CN107220303A - Recommendation method, device and the computer-readable medium of a kind of application - Google Patents
Recommendation method, device and the computer-readable medium of a kind of application Download PDFInfo
- Publication number
- CN107220303A CN107220303A CN201710324504.2A CN201710324504A CN107220303A CN 107220303 A CN107220303 A CN 107220303A CN 201710324504 A CN201710324504 A CN 201710324504A CN 107220303 A CN107220303 A CN 107220303A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- matrix
- application
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- 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
Abstract
The embodiment of the invention discloses the recommendation method of a kind of recommendation method of application, device and a kind of application of computer-readable medium, this method includes：The operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；Calculate the approximate matrix A of the principal matrix AK；Pass through the approximate matrix AKObtain user's similarity matrix；The neighbor user intersection of specific user is obtained with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value；According to the specific user, the operation history data of the heart and the neighbor user intersection calculate prediction scoring of the specific user to intended application in the application；According to the prediction scoring generation application recommendation list.The problem of present invention can solve big data matrix computations amount to a certain extent, shortens the calculating time, while improving application recommends precision.
Description
Technical field
The present invention relates to recommendation method, device and computer-readable Jie of communication technical field, more particularly to a kind of application
Matter.
Background technology
With the expansion of internet information amount and developing rapidly for ecommerce, problem of information overload is but more serious.
Either information consumer or information producer encounter very big challenge：On the one hand, for information consumer, he
Be increasingly difficult to from the data of magnanimity fast and accurately find to oneself valuable information；And on the other hand, for information
For the producer, they are difficult that the information for allowing oneself to produce is shown one's talent in mass data, allow information consumer to be concerned about certainly
Oneself.
How this problem is solvedThus commending system arises at the historic moment, and it is the important tool for solving the problems, such as this class.Push away
The recommending system of the task is exactly contact details consumer and information producer, and one side help information consumer has found useful to oneself
Information, the another aspect help information producer production information can conveniently be presented in letter interested in the information
Cease in face of consumer, so as to realize the win-win situation of both information consumer and information producer.
Two kinds of conventional implementations of SlopeOne algorithms：Weighted SlopeOne and Bi-Polar SlopeOne,
Its shortcoming is that algorithm does not account for similarity problem in calculating process.And it is bright according to the research of document, compared to traditional
Collaborative filtering, when data become than it is sparse when, SlopeOne algorithms are not outstanding.
Further, since very big using the data volume at center, it includes Apply Names, label, using description, icon, under
Carrying capacity, and the operation history of user include, and download, comment, scoring and temperature seniority among brothers and sisters etc..SlopeOne collaborative filterings are
All items attribute and user's operation history behavioral data are subjected to corresponding matrix operation, in the situation that data are constant
Under, server cluster, which calculates all data, certain difficulty.
The content of the invention
The embodiments of the invention provide recommendation method, device and the computer-readable medium of a kind of application, it is intended to no matter counts
According to dense or sparse, it can improve using the precision recommended, lift Consumer's Experience.
In view of this, there is provided a kind of recommendation method of application, the recommendation of the application for first aspect of the embodiment of the present invention
Method comprises the following steps：
The operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；
Calculate the approximate matrix A of the principal matrix AK；
Pass through the approximate matrix AKObtain user's similarity matrix；
The neighbour of specific user is obtained with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value
Occupy user's intersection；
According to the specific user, the operation history data of the heart and the neighbor user intersection calculate the spy in the application
Determine prediction of the user to intended application to score；
According to the prediction scoring generation application recommendation list.
In a kind of possible design, the approximate matrix A for calculating the principal matrix AKIncluding：
By principal matrix A described in singular value decomposition, the first matrix U, the second matrix ∑, the 3rd matrix V are obtained；
Dimension k is calculated, and obtains the first submatrix U accordinglyK, the second submatrix ∑K, the 3rd submatrix VK；
The approximate matrix AKFor the first submatrix UK, the second submatrix ∑KAnd the 3rd submatrix VKInversion
The product of matrix.
In a kind of possible design, the calculating dimension k includes：If the principal matrix A is n × m matrix,
K is equal with n or m intermediate values less one.
It is described to pass through the approximate matrix A in a kind of possible designKObtaining user's similarity matrix includes：
User's similarity matrix is obtained by below equation：
Wherein, simijRepresent user i and user j similarity.
It is described to pass through the approximate matrix A in a kind of possible designKObtaining user's similarity matrix includes：
The similarity between application is obtained using cosine similarity formula；
User's similarity matrix with reference to described in the Similarity Measure between the application.
It is described according to the specific user operation history data of the heart and described in the application in a kind of possible design
Neighbor user intersection, which calculates prediction scoring of the specific user to intended application, to be included：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
Second aspect of the embodiment of the present invention provides a kind of recommendation apparatus of application, and the recommendation apparatus of the application includes：
Memory, processor and the recommended program for being stored in the application that can be run on the memory and on the processor, it is described
The recommended program of application is realized the application as any one of claim 1 to 5 recommendation side during the computing device
The step of method.
It is described according to the specific user operation history data of the heart and described in the application in a kind of possible design
When prediction of the neighbor user intersection calculating specific user to intended application is scored, the recommended program of the application is by the place
Manage when device is performed to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
The third aspect of the embodiment of the present invention provides a kind of computer-readable medium, and the computer-readable medium storage has
Be stored with the recommended program of application on the recommended program of application, the computer-readable recording medium, the recommendation journey of the application
When sequence is executed by processor the step of the recommendation method of application of the realization as any one of claim 1 to 5.
It is described according to the specific user operation history data of the heart and described in the application in a kind of possible design
When prediction of the neighbor user intersection calculating specific user to intended application is scored, the recommended program of the application is by the place
Manage when device is performed to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
Recommendation method, device and the computer-readable medium for the application that the present invention is provided can be solved greatly to a certain extent
The problem of data matrix amount of calculation, while improving the accuracy recommended.
Brief description of the drawings
Fig. 1 is a kind of schematic diagram of recommendation method one embodiment of application of the invention；
Fig. 2 is a kind of schematic diagram of another embodiment of the recommendation method of application of the invention；
Fig. 3 is a kind of schematic diagram of another embodiment of the recommendation method of application of the invention；
Fig. 4 is the schematic diagram of the application for predicting a certain specific user scoring row preceding ten of the embodiment of the present invention；
Fig. 5 is the variance result schematic diagram of the prediction score and actual score to a certain application of the embodiment of the present invention；
The realization, functional characteristics and advantage of the object of the invention will be described further referring to the drawings in conjunction with the embodiments.
Embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
In follow-up description, the suffix using such as " module ", " part " or " unit " for representing element is only
Be conducive to the explanation of the present invention, itself there is no a specific meaning.Therefore, " module ", " part " or " unit " can be mixed
Ground is used.
Referring to Fig. 1, Fig. 1 is a kind of recommendation method of application of the invention, the recommendation method of the application includes following step
Suddenly：
301st, start；
302nd, the operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；
Usually, a terminal-pair answers a user；If be provided with account using a center certainly or end
The multiple users of end correspondence；Terminal needs the operation history data of the upload user heart in the application；The operation history data bag of user
Include the download corresponded to, comment, scoring etc.；In the specific implementation, Apply Names is combined to the operation of its corresponding all user
Historical data generation principal matrix A；Principal matrix can also be added using label, using retouching in another embodiment of the invention
State, icon, download and temperature seniority among brothers and sisters etc. data message；
303rd, the approximate matrix A of the principal matrix A is calculatedK；
304th, the approximate matrix A is passed throughKObtain user's similarity matrix；
305th, specific user is obtained with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value
Neighbor user intersection；
306th, according to the specific user, the operation history data of the heart and the neighbor user intersection calculate institute in the application
Prediction of the specific user to intended application is stated to score；
307th, according to the prediction scoring generation application recommendation list；
308th, terminate.
Alternatively, on the basis of the corresponding embodiments of above-mentioned Fig. 1, the recommendation method of application provided in an embodiment of the present invention
In another alternative embodiment, calculating principal matrix A approximate matrix A is being performedKStep when, as shown in Fig. 2 specific bag
Include：
401st, start；
402nd, by principal matrix A described in singular value decomposition, the first matrix U, the second matrix ∑, the 3rd matrix V are obtained；
The Eigenvalues Decomposition of matrix can obtain eigen vector, characteristic vector can for Description Matrix spy
Property, but it is that for square formation, in the world of reality, most of matrix is not square formation.And singular value decomposition is
One can be suitably used for arbitrary matrix a kind of decomposition method.
Singular value decomposition can be expressed with equation below：
A=U Σ VT
Wherein, A is N*M matrix, then obtained U is that (vector of the inside is orthogonal, U for N*N square formation
The vector of the inside is referred to as left singular vector), Σ is N*M matrix (except cornerwise element is all on 0, diagonal
Element is referred to as singular value), VT(V transposition) is N*N matrix, and the vector of the inside is also orthogonal, the vector title inside V
For right singular vector)；
In specific calculate, principal matrix A carry out transposition first can be obtained into AT, it will obtain a square formation, Wo Menyong
This square formation asks characteristic value to obtain：
(ATA)vi=λ vi
Here the v obtained, is exactly the right singular vector above us.In addition we can also obtain：
Here σ is exactly above said singular value；U is exactly left singular vector；
403rd, dimension k is calculated, and obtains the first submatrix U accordinglyK, the second submatrix ∑K, the 3rd submatrix VK；
Usually, if the principal matrix A is n × m matrix, k is equal with n or m intermediate values less one；
404th, the approximate matrix AKFor the first submatrix UK, the second submatrix ∑KAnd the 3rd submatrix VK's
It is inverted the product of matrix；
For the second matrix Σ, retain wherein k maximum singular value, obtain a new dimension for k × k, k × n's
Second submatrix matrix Σk.Accordingly, by deleting the first matrix U and the 3rd corresponding row or column of matrix V, dimension is obtained
Respectively m × k the first submatrix UkWith the 3rd submatrix Vk, then matrix A is reconstructed；OrderThen
Generate matrix A closest with matrix A in all matrixes of the order equal to kk；
405th, terminate.
It can be obtained by above-mentioned analysis, singular value can represent one to the close of set matrix and the matrix lower than its order
Degree.By the decomposition of this matrix, an approximate simplification matrix of matrix A can be found.
Alternatively, on the basis of the corresponding any embodiments of above-mentioned Fig. 1 or Fig. 2, application provided in an embodiment of the present invention
Recommendation method another alternative embodiment in, passing through the approximate matrix AKWhen obtaining user's similarity matrix, including：
User's similarity matrix is obtained by below equation：
Wherein, simijRepresent user i and user j similarity.
Alternatively, on the basis of the corresponding any embodiments of above-mentioned Fig. 1 or Fig. 2, application provided in an embodiment of the present invention
Recommendation method another alternative embodiment in, passing through the approximate matrix AKWhen obtaining user's similarity matrix, such as Fig. 3 institutes
Show, including：
501st, start；
502nd, the similarity between application is obtained using cosine similarity formula；
More specifically, below equation can be taken to obtain the similarity between application：
503rd, user's similarity matrix with reference to described in the Similarity Measure between the application；
That is, when calculating user's similarity matrix, it is necessary in view of the similarity between application；
504th, terminate.
It should be noted that if principal matrix A is N*M matrix, approximate matrix AkAlso N*M square
Battle array, if as the feature of correspondence user, using cosine similarity, every a line in approximate matrix can be obtained into a M*M
User's similarity matrix.
Alternatively, on the basis of the recommendation method for the application that above-mentioned any embodiment is provided, the embodiment of the present invention is carried
In another alternative embodiment of the recommendation method of the application of confession, passing through the approximate matrix AKObtain user's similarity matrix
When, including：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
The present invention also provides a kind of recommendation apparatus of application, and the recommendation apparatus of the application includes：Memory, processor and
It is stored in the recommended program for the application that can be run on the memory and on the processor, the recommended program quilt of the application
The step of recommendation method for the application that any embodiment of the present invention is provided is realized during the computing device.
More specifically, the recommended program of the application by the computing device to realize following steps：
The operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；
Usually, a terminal-pair answers a user；If be provided with account using a center certainly or end
The multiple users of end correspondence；Terminal needs the operation history data of the upload user heart in the application；The operation history data bag of user
Include the download corresponded to, comment, scoring etc.；In the specific implementation, Apply Names is combined to the operation of its corresponding all user
Historical data generation principal matrix A；Principal matrix can also be added using label, using retouching in another embodiment of the invention
State, icon, download and temperature seniority among brothers and sisters etc. data message；
Calculate the approximate matrix A of the principal matrix AK；
Pass through the approximate matrix AKObtain user's similarity matrix；
The neighbour of specific user is obtained with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value
Occupy user's intersection；
According to the specific user, the operation history data of the heart and the neighbor user intersection calculate the spy in the application
Determine prediction of the user to intended application to score；
According to the prediction scoring generation application recommendation list.
In one embodiment of the invention, as the approximate matrix A for performing the calculating principal matrix AKWhen, the application
Recommended program is by the computing device to realize following steps：
By principal matrix A described in singular value decomposition, the first matrix U, the second matrix ∑, the 3rd matrix V are obtained；
The Eigenvalues Decomposition of matrix can obtain eigen vector, characteristic vector can for Description Matrix spy
Property, but it is that for square formation, in the world of reality, most of matrix is not square formation.And singular value decomposition is
One can be suitably used for arbitrary matrix a kind of decomposition method.
Singular value decomposition can be expressed with equation below：
A=U Σ VT
Wherein, A is N*M matrix, then obtained U is that (vector of the inside is orthogonal, U for N*N square formation
The vector of the inside is referred to as left singular vector), Σ is N*M matrix (except cornerwise element is all on 0, diagonal
Element is referred to as singular value), VT(V transposition) is N*N matrix, and the vector of the inside is also orthogonal, the vector title inside V
For right singular vector)；
In specific calculate, principal matrix A carry out transposition first can be obtained into AT, it will obtain a square formation, Wo Menyong
This square formation asks characteristic value to obtain：
(ATA)vi=λ vi
Here the v obtained, is exactly the right singular vector above us.In addition we can also obtain：
Here σ is exactly above said singular value；U is exactly left singular vector；
Dimension k is calculated, and obtains the first submatrix U accordinglyK, the second submatrix ∑K, the 3rd submatrix VK；
Usually, if the principal matrix A is n × m matrix, k is equal with n or m intermediate values less one；
The approximate matrix AKFor the first submatrix UK, the second submatrix ∑KAnd the 3rd submatrix VKInversion
The product of matrix；
For the second matrix Σ, retain wherein k maximum singular value, obtain a new dimension for k × k, k × n's
Second submatrix matrix Σk.Accordingly, by deleting the first matrix U and the 3rd corresponding row or column of matrix V, dimension is obtained
Respectively m × k the first submatrix UkWith the 3rd submatrix Vk, then matrix A is reconstructed；OrderThen
Generate matrix A closest with matrix A in all matrixes of the order equal to kk。
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
User's similarity matrix is obtained by below equation：
Wherein, simijRepresent user i and user j similarity.
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
The similarity between application is obtained using cosine similarity formula；
More specifically, below equation can be taken to obtain the similarity between application：
User's similarity matrix with reference to described in the Similarity Measure between the application；
That is, when calculating user's similarity matrix, it is necessary in view of the similarity between application；
It should be noted that if principal matrix A is N*M matrix, approximate matrix AkAlso N*M square
Battle array, if as the feature of correspondence user, using cosine similarity, every a line in approximate matrix can be obtained into a M*M
User's similarity matrix.
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
The present invention also provides a kind of computer-readable medium, it is characterised in that deposited on the computer-readable recording medium
The recommended program of application is contained, realizes what any embodiment of the present invention was provided when the recommended program of the application is executed by processor
The step of recommendation method of application.
More specifically, the recommended program of the application by the computing device to realize following steps：
The operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；
Usually, a terminal-pair answers a user；If be provided with account using a center certainly or end
The multiple users of end correspondence；Terminal needs the operation history data of the upload user heart in the application；The operation history data bag of user
Include the download corresponded to, comment, scoring etc.；In the specific implementation, Apply Names is combined to the operation of its corresponding all user
Historical data generation principal matrix A；Principal matrix can also be added using label, using retouching in another embodiment of the invention
State, icon, download and temperature seniority among brothers and sisters etc. data message；
Calculate the approximate matrix A of the principal matrix AK；
Pass through the approximate matrix AKObtain user's similarity matrix；
The neighbour of specific user is obtained with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value
Occupy user's intersection；
According to the specific user, the operation history data of the heart and the neighbor user intersection calculate the spy in the application
Determine prediction of the user to intended application to score；
According to the prediction scoring generation application recommendation list.
In one embodiment of the invention, as the approximate matrix A for performing the calculating principal matrix AKWhen, the application
Recommended program is by the computing device to realize following steps：
By principal matrix A described in singular value decomposition, the first matrix U, the second matrix ∑, the 3rd matrix V are obtained；
The Eigenvalues Decomposition of matrix can obtain eigen vector, characteristic vector can for Description Matrix spy
Property, but it is that for square formation, in the world of reality, most of matrix is not square formation.And singular value decomposition is
One can be suitably used for arbitrary matrix a kind of decomposition method.
Singular value decomposition can be expressed with equation below：
A=U Σ VT
Wherein, A is N*M matrix, then obtained U is that (vector of the inside is orthogonal, U for N*N square formation
The vector of the inside is referred to as left singular vector), Σ is N*M matrix (except cornerwise element is all on 0, diagonal
Element is referred to as singular value), VT(V transposition) is N*N matrix, and the vector of the inside is also orthogonal, the vector title inside V
For right singular vector)；
In specific calculate, principal matrix A carry out transposition first can be obtained into AT, it will obtain a square formation, Wo Menyong
This square formation asks characteristic value to obtain：
(ATA)vi=λ vi
Here the v obtained, is exactly the right singular vector above us.In addition we can also obtain：
Here σ is exactly above said singular value；U is exactly left singular vector；
Dimension k is calculated, and obtains the first submatrix U accordinglyK, the second submatrix ∑K, the 3rd submatrix VK；
Usually, if the principal matrix A is n × m matrix, k is equal with n or m intermediate values less one；
The approximate matrix AKFor the first submatrix UK, the second submatrix ∑KAnd the 3rd submatrix VKInversion
The product of matrix；
For the second matrix Σ, retain wherein k maximum singular value, obtain a new dimension for k × k, k × n's
Second submatrix matrix Σk.Accordingly, by deleting the first matrix U and the 3rd corresponding row or column of matrix V, dimension is obtained
Respectively m × k the first submatrix UkWith the 3rd submatrix Vk, then matrix A is reconstructed；OrderThen
Generate matrix A closest with matrix A in all matrixes of the order equal to kk。
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
User's similarity matrix is obtained by below equation：
Wherein, simijRepresent user i and user j similarity.
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
The similarity between application is obtained using cosine similarity formula；
More specifically, below equation can be taken to obtain the similarity between application：
User's similarity matrix with reference to described in the Similarity Measure between the application；
That is, when calculating user's similarity matrix, it is necessary in view of the similarity between application；
It should be noted that if principal matrix A is N*M matrix, approximate matrix AkAlso N*M square
Battle array, if as the feature of correspondence user, using cosine similarity, every a line in approximate matrix can be obtained into a M*M
User's similarity matrix.
In one embodiment of the invention, when passing through the approximate matrix AKIt is described when obtaining user's similarity matrix
The recommended program of application is by the computing device to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be
Auseri；By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjComment
The number of users divided：
It should be noted that the operation history data of user's heart in the application can be carried out in the form of scoring in the present invention
Performance, the operation history data of the heart includes user and stops the number of times, right opened in the time of the application, one day user in the application
Operation that the application is carried out etc..
With reference to specific example, the present invention is further illustrated：
In the specific implementation, the step of being provided according to any of the above-described embodiment builds model, imports user, user in application
The operation history data at center；The scoring of each application is drawn, and user is recommended into the preceding application of ranking.Fig. 4 is refer to, is schemed
4 be the application for predicting a certain specific user scoring row preceding ten；More specifically, first is classified as subscriber-coded, second is classified as application
Software Coding, the 3rd is classified as prediction user's score；In the present embodiment, subscriber-coded is 65722, before its corresponding prediction scoring row
Ten application as secondary series encodes corresponding application.
Fig. 5 is refer to, after the present invention is shielded the mounted one of application of user, implements what the present invention was provided
Technical scheme, the actual score drawn by the prediction score of the application of shielding and according to user's history peration data carries out variance
Calculate, the variance drawn in the present embodiment is 4.10E-6, level off to zero；It can be seen that, the prediction score that the present invention is drawn has higher
Accuracy rate.
Recommendation method, device and the computer-readable medium of application provided in an embodiment of the present invention, pass through singular value decomposition
Dimensionality reduction is carried out to user's rating matrix so that it can take less service when calculating application center recommendation super large data volume
Device resource, the problem of big data matrix computations amount can be solved to a certain extent shortens the calculating time, while improving application
Recommend precision.
It should be noted that herein, term " comprising ", "comprising" or its any other variant are intended to non-row
His property is included, so that process, method, article or device including a series of key elements not only include those key elements, and
And also including other key elements being not expressly set out, or also include for this process, method, article or device institute inherently
Key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that including this
Also there is other identical element in process, method, article or the device of key element.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Understood based on such, technical scheme is substantially done to prior art in other words
Going out the part of contribution can be embodied in the form of software product, and the computer software product is stored in a storage medium
In (such as ROM/RAM, magnetic disc, CD), including some instructions are to cause a station terminal (can be mobile phone, computer, service
Device, air conditioner, or network equipment etc.) perform method described in each of the invention embodiment.
Embodiments of the invention are described above in conjunction with accompanying drawing, but the invention is not limited in above-mentioned specific
Embodiment, above-mentioned embodiment is only schematical, rather than restricted, one of ordinary skill in the art
Under the enlightenment of the present invention, in the case of present inventive concept and scope of the claimed protection is not departed from, it can also make a lot
Form, these are belonged within the protection of the present invention.
Claims (10)
1. a kind of recommendation method of application, it is characterised in that the recommendation method of the application comprises the following steps：
The operation history data of all users heart in the application is obtained, and accordingly generates principal matrix A；
Calculate the approximate matrix A of the principal matrix AK；
Pass through the approximate matrix AKObtain user's similarity matrix；
The neighbours for obtaining specific user with reference to user's similarity matrix and with the dynamic neighbor choice method based on threshold value use
Family intersection；
According to the specific user, the operation history data of the heart and the neighbor user intersection calculate the specific use in the application
Prediction of the family to intended application is scored；
According to the prediction scoring generation application recommendation list.
2. the recommendation method applied as claimed in claim 1, it is characterised in that the approximate square of the calculating principal matrix A
Battle array AKIncluding：
By principal matrix A described in singular value decomposition, the first matrix U, the second matrix ∑, the 3rd matrix V are obtained；
Dimension k is calculated, and obtains the first submatrix U accordinglyK, the second submatrix ∑K, the 3rd submatrix VK；
The approximate matrix AKFor the first submatrix UK, the second submatrix ∑KAnd the 3rd submatrix VKInversion matrix
Product.
3. the recommendation method applied as claimed in claim 2, it is characterised in that the calculating dimension k includes：If the main square
Battle array A is n × m matrix, then k is equal with n or m intermediate values less one.
4. the recommendation method applied as claimed in claim 1, it is characterised in that described to pass through the approximate matrix AKObtain and use
Family similarity matrix includes：
User's similarity matrix is obtained by below equation：
<mrow>
<msub>
<mi>sim</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>A</mi>
<mi>k</mi>
<mi>i</mi>
</msubsup>
<mo>&CenterDot;</mo>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>A</mi>
<mi>k</mi>
<mi>j</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
</mrow>
<mrow>
<mo>|</mo>
<mo>|</mo>
<msubsup>
<mi>A</mi>
<mi>k</mi>
<mi>i</mi>
</msubsup>
<mo>|</mo>
<mo>|</mo>
<mo>&times;</mo>
<mo>|</mo>
<mo>|</mo>
<msubsup>
<mi>A</mi>
<mi>k</mi>
<mi>j</mi>
</msubsup>
<mo>|</mo>
<mo>|</mo>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
Wherein, simijRepresent user i and user j similarity.
5. the recommendation method applied as claimed in claim 1, it is characterised in that described to pass through the approximate matrix AKObtain and use
Family similarity matrix includes：
The similarity between application is obtained using cosine similarity formula；
User's similarity matrix with reference to described in the Similarity Measure between the application.
6. the recommendation method applied as claimed in claim 1, it is characterised in that it is described according to the specific user in the application
The operation history data of the heart and the neighbor user intersection, which calculate prediction scoring of the specific user to intended application, to be included：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be Auseri；
By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
<mrow>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mo>&Element;</mo>
<msub>
<mi>DATA</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
<mo>;</mo>
</mrow>
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjThere is scoring
Number of users：
<mrow>
<msub>
<mi>RE</mi>
<mrow>
<mi>u</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>+</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>.</mo>
</mrow>
1
7. a kind of recommendation apparatus of application, it is characterised in that the recommendation apparatus of the application includes：Memory, processor and deposit
The recommended program for the application that can be run on the memory and on the processor is stored up, the recommended program of the application is by institute
When stating computing device the step of the recommendation method of application of the realization as any one of claim 1 to 5.
8. the recommendation apparatus applied as claimed in claim 7, it is characterised in that it is described according to the specific user in the application
It is described when the prediction of the operation history data of the heart and the neighbor user intersection calculating specific user to intended application is scored
The recommended program of application by during the computing device to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be Auseri；
By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
<mrow>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mo>&Element;</mo>
<msub>
<mi>DATA</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
<mo>;</mo>
</mrow>
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjThere is scoring
Number of users：
<mrow>
<msub>
<mi>RE</mi>
<mrow>
<mi>u</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>+</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>.</mo>
</mrow>
9. a kind of computer-readable medium, it is characterised in that the recommendation for the application that is stored with the computer-readable recording medium
Program, realizes pushing away for the application as any one of claim 1 to 5 when the recommended program of the application is executed by processor
The step of recommending method.
10. computer-readable medium as claimed in claim 9, it is characterised in that described to be applied according to the specific user
When the prediction of the operation history data at center and the neighbor user intersection calculating specific user to intended application is scored, institute
State the recommended program of application by during the computing device to realize following steps：
The set of applications of scoring for defining the specific user is Ru, to applying Item in the principal matrix AiScoring be Auseri；
By below equation set of computations RuIn each apply ItemjWith intended application ItemiEqual difference Disij；
<mrow>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mo>&Element;</mo>
<msub>
<mi>DATA</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
<mo>;</mo>
</mrow>
Calculate prediction scoring REui, wherein NumijRepresent that the neighbor user closes pooled applications ItemiAnd ItemjThere is scoring
Number of users：
<mrow>
<msub>
<mi>RE</mi>
<mrow>
<mi>u</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Dis</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>+</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>u</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>R</mi>
<mi>u</mi>
</msub>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>Num</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>.</mo>
</mrow>
2
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710324504.2A CN107220303A (en) | 2017-05-10 | 2017-05-10 | Recommendation method, device and the computer-readable medium of a kind of application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710324504.2A CN107220303A (en) | 2017-05-10 | 2017-05-10 | Recommendation method, device and the computer-readable medium of a kind of application |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107220303A true CN107220303A (en) | 2017-09-29 |
Family
ID=59944062
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710324504.2A Pending CN107220303A (en) | 2017-05-10 | 2017-05-10 | Recommendation method, device and the computer-readable medium of a kind of application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107220303A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108355349A (en) * | 2018-03-14 | 2018-08-03 | 张伟东 | Games system |
| CN109087138A (en) * | 2018-07-26 | 2018-12-25 | 北京京东金融科技控股有限公司 | Data processing method and system, computer system and readable storage medium storing program for executing |
| CN109214882A (en) * | 2018-07-09 | 2019-01-15 | 西北大学 | A kind of Method of Commodity Recommendation |
| CN109635191A (en) * | 2018-12-05 | 2019-04-16 | 广州市百果园信息技术有限公司 | Similarity determines method, apparatus, storage medium and computer equipment |
| CN110163460A (en) * | 2018-03-30 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of method and apparatus determined using score value |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101540874A (en) * | 2009-04-23 | 2009-09-23 | 中山大学 | Interactive TV program recommendation method based on collaborative filtration |
| US20090299996A1 (en) * | 2008-06-03 | 2009-12-03 | Nec Laboratories America, Inc. | Recommender system with fast matrix factorization using infinite dimensions |
| CN103093376A (en) * | 2013-01-16 | 2013-05-08 | 北京邮电大学 | Clustering collaborative filtering recommendation system based on singular value decomposition algorithm |
| CN103678431A (en) * | 2013-03-26 | 2014-03-26 | 南京邮电大学 | Recommendation method based on standard labels and item grades |
| CN104657373A (en) * | 2013-11-20 | 2015-05-27 | 腾讯科技（上海）有限公司 | Application information pushing method and device |
| CN105025091A (en) * | 2015-06-26 | 2015-11-04 | 南京邮电大学 | Shop recommendation method based on position of mobile user |
| CN105282616A (en) * | 2015-11-13 | 2016-01-27 | 云南大学 | Combined collaborative filtering IPTV program recommendation method |
| CN105302880A (en) * | 2015-10-14 | 2016-02-03 | 合一网络技术(北京)有限公司 | Content correlation recommendation method and apparatus |
| CN105354260A (en) * | 2015-10-22 | 2016-02-24 | 中南大学 | Mobile application recommendation method with social network and project feature fused |
| CN105426514A (en) * | 2015-11-30 | 2016-03-23 | 扬州大学 | Personalized mobile APP recommendation method |
| CN105809275A (en) * | 2016-02-29 | 2016-07-27 | 海信集团有限公司 | Item scoring prediction method and apparatus |
-
2017
- 2017-05-10 CN CN201710324504.2A patent/CN107220303A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090299996A1 (en) * | 2008-06-03 | 2009-12-03 | Nec Laboratories America, Inc. | Recommender system with fast matrix factorization using infinite dimensions |
| CN101540874A (en) * | 2009-04-23 | 2009-09-23 | 中山大学 | Interactive TV program recommendation method based on collaborative filtration |
| CN103093376A (en) * | 2013-01-16 | 2013-05-08 | 北京邮电大学 | Clustering collaborative filtering recommendation system based on singular value decomposition algorithm |
| CN103678431A (en) * | 2013-03-26 | 2014-03-26 | 南京邮电大学 | Recommendation method based on standard labels and item grades |
| CN104657373A (en) * | 2013-11-20 | 2015-05-27 | 腾讯科技（上海）有限公司 | Application information pushing method and device |
| CN105025091A (en) * | 2015-06-26 | 2015-11-04 | 南京邮电大学 | Shop recommendation method based on position of mobile user |
| CN105302880A (en) * | 2015-10-14 | 2016-02-03 | 合一网络技术(北京)有限公司 | Content correlation recommendation method and apparatus |
| CN105354260A (en) * | 2015-10-22 | 2016-02-24 | 中南大学 | Mobile application recommendation method with social network and project feature fused |
| CN105282616A (en) * | 2015-11-13 | 2016-01-27 | 云南大学 | Combined collaborative filtering IPTV program recommendation method |
| CN105426514A (en) * | 2015-11-30 | 2016-03-23 | 扬州大学 | Personalized mobile APP recommendation method |
| CN105809275A (en) * | 2016-02-29 | 2016-07-27 | 海信集团有限公司 | Item scoring prediction method and apparatus |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108355349A (en) * | 2018-03-14 | 2018-08-03 | 张伟东 | Games system |
| CN110163460A (en) * | 2018-03-30 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of method and apparatus determined using score value |
| CN110163460B (en) * | 2018-03-30 | 2023-09-19 | 腾讯科技（深圳）有限公司 | Method and equipment for determining application score |
| CN109214882A (en) * | 2018-07-09 | 2019-01-15 | 西北大学 | A kind of Method of Commodity Recommendation |
| CN109214882B (en) * | 2018-07-09 | 2021-06-25 | 西北大学 | Commodity recommendation method |
| CN109087138A (en) * | 2018-07-26 | 2018-12-25 | 北京京东金融科技控股有限公司 | Data processing method and system, computer system and readable storage medium storing program for executing |
| CN109635191A (en) * | 2018-12-05 | 2019-04-16 | 广州市百果园信息技术有限公司 | Similarity determines method, apparatus, storage medium and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sur et al. | A modern maximum-likelihood theory for high-dimensional logistic regression | |
| CN107220303A (en) | Recommendation method, device and the computer-readable medium of a kind of application | |
| Todorov | Efficient computation of optimal actions | |
| CN111797321A (en) | Personalized knowledge recommendation method and system for different scenes | |
| CN111695965B (en) | Product screening method, system and equipment based on graphic neural network | |
| Hassan et al. | Genetic algorithm approaches for improving prediction accuracy of multi-criteria recommender systems | |
| CN112766649B (en) | Target object evaluation method based on multi-scoring card fusion and related equipment thereof | |
| CN112115377A (en) | Graph neural network link prediction recommendation method based on social relationship | |
| CN106774975A (en) | Input method and device | |
| CN105825269B (en) | A kind of feature learning method and system based on parallel automatic coding machine | |
| CN110706055A (en) | Commodity information pushing method and device, storage medium and computer equipment | |
| CN113191838A (en) | Shopping recommendation method and system based on heterogeneous graph neural network | |
| CN114461929A (en) | Recommendation method based on collaborative relationship graph and related device | |
| Khan et al. | Generalized fractional strategy for recommender systems with chaotic ratings behavior | |
| CN113609345B (en) | Target object association method and device, computing equipment and storage medium | |
| CN111339435A (en) | Matrix decomposition completion hybrid recommendation method based on potential factors | |
| Selsaas et al. | AFFM: Auto feature engineering in field-aware factorization machines for predictive analytics | |
| CN112800207B (en) | Commodity information recommendation method and device and storage medium | |
| CN104035978A (en) | Association discovering method and system | |
| CN108665182B (en) | Patent litigation risk prediction method | |
| CN113988951A (en) | Commodity recommendation learning model construction method based on tensor decomposition and collaborative filtering | |
| Huang et al. | Improved collaborative filtering personalized recommendation algorithm based on k-means clustering and weighted similarity on the reduced item space | |
| JP5182509B2 (en) | Information recommendation device, information recommendation method, and information recommendation program | |
| CN114842247B (en) | Characteristic accumulation-based graph convolution network semi-supervised node classification method | |
| Cai et al. | Personalised recommendation algorithm based on covariance |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170929 |
|
| RJ01 | Rejection of invention patent application after publication |