3, adjustment is just
The mode of beginning central point is：Each data object in each classification is set as class center, each class center and phase is calculated
Answer the distance between the data object of other in classification sum so that apart from the minimum class center of sum as it is final it is initial in
Heart point.
4. the implementation method of personalized recommendation according to claim 1, it is characterised in that in the step D5, central point
The acquisition process of Candidate Set includes：Calculate about subtrahend according to concentration each data object nuclear space distance, so as to collect
To central point Candidate Set.
5. a kind of personalized recommendation realizes system, it is characterised in that including：
Original user behavioural characteristic data set acquisition module, for obtaining user behavior characteristic, according to the user behavior
Characteristic obtains original user behavioural characteristic data set；
Nuclear space mapping block, reflects for carrying out nuclear space to original user behavioural characteristic data set using different group of multinuclear nuclear matrix
Penetrate, obtain nuclear space data set；
The nuclear space mapping block is specifically included：Nuclear matrix acquiring unit, for pre-setting one group of kernel function, calculates each
Nuclear matrix of the kernel function on original user behavioural characteristic data set, and determine the optimal running parameter of each nuclear matrix；
Optimum combination coefficient calculation unit, the optimal running parameter of each nuclear matrix is combined for the interior point method using Semidefinite Programming
Calculate the optimum combination coefficient of different group of multinuclear nuclear matrix；
Linear combination unit, for each nuclear matrix to be carried out into linear combination according to the optimum combination coefficient, is obtained more than different group
Core nuclear matrix,For nuclear matrix,For the coefficient of nuclear matrix；
Nuclear space map unit, for carrying out nuclear space mapping to user behavior characteristic using the different group of multinuclear nuclear matrix,
Obtain nuclear space data set；
Data are intensive to subtract module, for being trained by different group of multinuclear nuclear matrix to original user behavioural characteristic data set, builds
Vertical initial different group of multinuclear user behavior examination criteria, nuclear space is about subtracted with the initial different group of multinuclear user behavior examination criteria of foundation
Data set, is about subtracted data set；
Optimization module, for will about subtract data set, different group of multinuclear nuclear matrix, initial different group of multinuclear user behavior examination criteria input
Into nuclear space clustering algorithm, output different group of multinuclear user behavior examination criteria of optimization and cluster category result；
Recommending module, for being entered according to different group of multinuclear user behavior examination criteria of optimization and cluster category result to about subtracting data set
Row classification and Detection obtains classification results collection, according to the classification results integrate as user recommend it is associated with user behavior feature classification
Content；
The optimization module includes：
Data object map unit, for by data object of the about subtrahend according to concentrationCore is mapped to different group of multinuclear nuclear matrix empty
Between in, the data object after being mapped is；
Preliminary division unit, for choosing m object as quasi- initial center point from about subtrahend according to concentration, with quasi- initial center point
To data objectCarry out Preliminary division；
Initial center point adjustment unit, is adjusted and obtains most for being directed at initial center point in the classification that Preliminary division is obtained
Whole initial center point；
Repartition unit, for according to final initial center point again to data objectDivide the class optimized
Category result Zuo Wei not clustered；
Criteria optimization unit, is replaced for increasing central point Candidate Set to initial center point, and continuous iteration updates central point
Until central point no longer changes, obtain optimizing different group of multinuclear user behavior examination criteria.
6. personalized recommendation according to claim 5 realizes system, it is characterised in that the data are intensive subtracts module bag
Include：
Primary standard determining unit, for the middle solving result by different group of multinuclear nuclear matrix and Semidefinite Programming, it is determined that initial different
Group multinuclear user behavior examination criteria；
About subtract unit, for initial different group of multinuclear user behavior examination criteria about to be subtracted into nuclear space data set, obtain about subtrahend evidence
Collection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310297189.0A CN103544206B (en) | 2013-07-16 | 2013-07-16 | A kind of realization method and system of personalized recommendation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310297189.0A CN103544206B (en) | 2013-07-16 | 2013-07-16 | A kind of realization method and system of personalized recommendation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103544206A CN103544206A (en) | 2014-01-29 |
| CN103544206B true CN103544206B (en) | 2017-09-15 |
Family
ID=49967659
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310297189.0A Expired - Fee Related CN103544206B (en) | 2013-07-16 | 2013-07-16 | A kind of realization method and system of personalized recommendation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103544206B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104954873B (en) * | 2014-03-26 | 2018-10-26 | Tcl集团股份有限公司 | A kind of smart television video method for customizing and system |
| CN108829846B (en) * | 2018-06-20 | 2021-09-10 | 中国联合网络通信集团有限公司 | Service recommendation platform data clustering optimization system and method based on user characteristics |
| CN112057079B (en) * | 2020-08-07 | 2022-07-29 | 中国科学院深圳先进技术研究院 | Behavior quantification method and terminal based on state and map |
| CN112887743B (en) * | 2021-01-19 | 2023-04-07 | 北京映客芝士网络科技有限公司 | Information pushing method and device for live broadcast platform, electronic equipment and storage medium |
| CN113852867B (en) * | 2021-05-27 | 2023-09-08 | 天翼数字生活科技有限公司 | Method and device for recommending programs based on kernel density estimation |
| CN115563186A (en) * | 2022-05-23 | 2023-01-03 | 谢发泽 | User behavior intention output method based on big data analysis and big data system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289478A (en) * | 2011-08-01 | 2011-12-21 | 江苏广播电视大学 | System and method for recommending video on demand based on fuzzy clustering |
| CN103077220A (en) * | 2012-12-29 | 2013-05-01 | 中国科学院深圳先进技术研究院 | User group correlation degree-based personalized recommendation method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IL134943A0 (en) * | 2000-03-08 | 2001-05-20 | Better T V Technologies Ltd | Method for personalizing information and services from various media sources |
| US8620764B2 (en) * | 2011-05-09 | 2013-12-31 | Telefonaktiebolaget L M Ericsson (Publ) | Method for providing a recommendation such as a personalized recommendation, recommender system, and computer program product comprising a recommender computer program |
-
2013
- 2013-07-16 CN CN201310297189.0A patent/CN103544206B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289478A (en) * | 2011-08-01 | 2011-12-21 | 江苏广播电视大学 | System and method for recommending video on demand based on fuzzy clustering |
| CN103077220A (en) * | 2012-12-29 | 2013-05-01 | 中国科学院深圳先进技术研究院 | User group correlation degree-based personalized recommendation method and system |
Non-Patent Citations (2)
| Title |
|---|
| 一种高效的K-medoids聚类算法;夏宁霞 等;《计算机应用研究》;20101231;第27卷(第12期);4517-4519 * |
| 用核K-means聚类和半定规划SVM实现垃圾标签检测;覃华 等;《计算机应用研究》;20130430;第30卷(第4期);1179-1182,1186 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103544206A (en) | 2014-01-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103544206B (en) | A kind of realization method and system of personalized recommendation | |
| CN104679743B (en) | A kind of method and device of the preference pattern of determining user | |
| EP4181026A1 (en) | Recommendation model training method and apparatus, recommendation method and apparatus, and computer-readable medium | |
| CN103544623A (en) | Web service recommendation method based on user preference feature modeling | |
| CN108090208A (en) | Fused data processing method and processing device | |
| CN110705448A (en) | Human body detection method and device | |
| CN108062573A (en) | Model training method and device | |
| CN106776928A (en) | Recommend method in position based on internal memory Computational frame, fusion social environment and space-time data | |
| CN105631415A (en) | Video pedestrian recognition method based on convolution neural network | |
| CN107851174A (en) | The apparatus and method of linguistic indexing of pictures | |
| CN102708130A (en) | Scalable engine that computes user micro-segments for offer matching | |
| CN104102696A (en) | Content recommendation method and device | |
| CN105095219A (en) | Weibo recommendation method and terminal | |
| CN107122447A (en) | The network searching system and control method of a kind of multi-data source fusion based on preference | |
| CN105430100A (en) | User value assessment method and device, and application method and system thereof | |
| CN104778237A (en) | Individual recommending method and system based on key users | |
| Miao et al. | Label3DMaize: toolkit for 3D point cloud data annotation of maize shoots | |
| CN104657457B (en) | A kind of user evaluates data processing method, video recommendation method and the device of video | |
| CN102073718B (en) | System and method for explaining, erasing and modifying search result in probabilistic database | |
| CN108363730A (en) | A kind of content recommendation method, system and terminal device | |
| CN108376164A (en) | A kind of methods of exhibiting and device of potentiality main broadcaster | |
| CN106919706A (en) | Data updating method and device | |
| CN109190040A (en) | Personalized recommendation method and device based on coevolution | |
| CN105159918A (en) | Trust correlation based microblog network community discovery method | |
| CN116739739A (en) | Loan amount evaluation method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 516001 technical center, 8 floor, TCL Arts crafts emporium, 6 South Goose Road, Huizhou, Guangdong Patentee after: TCL Technology Group Co.,Ltd. Address before: 516001 technical center, 8 floor, TCL Arts crafts emporium, 6 South Goose Road, Huizhou, Guangdong Patentee before: TCL Corp. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200416 Address after: 518000 Guangdong city of Shenzhen province Qianhai Shenzhen Hong Kong cooperation zone before Bay Road No. 1 building 201 room A (located in Shenzhen Qianhai business secretary Co. Ltd.) Patentee after: SHENZHEN LEINIAO NETWORK MEDIA CO.,LTD. Address before: 516001 technical center, 8 floor, TCL Arts crafts emporium, 6 South Goose Road, Huizhou, Guangdong Patentee before: TCL Technology Group Co.,Ltd. |
|
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170915 |