CN109389168A - Project recommendation model training method, item recommendation method and device - Google Patents
Project recommendation model training method, item recommendation method and device Download PDFInfo
- Publication number
- CN109389168A CN109389168A CN201811152399.XA CN201811152399A CN109389168A CN 109389168 A CN109389168 A CN 109389168A CN 201811152399 A CN201811152399 A CN 201811152399A CN 109389168 A CN109389168 A CN 109389168A
- Authority
- CN
- China
- Prior art keywords
- project
- user
- vector
- users
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
Abstract
The invention discloses a kind of project recommendation model training method item recommendation method and devices, project recommendation model training method of the present invention includes: the operation information based at least one user at least one project, and generating the first of each user indicates that the second of vector and each project indicates vector；At least one sample of users is selected, at least one described sample of users is some or all of of at least one user；And for each of at least one described sample of users, vector is indicated to the operation order of the project operated by it and its second of operated project based on the sample of users, and the first of the sample of users indicates vector, generates the training data of the project recommendation model；The project recommendation model is trained using the training data.The application can better describe the interest preference of user, improve the validity of model, improve the accuracy that application model carries out project recommendation.
Description
Technical field
This application involves depth learning technology fields, in particular to project recommendation model training method, project recommendation
Method and device.
Background technique
In project recommendation problem, need to consider user to the succession of different project operations, this is to study user
Current interest preference has important role.
It is currently based on the method that context relationship is modeled, such as: hidden Markov model (Hidden Markov
Model, HMM), Recognition with Recurrent Neural Network (Recurrent Neural Networks, RNN) and long memory network (Long in short-term
Short-Term Memory, LSTM) etc., there are still omit sequence information, train difficulty big.By these model applications
When project recommendation, can also there be the problems such as can not describing user's entirety interest, current interest, recommending accuracy not high.
Summary of the invention
In view of this, the embodiment of the present application is designed to provide a kind of project recommendation model training method, device and item
Mesh recommended method and device can better describe the interest preference of user, improve the validity of model, improve application model into
The accuracy of row project recommendation.
In a first aspect, the embodiment of the present application provides a kind of project recommendation model training method, comprising:
Based at least one user to the operation information of at least one project, generate the first of each user indicate to
Second expression vector of amount and each project；
Select at least one sample of users, at least one described sample of users be at least one user part or
All；And for each of at least one described sample of users, based on the sample of users to the behaviour of the project operated by it
The first of the second expression vector and the sample of users of work sequence and its operated project indicates vector, generates the item
The training data of mesh recommended models；
The project recommendation model is trained using the training data.
Optionally, the operation information includes selection and/or score information of the user to the project.
Optionally, the first expression vector is to indicate that the latent factor of the fancy grade of the user indicates vector, institute
Stating the latent factor that the second expression vector is the project indicates vector.
Optionally, it is described based at least one user to the operation information of at least one project, generate each user
First indicate that the second of vector and each project indicates vector, comprising: according at least one described user at least
The operation information of one project generates user's rating matrix；
Based on latent factor model, user's rating matrix is decomposed, generates and likes journey with user's latent factor
Spend matrix and project latent factor representing matrix；
According to user's latent factor fancy grade matrix, determine that the first of each user indicates vector；And
According to the project latent factor representing matrix, determine that the second of each project indicates vector.
Optionally, it is described based on the sample of users to the operation order of the project operated by it and its operated project
Second indicates that the first of vector and the sample of users indicates vector, generates the training data of the project recommendation model, wraps
It includes:
The operation order of project based on sample of users operation is chosen pre- from project operated by the sample of users
If the project in quantity or preset time period, and second based on selected project indicates the of vector and the sample of users
One indicates vector, generates the training input data of the project recommendation model；And
Vector is indicated by second of the project after preset quantity or preset time period, as the project recommendation model
Training monitoring data.
Optionally, it is described based on selected project second expression vector and the sample of users first indicate to
Amount, generates the training input data of the project recommendation model, comprising: according to the sample of users to the operation order of project, group
Close the project in the selected preset quantity or preset time period second indicates vector, forms project time sequence matrix；
Project time sequence matrix corresponding to the sample of users and its first are indicated into vector, as the project recommendation model
Training input data.
In a kind of possible embodiment, the project recommendation model is convolutional neural networks CNN model.
Second aspect, the embodiment of the present application provide a kind of item recommendation method, comprising:
Based on target user to the operation information of at least one project, selection is in the preset time at current time or pre-
If the project of quantity；
Based on the target user to the operation order of selected project, using the second of selected project indicate to
The first expression vector of amount and the target user, generate input data；
In the input data cuit recommended models, will obtain the prediction of prediction term purpose indicates vector, the project
Recommended models project recommendation model training method as described in first aspect any one obtains；
Calculate the similarity between the prediction expression vector and the second expression vector of project；Based on the phase being calculated
Like degree, determines the project recommended to the target user and recommend the target user.
The third aspect, the embodiment of the present application provide a kind of project recommendation model training apparatus, comprising:
It indicates vector generation module, for the operation information based at least one user at least one project, generates each
The first of a user indicates that the second of vector and each project indicates vector；
Training data generation module, for selecting at least one sample of users, at least one described sample of users is described
At least one user's is some or all of；And it for each of at least one described sample of users, is used based on the sample
Family indicates the of vector and the sample of users to the operation order of the project operated by it and its second of operated project
One indicates vector, generates the training data of the project recommendation model；
Training module, for being trained using the training data to the project recommendation model.
Fourth aspect, the embodiment of the present application provide a kind of project recommendation device, comprising:
Selecting module is selected for the operation information based on target user at least one project apart from current time
In preset time or the project of preset quantity；
Input data generation module uses institute for the operation order based on the target user to selected project
The second of the project of selection indicates that the first of vector and the target user indicates vector, generates input data；
Vector generation module is indicated, for obtaining prediction term purpose in the input data cuit recommended models
Prediction indicates that vector, project recommendation model project recommendation model training method as described in first aspect any one obtain
It arrives；
Similarity calculation module indicates that vector indicates similar between vector to the second of project for calculating the prediction
Degree；
Recommending module, for determining the project recommended to the target user and recommendation based on the similarity being calculated
To the target user.
The embodiment of the present application provides a kind of project recommendation model training method, item recommendation method and device, Neng Gouji
In at least one user to the operation information of at least one project, generate each user first indicates vector and each project
Second indicate vector, then for operation order and each sample of users institute of the sample of users to project determined from user
The second of operation item indicates that the first of vector and sample of users indicates vector, generates the training data of project recommended models,
And project recommendation model is trained, the interest preference of user can be better described, the validity of model is improved.
To enable the above objects, features, and advantages of the application to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of flow chart of project recommendation model training method provided by the embodiment of the present application；
Fig. 2 shows the first tables in project recommendation model training method provided by the embodiment of the present application, obtaining user
Showing the second of each project of vector sum indicates the flow chart of the specific method of vector；
Fig. 3 is shown in project recommendation model training method provided by the embodiment of the present application, generates project recommended models
Training data specific method flow chart；
Fig. 4 shows the process schematic for choosing project provided by the embodiment of the present application by sliding window.
Fig. 5 shows a kind of flow chart of item recommendation method provided by the embodiment of the present application；
Fig. 6 shows a kind of schematic diagram of project recommendation model training apparatus 600 provided by the embodiment of the present application；
Fig. 7 shows a kind of schematic diagram of project recommendation device 700 provided by the embodiment of the present application；
Fig. 8 shows the structural schematic diagram of a kind of electronic equipment 800 provided by the embodiment of the present application；
Fig. 9 shows the structural schematic diagram of a kind of electronic equipment 900 provided by the embodiment of the present application；
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
Middle attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only
It is some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is real
The component for applying example can be arranged and be designed with a variety of different configurations.Therefore, below to the application's provided in the accompanying drawings
The detailed description of embodiment is not intended to limit claimed scope of the present application, but is merely representative of the selected reality of the application
Apply example.Based on embodiments herein, those skilled in the art institute obtained without making creative work
There are other embodiments, shall fall in the protection scope of this application.
Unlike the prior art, the embodiment of the present application modeling when, according to user to different project operations before
Order afterwards, obtain user it is long-term, it is static, be not easy the hobby feature changed easily and user is recent, dynamic hobby
Feature is modeled according to both hobby features, and the project recommendation model enabled better describes the emerging of user
Interesting preference improves the validity of model, improves the accuracy that application model carries out project recommendation.
For example, carrying out film recommendation to user, found from the scoring record that user watches film, user watches and gives
Give that action movie in the film of favorable comment is in the majority, however user is watched and is given recently in 10 films of favorable comment, romance movie accounts for 7
Portion, in remaining 3 film, only 2 parts are action movies.It can be concluded that action movie be user it is long-term, it is static, be not easy
The hobby changed easily, and romance movie can be described as that user is recent, dynamic hobby.Under current time node, recommend
This user's love film is to be more in line with the current film interest preference of user than recommending action movie better choice.
In order to achieve the above object, the embodiment of the present application provides a kind of project recommendation model training method, project recommendation side
Method and device can generate the first table of each user based at least one user to the operation information of at least one project
Showing the second of vector and each project indicates vector, then suitable to the operation of project for the sample of users determined from user
Second of project operated by sequence and each sample of users indicates that the first of vector and sample of users indicates vector, generates project
The training data of recommended models, and project recommendation model is trained, the interest preference of user is better described, model is improved
Validity, improve application model carry out project recommendation accuracy.
To be instructed to a kind of project recommendation model disclosed in the embodiment of the present application first convenient for understanding the present embodiment
Practice method to describe in detail.
It is shown in Figure 1, project recommendation model training method provided by the embodiments of the present application, comprising: S101~S103.
S101: based at least one user to the operation information of at least one project, the first of each user is generated
Indicating the second of vector and each project indicates vector.
When specific implementation, project can refer to for the main body of user's selection on particular platform, such as in a purchase
On object platform, every sample commodity are exactly a project；In a video website, each video (such as film) or some be
The video (such as one includes the TV play for collecting collection of dramas more) of column is exactly a project；Either image, etc..Different projects
It can come from a platform, can be from multiple platforms.It is not intended to limit the type of project herein.In other words, project can also be with
It is the main body for recommending user, recommendation provided by the embodiment of the present application determines that model can be determined from each project and wants
Recommend the destination item of user.
Here, at least one user to the operation information of at least one project may include user to the selection of project and/
Or score information.By taking project is the commodity on shopping platform as an example, operation of the user to project can be user and look into commodity
The operation such as see, shopping cart, purchase, comment be added；Wherein, it checks, shopping cart is added, purchase all can serve as user to project
Selection；Comment can be used as scoring of the user to project.By taking project is the video in video website as an example, behaviour of the user to project
User be can be to operations such as the click viewings, addition concern, comment of project；Wherein, viewing is clicked, addition concern is all to use
Selection of the family to project；Comment on the scoring as user items.
Operation information can be and believe to all users at current time the operation of each project after platform foundation
Breath.Optionally, obtain adapting to the project recommendation model of present case in order to training, acquired user is at least one
Purpose operation information can be the operation information obtained in current time preset time.
In addition, may have the case where project undercarriage for each platform.In response to this, due to user
Operation behavior can be made for the project of undercarriage, which is also that can characterize the hobby of user to a certain extent
It is biased to, therefore the project of undercarriage can also be used as the project for needing to operate in the application.
After obtaining at least one user to the operation information of at least one project, user is generated based on the operation information
First indicate vector sum second indicate vector.Wherein, the hobby that the first expression vector is used to characterize each user is biased to, therefore
First indicates that vector can indicate vector for latent factor of the expression user to the fancy grade of each project；Second indicates vector
It is then for characterizing each project, the latent factor that can be project indicates vector.
Specifically, shown in Figure 2, the embodiment of the present application using following S201~S204 obtain user first indicate to
Second expression vector of amount and each project:
S201: according at least one described user to the operation information of at least one project, user's rating matrix is generated.This
Locate, each element in user's rating matrix, is the scoring for the operation that some project occurs for a certain user.The scoring can be with
It is obtained using any one following generating mode:
(1): each user is directed to, as long as operation behavior has occurred to some project in the user, regardless of this is to same project
Operation behavior have occurred several times, all the user can be set default score value to the scoring of the project.
Such as: there are user's first, second and third, is respectively as follows: user's first for project A, B, C, D operation occurred and project A is sent out
The operation clicked and checked is given birth to, to the operation that project B is evaluated, to project C and project D without operation；Then user's first pair
The scoring of project A and project B set 1, due to its to project C and project D without operation, user's first comments project C and project D
Split 0.The operation that user's second buys project B, it is equal to project A and project C to the operation that project D is checked
Without operation, then user's second sets 1 to the scoring of project B and project D, due to rising to project A and project C without operation, user
Second sets 0 to the scoring of project A and project C.Purchase operation has occurred to project A in user third, has occurred to project B and checks operation,
Purchase operation is had occurred to project C, to project D without operation, then user third sets 1 to the scoring of project A, project B and project C, right
The scoring of project D sets 0.The user's rating matrix then formed is as shown in table 1 below:
Table 1
| Project A | Project B | Project C | Project D | |
| User's first | 1 | 1 | 0 | 0 |
| User's second | 0 | 1 | 0 | 1 |
| User third | 1 | 1 | 1 | 0 |
(2) be directed to each user, some each pair of project of the user have occurred once-through operation behavior it is necessary to by the user to this
The scoring of project increases setting score value, such as: adding 1.
Such as: there are user's first, second and third, is respectively as follows: user's first for project A, B, C, D operation occurred and project A is sent out
It has given birth to 2 times and has clicked the operation checked, project B has had occurred the operation of 1 evaluation, project C has had occurred the operation of 3 purchases,
To project D without operation；Then user's first is respectively as follows: 2,1,3,0 to the scoring of project A, project B, project C and project D, since its is right
Project D is without operation, therefore user's first sets 0 to the scoring of project D.User's second project A has occurred the operation of 1 evaluation, to item
The operation of 1 purchase has occurred in mesh B, 2 operations checked has occurred to project D, to project C without operation, then user's second is to item
The scoring of mesh A, project B, project C and project D are respectively as follows: 1,1,0,2.User third, without operation, has occurred 3 to project B to project A
It is secondary to check operation, 1 purchase operation is had occurred to project C, has no way of operating to project D, then user third is to project A, project B, item
The scoring of mesh C and project D are respectively as follows: 0,3,1,0.The user's rating matrix then formed is as shown in table 2 below:
Table 2
| Project A | Project B | Project C | Project D | |
| User's first | 2 | 1 | 3 | 0 |
| User's second | 1 | 1 | 0 | 2 |
| User third | 0 | 3 | 1 | 0 |
(3) user occurs project as user for the evaluation data such as scoring of each project or grade
The scoring of operation.
In addition, scoring acquisition pattern provided by the embodiments of the present application is not limited to above-mentioned three kinds, it can also be by above-mentioned at least two
Item combines.For example, if user operates for some project, and being carried out to the project when combining (2) and (3)
The scoring of scoring, the then operation that the project occurs for the user uses scoring of the user to project；If user is directed to some
Project is operated, but and score the project, then it is scored based on user the operation of the project.
Further, it is also possible to there is other scoring acquisition patterns, the different operation behavior of for example, user assigns different comment
Score value, for user when each operation behavior occurs to some project, the scoring assigned for the project is corresponding according to operation behavior
Score value carries out assignment.
For example, user checks behavior to project, corresponding score value is 1；Buying behavior, corresponding score value occurs
It is 4；Splitting glass opaque occurs, corresponding score value is 2；Occur that blacklist behavior is added, corresponding score value is -3；It evaluates
Behavior, corresponding score value are 2, finally carry out the corresponding score value of each secondary operation behavior that user occurs for some project
It is cumulative, obtain the scoring that user project occurs operation behavior.
S202: being based on latent factor model, decompose to user's rating matrix, generates and likes with user's latent factor
Good degree matrix and project latent factor representing matrix.
When specific implementation, user's rating matrix is expressed as Rn×m；Wherein, n indicates the quantity of user；M indicates project
Quantity, the scoring factor matrix established can be as shown in table 3 below:
Table 3
| Project 1 | Project 2 | …… | Project m | |
| User 1 | R11 | R12 | …… | R1m |
| …… | …… | …… | …… | …… |
| User n | Rn1 | Rn2 | …… | Rnm |
User's latent factor fancy grade matrix of generation is expressed as: Pn×f；Wherein, f indicates the quantity of latent factor.It is raw
At project latent factor representing matrix indicate are as follows: Qf×m.User's rating matrix is decomposed, the happiness of user's latent factor is generated
The process of good degree matrix and project latent factor representing matrix can be expressed as follows shown in table 4:
Table 4
S203: according to user's latent factor fancy grade matrix, determine that the first of each user indicates vector；
And according to the project latent factor representing matrix, determine that the second of each project indicates vector.
After user's latent factor fancy grade matrix has been determined, for each user in user's latent factor fancy grade
Element corresponding with each latent factor in matrix determines that the first of the user indicates vector, such as dives shown in above-mentioned table 4
In factor fancy grade matrix, the first of user 1 indicates vector are as follows: (P11, P12 ..., P1f)；The first of user 2 indicates
Vector are as follows: (P21, P22 ..., P2f)；... the first of user n indicates vector are as follows: (Pn1, Pn2 ... Pnf).
Similarly, the second expression vector of each project can also obtain in the same fashion, such as shown in the above-mentioned table 4
Project latent factor representing matrix in, the second of project 1 indicates vector are as follows: (Q11, Q21 ..., Qf1)；The second of project 2
Indicate vector are as follows: (Q12, Q22 ..., Qf2)；... the second of project m indicates vector are as follows: (Q1m, Q2m ..., Qfm).
It is provided by the embodiments of the present application after generate user first indicates the second expression vector of each project of vector sum
Project recommendation model training method will also indicate that vector sum second indicates vector composing training data based on first, specific as follows
It states shown in S102.
S102: selecting at least one sample of users, at least one described sample of users is the portion of at least one user
Divide or whole；And for each of at least one described sample of users, based on the sample of users to the item operated by it
The first of the second expression vector and the sample of users of purpose operation order and its operated project indicates vector, generates
The training data of the project recommendation model.
When specific implementation again, sample of users be screened from user it is all or part of.
The case where for sample of users being the part screened in user, due to indicating vector generate user first
During indicating vector with the second of project, the quantity of user, the needs of training can be combined, appropriate number of user is selected
As sample of users, training data is generated.
Furthermore, it is also possible to that there are the operation informations of certain user is more long apart from current time, certain user is currently
In disabled state, or the state being lost, based on the operation information of this certain customers come composing training data, training
When can be not suitable for the information of these users, for this purpose, can be screened from user currently also in the user of more active state
As sample of users.Wherein, it is currently at more active user, can be has project in current time preset duration
The user of operation behavior.
In addition, before the training data for generating some sample of users, it can also operation to the sample of users to project
Data carry out data cleansing, such as the number of entry that operates within a certain period of time of the user is considerably less or the time is more long
Far, training is helped less, such as once-through operation behavior only has occurred between in July, 2018 in September, 2017, then just
This operation behavior can be rejected from the operation behavior of the sample of users.
Specifically, shown in Figure 3, the embodiment of the present application generates the training number of project recommended models using following manner
According to:
S301: the operation order of the project based on sample of users operation, from project operated by the sample of users, choosing
The project in preset quantity or preset time period is taken, and second based on selected project indicates vector and the sample of users
First indicate vector, generate the training input data of the project recommendation model.
S302: vector is indicated by second of the project after preset quantity or preset time period, as the project recommendation
The training monitoring data of model.Here, second of the project after preset quantity or preset time period indicates vector, corresponding user
The project of institute's practical operation indicates that vector as training monitoring data, realizes the supervised training to model for the second of the project.
For example, model carries out operation to input data, operation result namely prediction result are compared with training monitoring data, root
According to the parameter of comparison result adjustment model.
Herein, it should be noted that above-mentioned S301 and S302 has no the sequencing of execution, may be performed simultaneously, can also
To first carry out any one therein.When specific implementation, when being trained to model, there may be training
The case where data deficiencies, the model obtained by this data can have that accuracy rate is low.In order to avoid above situation
It generates, the embodiment of the present application can be from the project operated by sample of users, multiple selected part sample data, defeated with building training
Enter data, the sample data chosen every time can be entirely different, there can also be part identical.When selection, used based on sample
The operation order of the project of family operation carries out.The project of preset quantity can be chosen, when selection, when can also choose default
Between project in section.
(1) for the project for choosing preset quantity the case where, such as the project of user A operation includes W1~W1000 total
1000 datas choose 100 projects according to preset quantity 100 from W1~W1000 every time, and it is corresponding to construct this time selection
Training input data,
The project chosen for the first time are as follows: W1~W100, trained monitoring data corresponding at this time are W101；
Second of project chosen are as follows: W11~W110, trained monitoring data corresponding at this time are W111；
The project that third time is chosen are as follows: W21~W120, trained monitoring data corresponding at this time are W121；
The project of 4th selection are as follows: W201~W300, trained monitoring data corresponding at this time are W301.
The project of 5th selection are as follows: W501~W600, trained monitoring data corresponding at this time are W601.……
It is chosen for each, choosing the second of selected project according to this time indicates the of vector and the sample of users
One indicates vector, generates the training input data of the project recommendation model.It herein, can be with when carrying out project selection
It is carried out by sliding window；That is, the operation order of the project based on sample of users operation, to sample of users operation
All items are ranked up, and then use a sliding window, which " can enclose " project for living preset quantity, will slide
Dynamic window is since any one position of the project after sequence, after drawing a circle to approve the project of preset quantity, using the project of delineation as
The project of selection is instructed the project outside the adjacent sliding window of the last one project in the project with delineation as corresponding supervision
Practice data；After sliding window is moved back according to certain step-length, repeat, until having slided selected project.
The process schematic of project, the project packet of user A operation are chosen provided by shown in Figure 4 by sliding window
W1~W1000 totally 1000 data is included, the size of sliding window is 20, step-length 5, and sliding window is drawn a circle to approve since W1 for the first time
Project, as shown in a in Fig. 4, including W1~W20, W21 are as corresponding trained monitoring data；Project such as Fig. 4 of second of delineation
Shown in middle b, comprising: W6~W26, W27 are as corresponding trained monitoring data ... in this way, sliding window is moved, to obtain
Multiple groups training input data corresponding with the sample of users and training monitoring data.Size, the step-length of the above sliding window be all
Example, herein to it and with no restrictions.
(2) for the project in preset time period is chosen the case where, chooses project from project operated by sample of users
When, different sample of users can have the identical period.For example, when selecting one section before current point in time
Between, project operated by sample of users.The division of operations that family can also be mixed the sample with is multiple periods, and is based on after dividing
Period, therefrom choose sample data project operated by some or certain periods.
The preset time period can be the identical different time sections of time span, example for different sample of users
Make as chosen project operated by user's first user's first from the period between on January 10,1 day to 2018 January in 2018
For the project of selection；Choose user's second user's second institute from the period between on March 10,1 day to 2018 March in 2018
The project of operation is as the project chosen；It is also possible to the completely the same period, such as user's first and user's second, all selects
Take two users from 2 months 2018 periods between No. 1 to 2 months 2018 No. 10, the item of user's first and the operation of user's second
Mesh is as the project respectively chosen.In response to this, can be different to the number of entry selected by different user, some users
Frequent operation, then the project of the user chosen is with regard to more；The operation of some users is less, then the project of the user chosen
With regard to less.In the case, the project of selection can also be further processed, such as: the user institute less to project is right
The data answered carry out zero padding operation, or can also be more to project user corresponding to training data carry out rejecting operation,
Deng.The above time segment length is only example, herein to it and with no restrictions.Furthermore, it is possible to according to the item types of recommendation, if
Set the period of appropriate length.
In addition, the embodiment of the present application also provides a kind of the second expression vector based on selected project and the sample is used
The first of family indicates vector, generates the specific method of the training input data of the project recommendation model, comprising: according to the sample
User to the operation order of project, combine second of the project in the selected preset quantity or preset time period indicate to
Amount forms project time sequence matrix, i.e., arranges project expression vector row wise or column wise, be combined into matrix；By the sample of users institute
Corresponding project time sequence matrix and its first expression vector, the training input data as the project recommendation model.
S103: the project recommendation model is trained using the training data.
When specific implementation, project recommendation model can be convolutional neural networks (ConvolutionalNeural
Network, CNN) model.Exquisite ability of the CNN by extracting characteristics of image, in computer vision field, such as target detection, vision
Understand etc., the success for having obtained huge.Convolution process in convolutional neural networks can be understood as a kind of abstract process, will be small
Information Statistics in region abstract；Pond process can be understood as the screening in the description of extensive feature and extract most one
As, most representational information.Multiple convolution sum pond process, facilitates characteristics of image by shallow and deeply extract,
Then it is connect again with full articulamentum, it is final to obtain profound character representation.With convolutional neural networks CNN, pass through convolution, pond
The operation such as change, excavate user's latent factor characteristic sequence " figure ", to obtain the instant interest preference transfer mode of user, then by its with
Prolonged, the static interest preference feature of user merges, and description user currently recommend quasi- by whole interest preference, raising
True property.
When being trained using training data to convolutional neural networks, due to the training input data of training data
It is the operation order according to sample of users to project, by the second table of the project in the preset quantity or preset time period of selection
Show that vector is formed by the time sequence matrix of project, the second expression vector of each project treated as a whole,
A pixel being similar in image.
CNN includes convolutional layer, pond layer and full articulamentum.It can be trained using following manner:
(1) the project time sequence matrix in training input data is input to convolutional layer, uses multiple convolution of different sizes
Core carries out convolution algorithm to training input data, obtains median feature vector corresponding with each convolution kernel.
(2) after being extracted multiple median feature vectors of time sequence matrix, multiple median feature vectors enter pond layer.
In the layer of pond, multiple median feature vectors can be subjected to Fusion Features, obtained corresponding with the time sequence matrix
Indicate that the vector of user interest preference transfer mode indicates, referred to as interest preference transfer indicates vector, the interest preference transfer table
Show vector for characterizing the interest preference of sample of users within a short period of time；Then by the interest preference shift representing matrix and when
First expression vector of the corresponding sample of users of sequence matrix is spliced, and forming interest preference transfer indicates that vector sum user is potential
The splicing vector of factor fancy grade vector.Wherein, user's latent factor fancy grade vector can characterize user for a long time, no
The interest preference of malleable.
(3) after obtaining splicing vector, splicing vector is sent into full articulamentum.Full articulamentum can will splice in vector
Interest preference transfer indicate vector sum user latent factor fancy grade vector carry out Fusion Features, ultimately forming can either be pre-
Measure the feature vector of the interested next project of user.
The feature vector of the interested next project of user is obtained in prediction, it will be able to the feature vector based on the prediction
With the training monitoring data of the sample of users, the parameter of CNN is adjusted.
When the parameter to volume CNN is adjusted, seek to so that the user predicted by CNN is interested
The second of next project that the feature vector of next project and user are actually operated indicates vector, Ye Jixun
Practice monitoring data as far as possible close.
After more wheels training to convolutional neural networks model, using obtained CNN as project recommendation model.
The embodiment of the present application can generate each user based at least one user to the operation information of at least one project
First indicate that the second of vector and each project indicates vector, then for the sample of users determined from user to project
Operation order and each sample of users operated by project second indicate that the first of vector and sample of users indicates vector,
The training data of generation project recommended models, and project recommendation model is trained, the interest preference of user is preferably utilized,
Improve the validity of model.
Shown in Figure 5, the embodiment of the present application also provides a kind of item recommendation method, including S501~S504:
S501: based on target user to the operation information of at least one project, the preset time apart from current time is selected
Interior or preset quantity project.Operated project or operated recently specific i.e. in selection target user nearest a period of time
The project of quantity.
S502: based on the target user to the operation order of selected project, the second of selected project is used
Indicating the first of vector and the target user indicates vector, generates input data.
Herein, the generation method of input data is similar with the training generation method of input data, and details are not described herein.
S503: in the input data cuit recommended models, will obtain the prediction of prediction term purpose indicates vector, described
Project recommendation model is obtained by project recommendation model training method provided by the embodiments of the present application.
Herein, obtaining prediction according to input data indicates the process and the above-mentioned feature vector for obtaining sample of users of vector
Process is similar, and details are not described herein.
S504: the similarity between the prediction expression vector and the second expression vector of project is calculated；Based on calculating
The similarity arrived determines the project recommended to the target user and recommends the target user.
Herein, can successively calculate the second of each project indicates that vector indicates that the cosine between vector is similar to prediction
Degree, and similarity highest second is indicated into the corresponding project of vector, it is determined as wanting project recommended to the user, then by this
Mesh recommends target user.
When being recommended using aforementioned project recommendation model, can preferably utilize user interest preference so as to
The project that family is recommended can be more in line with the hobby of user.
Based on the same inventive concept, item corresponding with project recommendation model training method is additionally provided in the embodiment of the present application
Mesh recommended models training device, the principle solved the problems, such as due to the device in the embodiment of the present application and the above-mentioned item of the embodiment of the present application
Mesh recommended models training method is similar, therefore the implementation of device may refer to the implementation of method, and overlaps will not be repeated.
Shown in Figure 6, project recommendation model training apparatus 600 provided by the embodiments of the present application includes:
It indicates vector generation module 61, for the operation information based at least one user at least one project, generates
The first of each user indicates that the second of vector and each project indicates vector；
Training data generation module 62, for selecting at least one sample of users, at least one described sample of users is institute
State some or all of of at least one user；And for each of at least one described sample of users, it is based on the sample
User indicates vector and the sample of users to the operation order of the project operated by it and its second of operated project
First indicates vector, generates the training data of the project recommendation model；
Training module 63, for being trained using the training data to the project recommendation model.
The embodiment of the present application can generate each user based at least one user to the operation information of at least one project
First indicate that the second of vector and each project indicates vector, then for the sample of users determined from user to project
Operation order and each sample of users operated by project second indicate that the first of vector and sample of users indicates vector,
The training data of generation project recommended models, and project recommendation model is trained, training process difficulty is small, expends the time
It is short, and can establish interrelated between different time nodes, take into account generality.
In a kind of possible embodiment, the operation information includes the user to the selection of the project and/or comments
Divide information.
In a kind of possible embodiment, the first expression vector be indicate the user fancy grade it is potential because
Subrepresentation vector, described second indicates that the latent factor that vector is the project indicates vector.
In a kind of possible embodiment, indicate that vector generation module 61 is specifically used for generating each institute using following manner
Stating the first of user indicates that the second of vector and each project indicates vector: according at least one described user at least
The operation information of one project generates user's rating matrix；Based on latent factor model, user's rating matrix is divided
Solution generates and user's latent factor fancy grade matrix and project latent factor representing matrix；According to user's latent factor
Fancy grade matrix determines that the first of each user indicates vector；And square is indicated according to the project latent factor
Battle array determines that the second of each project indicates vector.
In a kind of possible embodiment, training data generation module 62 is specifically used for generating the item using following manner
The training data of mesh recommended models: the operation order of the project based on sample of users operation, operated by the sample of users
In project, the project in preset quantity or preset time period is chosen, and second based on selected project indicates vector and institute
State sample of users first indicates vector, generates the training input data of the project recommendation model；And by preset quantity or
Second of project after preset time period indicates vector, the training monitoring data as the project recommendation model.
In a kind of possible embodiment, training data generation module 62 is specifically used for being based on using following manner selected
Project second indicate that the first of vector and the sample of users indicates vector, the training for generating the project recommendation model is defeated
Enter data: the operation order according to the sample of users to project combines in the selected preset quantity or preset time period
Project second indicate vector, formed project time sequence matrix；By project time sequence matrix corresponding to the sample of users and its
One indicates vector, the training input data as the project recommendation model.
In a kind of possible embodiment, project recommendation model is convolutional neural networks CNN model.
Based on the same inventive concept, project recommendation dress corresponding with item recommendation method is additionally provided in the embodiment of the present application
It sets, since the principle that the device in the embodiment of the present application solves the problems, such as is similar to the above-mentioned item recommendation method of the embodiment of the present application,
Therefore the implementation of device may refer to the implementation of method, and overlaps will not be repeated.
Shown in Figure 7, project recommendation device 700 provided by the embodiments of the present application includes:
Selecting module 71 is selected for the operation information based on target user at least one project apart from current time
Preset time in or preset quantity project；
Input data generation module 72 is used for the operation order based on the target user to selected project
The second of selected project indicates that the first of vector and the target user indicates vector, generates input data；
Vector generation module 73 is indicated, for obtaining prediction project in the input data cuit recommended models
Prediction indicate vector, the project recommendation model obtains by project recommendation model training method provided by the embodiments of the present application；
Similarity calculation module 74, for calculating the phase between the prediction expression vector and the second expression vector of project
Like degree；
Recommending module 75, for determining the project recommended to the target user and pushing away based on the similarity being calculated
It recommends to the target user.
When being recommended using aforementioned project recommendation model, can preferably utilize user interest preference so as to
The project that family is recommended can be more in line with the hobby of user.
Corresponding to the project recommendation model training method in Fig. 1, the embodiment of the present application also provides a kind of computer equipments
800, as shown in figure 8, the equipment 800 includes memory 801, processor 802 and is stored on the memory 801 and can be at this
The computer program run on reason device 802, wherein above-mentioned processor 802 realizes above-mentioned project when executing above-mentioned computer program
The step of recommended models training method.
Specifically, above-mentioned memory 801 and processor 802 can be general memory and processor, do not do have here
Body limits, and when the computer program of 802 run memory 801 of processor storage, is able to carry out above-mentioned project recommendation model instruction
Practice method, to solve the problems, such as that generality and training difficulty cannot be considered in terms of, and then reach small to the training process difficulty of model,
It is short to expend the time, and can establish interrelated between different time nodes, takes into account recapitulative effect.
Corresponding to the project recommendation model training method in Fig. 1, the embodiment of the present application also provides a kind of computer-readable
Storage medium is stored with computer program on the computer readable storage medium, which holds when being run by processor
The step of row above-mentioned project recommendation model training method.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, above-mentioned project recommendation model training method is able to carry out, to solve generality and training is difficult
The problem of degree cannot be considered in terms of, and then reach small to the training process difficulty of model, it is short to expend the time, and different time can be established
It is interrelated between node, take into account recapitulative effect.
Corresponding to the item recommendation method in Fig. 5, the embodiment of the present application also provides a kind of computer equipments 900, such as Fig. 9
Shown, which includes memory 901, processor 902 and is stored on the memory 801 and can transport on the processor 902
Capable computer program, wherein above-mentioned processor 902 realizes the step of above-mentioned item recommendation method when executing above-mentioned computer program
Suddenly.
Specifically, above-mentioned memory 901 and processor 902 can be general memory and processor, do not do have here
Body limits, and when the computer program of 902 run memory 902 of processor storage, is able to carry out above-mentioned item recommendation method, from
And the interest preference of user can be better described, improve the validity of model.
Corresponding to the project recommendation model training method in Fig. 5, the embodiment of the present application also provides a kind of computer-readable
Storage medium is stored with computer program on the computer readable storage medium, which holds when being run by processor
The step of row above-mentioned item recommendation method.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, above-mentioned item recommendation method is able to carry out, so as to preferably utilize the interest of user inclined
It is good, enable project recommended to the user to be more in line with the hobby of user.Project recommendation mould provided by the embodiment of the present application
The computer program product of type training method, device and item recommendation method and device, the calculating including storing program code
Machine readable storage medium storing program for executing, the instruction that said program code includes can be used for executing previous methods method as described in the examples, tool
Body, which is realized, can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description
It with the specific work process of device, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.In the application
In provided several embodiments, it should be understood that disclosed systems, devices and methods, it can be real by another way
It is existing.The apparatus embodiments described above are merely exemplary, for example, the division of the unit, only a kind of logic function
It can divide, can there is other division mode in actual implementation, in another example, multiple units or components can be combined or be can integrate
To another system, or some features can be ignored or not executed.Another point, shown or discussed mutual coupling
Or direct-coupling or communication connection can be the indirect coupling or communication connection by some communication interfaces, device or unit, it can
To be electrically mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, each functional unit in each embodiment of the application can integrate in one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in the executable non-volatile computer-readable storage medium of a processor.Based on this understanding, the application
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the application
State all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (Read-Only
Memory, ROM), random access memory (Random Access Memory, RAM), magnetic or disk etc. is various to deposit
Store up the medium of program code.
Finally, it should be noted that embodiment described above, the only specific embodiment of the application, to illustrate the application
Technical solution, rather than its limitations, the protection scope of the application is not limited thereto, although with reference to the foregoing embodiments to this Shen
It please be described in detail, those skilled in the art should understand that: anyone skilled in the art
Within the technical scope of the present application, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of the embodiment of the present application technical solution, should all cover the protection in the application
Within the scope of.Therefore, the protection scope of the application shall be subject to the protection scope of the claim.
Claims (10)
1. a kind of project recommendation model training method characterized by comprising
Based at least one user to the operation information of at least one project, generate the first of each user indicate vector with
And the second of each project indicates vector；
At least one sample of users is selected, at least one described sample of users is the part or complete of at least one user
Portion；Operation and for each of at least one described sample of users, based on the sample of users to the project operated by it
The first of the second expression vector and the sample of users of sequence and its operated project indicates vector, generates the project
The training data of recommended models；
The project recommendation model is trained using the training data.
2. the method according to claim 1, wherein the operation information includes the user to the project
Selection and/or score information.
3. the method according to claim 1, wherein described first indicates that vector is the hobby for indicating the user
The latent factor of degree indicates vector, and described second indicates that the latent factor that vector is the project indicates vector.
4. according to the method described in claim 3, it is characterized in that, it is described based at least one user at least one project
Operation information, generating the first of each user indicates that the second of vector and each project indicates vector, comprising:
According at least one described user to the operation information of at least one project, user's rating matrix is generated；
Based on latent factor model, user's rating matrix is decomposed, is generated and user's latent factor fancy grade square
Battle array and project latent factor representing matrix；
According to user's latent factor fancy grade matrix, determine that the first of each user indicates vector；And
According to the project latent factor representing matrix, determine that the second of each project indicates vector.
5. the method according to claim 1, wherein it is described based on the sample of users to the project operated by it
The first of the second expression vector and the sample of users of operation order and its operated project indicates vector, generates institute
State the training data of project recommendation model, comprising:
The operation order of project based on sample of users operation chooses present count from project operated by the sample of users
Project in amount or preset time period, and the first table of the based on selected project second expression vector and the sample of users
Show vector, generates the training input data of the project recommendation model；And
Vector is indicated by second of the project after preset quantity or preset time period, the training as the project recommendation model
Monitoring data.
6. according to the method described in claim 5, it is characterized in that, it is described based on selected project second indicate vector with
The first of the sample of users indicates vector, generates the training input data of the project recommendation model, comprising:
According to the sample of users to the operation order of project, the item in the selected preset quantity or preset time period is combined
Purpose second indicates vector, forms project time sequence matrix；
Project time sequence matrix corresponding to the sample of users and its first are indicated into vector, the instruction as the project recommendation model
Practice input data.
7. -6 any method according to claim 1, which is characterized in that the project recommendation model is convolutional neural networks
CNN model.
8. a kind of item recommendation method characterized by comprising
Based on target user to the operation information of at least one project, selection is in the preset time at current time or present count
The project of amount；
Based on the target user to the operation order of selected project, using the second of selected project indicate vector and
The first of the target user indicates vector, generates input data；
In the input data cuit recommended models, will obtain the prediction of prediction term purpose indicates vector, the project recommendation
Model is obtained by project recommendation model training method as claimed in claim 1 to 7；
Calculate the similarity between the prediction expression vector and the second expression vector of project；It is similar based on what is be calculated
Degree determines the project recommended to the target user and recommends the target user.
9. a kind of project recommendation model training apparatus characterized by comprising
Indicate that vector generation module generates each institute for the operation information based at least one user at least one project
Stating the first of user indicates that the second of vector and each project indicates vector；
Training data generation module, for selecting at least one sample of users, at least one described sample of users be it is described at least
One user's is some or all of；And for each of at least one described sample of users, it is based on the sample of users pair
First table of the second expression vector and the sample of users of the operation order of the project operated by it and its operated project
Show vector, generates the training data of the project recommendation model；
Training module, for being trained using the training data to the project recommendation model.
10. a kind of project recommendation device characterized by comprising
Selecting module selects presetting apart from current time for the operation information based on target user at least one project
In time or the project of preset quantity；
Input data generation module, for the operation order based on the target user to selected project, using selected
Project second indicate that the first of vector and the target user indicates vector, generate input data；
Vector generation module is indicated, for the prediction of prediction term purpose in the input data cuit recommended models, will to be obtained
Indicate that vector, the project recommendation model are obtained by project recommendation model training method as claimed in claim 1 to 7；
Similarity calculation module, for calculating the similarity between the prediction expression vector and the second expression vector of project；
Recommending module, for determining the project recommended to the target user and recommending institute based on the similarity being calculated
State target user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811152399.XA CN109389168B (en) | 2018-09-29 | 2018-09-29 | Project recommendation model training method, project recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811152399.XA CN109389168B (en) | 2018-09-29 | 2018-09-29 | Project recommendation model training method, project recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109389168A true CN109389168A (en) | 2019-02-26 |
| CN109389168B CN109389168B (en) | 2021-07-13 |
Family
ID=65418419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811152399.XA Active CN109389168B (en) | 2018-09-29 | 2018-09-29 | Project recommendation model training method, project recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109389168B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902446A (en) * | 2019-04-09 | 2019-06-18 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information prediction model |
| CN109918574A (en) * | 2019-03-28 | 2019-06-21 | 北京卡路里信息技术有限公司 | Item recommendation method, device, equipment and storage medium |
| CN110297848A (en) * | 2019-07-09 | 2019-10-01 | 深圳前海微众银行股份有限公司 | Recommended models training method, terminal and storage medium based on federation's study |
| CN110689937A (en) * | 2019-09-05 | 2020-01-14 | 郑州金域临床检验中心有限公司 | Coding model training method, system and equipment and detection item coding method |
| CN110851718A (en) * | 2019-11-11 | 2020-02-28 | 重庆邮电大学 | Movie recommendation method based on long-time memory network and user comments |
| CN111091888A (en) * | 2019-11-25 | 2020-05-01 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
| CN111242748A (en) * | 2020-02-21 | 2020-06-05 | 腾讯科技（深圳）有限公司 | Method, apparatus, and storage medium for recommending items to a user |
| WO2020228514A1 (en) * | 2019-05-13 | 2020-11-19 | 腾讯科技（深圳）有限公司 | Content recommendation method and apparatus, and device and storage medium |
| CN111026858B (en) * | 2019-11-29 | 2021-06-11 | 腾讯科技（深圳）有限公司 | Project information processing method and device based on project recommendation model |
| CN113312644A (en) * | 2021-06-15 | 2021-08-27 | 杭州金智塔科技有限公司 | Cross-domain recommendation model training method and system based on privacy protection |
| CN113538185A (en) * | 2021-07-15 | 2021-10-22 | 山西安弘检测技术有限公司 | Simulation training method and device for radiological hygiene detection |
| WO2022057658A1 (en) * | 2020-09-15 | 2022-03-24 | 腾讯科技（深圳）有限公司 | Method and apparatus for training recommendation model, and computer device and storage medium |
| CN114418246A (en) * | 2022-03-30 | 2022-04-29 | 深圳谱盈科技有限公司 | Project prediction system based on big data |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101814068A (en) * | 2009-02-24 | 2010-08-25 | 日电（中国）有限公司 | Rating prediction based project recommending method for time-sequence control and system thereof |
| CN103823908A (en) * | 2014-03-21 | 2014-05-28 | 北京飞流九天科技有限公司 | Method and server for content recommendation on basis of user preferences |
| CN104935963A (en) * | 2015-05-29 | 2015-09-23 | 中国科学院信息工程研究所 | Video recommendation method based on timing sequence data mining |
| CN105069122A (en) * | 2015-08-12 | 2015-11-18 | 天津大学 | Personalized recommendation method and recommendation apparatus based on user behaviors |
| CN105354330A (en) * | 2015-11-27 | 2016-02-24 | 南京邮电大学 | Sparse data preprocessing based collaborative filtering recommendation method |
| CN105354729A (en) * | 2015-12-14 | 2016-02-24 | 电子科技大学 | Commodity recommendation method in electronic commerce system |
| CN105740444A (en) * | 2016-02-02 | 2016-07-06 | 桂林电子科技大学 | User score-based project recommendation method |
| CN106294758A (en) * | 2016-09-23 | 2017-01-04 | 华南师范大学 | Collaborative recommendation method based on the change of user cognition degree |
| CN106503028A (en) * | 2015-09-08 | 2017-03-15 | 北京邮电大学 | Recommend method and system |
| CN107368519A (en) * | 2017-06-05 | 2017-11-21 | 桂林电子科技大学 | A kind of cooperative processing method and system for agreeing with user interest change |
| US9904949B1 (en) * | 2013-06-21 | 2018-02-27 | Amazon Technologies, Inc. | Product recommendations |
| CN108243357A (en) * | 2018-01-25 | 2018-07-03 | 北京搜狐新媒体信息技术有限公司 | A kind of video recommendation method and device |
-
2018
- 2018-09-29 CN CN201811152399.XA patent/CN109389168B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101814068A (en) * | 2009-02-24 | 2010-08-25 | 日电（中国）有限公司 | Rating prediction based project recommending method for time-sequence control and system thereof |
| US9904949B1 (en) * | 2013-06-21 | 2018-02-27 | Amazon Technologies, Inc. | Product recommendations |
| CN103823908A (en) * | 2014-03-21 | 2014-05-28 | 北京飞流九天科技有限公司 | Method and server for content recommendation on basis of user preferences |
| CN104935963A (en) * | 2015-05-29 | 2015-09-23 | 中国科学院信息工程研究所 | Video recommendation method based on timing sequence data mining |
| CN105069122A (en) * | 2015-08-12 | 2015-11-18 | 天津大学 | Personalized recommendation method and recommendation apparatus based on user behaviors |
| CN106503028A (en) * | 2015-09-08 | 2017-03-15 | 北京邮电大学 | Recommend method and system |
| CN105354330A (en) * | 2015-11-27 | 2016-02-24 | 南京邮电大学 | Sparse data preprocessing based collaborative filtering recommendation method |
| CN105354729A (en) * | 2015-12-14 | 2016-02-24 | 电子科技大学 | Commodity recommendation method in electronic commerce system |
| CN105740444A (en) * | 2016-02-02 | 2016-07-06 | 桂林电子科技大学 | User score-based project recommendation method |
| CN106294758A (en) * | 2016-09-23 | 2017-01-04 | 华南师范大学 | Collaborative recommendation method based on the change of user cognition degree |
| CN107368519A (en) * | 2017-06-05 | 2017-11-21 | 桂林电子科技大学 | A kind of cooperative processing method and system for agreeing with user interest change |
| CN108243357A (en) * | 2018-01-25 | 2018-07-03 | 北京搜狐新媒体信息技术有限公司 | A kind of video recommendation method and device |
Non-Patent Citations (4)
| Title |
|---|
| ZHIJUN ZHANG 等: ""Social recommendation model combining trust propagation and sequential behaviors"", 《HTTPS://LINK.SPRINGER.COM/ARTICLE/10.1007/S10489-015-0681-Y》 * |
| 孙光福 等: ""基于时序行为的协同过滤推荐算法"", 《软件学报》 * |
| 张志军: ""社交网络中个性化推荐模型及算法研究"", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 蔡海尼 等: ""基于时序模型和矩阵分解的推荐算法"", 《计算机应用研究》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918574A (en) * | 2019-03-28 | 2019-06-21 | 北京卡路里信息技术有限公司 | Item recommendation method, device, equipment and storage medium |
| CN109902446A (en) * | 2019-04-09 | 2019-06-18 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information prediction model |
| CN109902446B (en) * | 2019-04-09 | 2023-07-11 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information prediction model |
| WO2020228514A1 (en) * | 2019-05-13 | 2020-11-19 | 腾讯科技（深圳）有限公司 | Content recommendation method and apparatus, and device and storage medium |
| CN110297848A (en) * | 2019-07-09 | 2019-10-01 | 深圳前海微众银行股份有限公司 | Recommended models training method, terminal and storage medium based on federation's study |
| CN110297848B (en) * | 2019-07-09 | 2024-02-23 | 深圳前海微众银行股份有限公司 | Recommendation model training method, terminal and storage medium based on federal learning |
| CN110689937A (en) * | 2019-09-05 | 2020-01-14 | 郑州金域临床检验中心有限公司 | Coding model training method, system and equipment and detection item coding method |
| CN110851718B (en) * | 2019-11-11 | 2022-06-28 | 重庆邮电大学 | Movie recommendation method based on long and short term memory network and user comments |
| CN110851718A (en) * | 2019-11-11 | 2020-02-28 | 重庆邮电大学 | Movie recommendation method based on long-time memory network and user comments |
| CN111091888A (en) * | 2019-11-25 | 2020-05-01 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
| CN111091888B (en) * | 2019-11-25 | 2023-10-31 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
| CN111026858B (en) * | 2019-11-29 | 2021-06-11 | 腾讯科技（深圳）有限公司 | Project information processing method and device based on project recommendation model |
| CN111242748B (en) * | 2020-02-21 | 2023-04-07 | 腾讯科技（深圳）有限公司 | Method, apparatus, and storage medium for recommending items to a user |
| CN111242748A (en) * | 2020-02-21 | 2020-06-05 | 腾讯科技（深圳）有限公司 | Method, apparatus, and storage medium for recommending items to a user |
| WO2022057658A1 (en) * | 2020-09-15 | 2022-03-24 | 腾讯科技（深圳）有限公司 | Method and apparatus for training recommendation model, and computer device and storage medium |
| CN113312644B (en) * | 2021-06-15 | 2022-05-24 | 杭州金智塔科技有限公司 | Cross-domain recommendation model training method and system based on privacy protection |
| CN113312644A (en) * | 2021-06-15 | 2021-08-27 | 杭州金智塔科技有限公司 | Cross-domain recommendation model training method and system based on privacy protection |
| CN113538185A (en) * | 2021-07-15 | 2021-10-22 | 山西安弘检测技术有限公司 | Simulation training method and device for radiological hygiene detection |
| CN114418246A (en) * | 2022-03-30 | 2022-04-29 | 深圳谱盈科技有限公司 | Project prediction system based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109389168B (en) | 2021-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389168A (en) | Project recommendation model training method, item recommendation method and device | |
| CN109902706B (en) | Recommendation method and device | |
| CN109767301B (en) | Recommendation method and system, computer device and computer readable storage medium | |
| EP3862893A1 (en) | Recommendation model training method, recommendation method, device, and computer-readable medium | |
| CN108460619B (en) | Method for providing collaborative recommendation model fusing explicit and implicit feedback | |
| CN111506820B (en) | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium | |
| CN109903103B (en) | Method and device for recommending articles | |
| CN105844508B (en) | Commodity recommendation method based on dynamic periodic neural network | |
| CN116664719B (en) | Image redrawing model training method, image redrawing method and device | |
| CN115270001B (en) | Privacy protection recommendation method and system based on cloud collaborative learning | |
| CN109033453B (en) | RBM and differential privacy protection based clustering movie recommendation method and system | |
| CN109670161A (en) | Commodity similarity calculating method and device, storage medium, electronic equipment | |
| CN110321473A (en) | Diversity preference information method for pushing, system, medium and equipment based on multi-modal attention | |
| CN110287415A (en) | A kind of content recommendation method, device and calculate equipment | |
| CN112308669A (en) | Commodity recommendation method and device, storage medium and terminal equipment | |
| CN110781377B (en) | Article recommendation method and device | |
| CN113610610B (en) | Session recommendation method and system based on graph neural network and comment similarity | |
| CN105260458A (en) | Video recommendation method for display apparatus and display apparatus | |
| CN112364236A (en) | Target object recommendation system, method and device, and data processing method and device | |
| CN112559904B (en) | Conversational social recommendation method based on door mechanism and multi-modal graph network | |
| CN111652673B (en) | Intelligent recommendation method, device, server and storage medium | |
| CN110020910A (en) | Object recommendation method and apparatus | |
| Dugaro et al. | Physical properties of terrestrial planets and water delivery in the habitable zone using N-body simulations with fragmentation | |
| CN112734519B (en) | Commodity recommendation method based on convolution self-encoder network | |
| CN106919946A (en) | A kind of method and device of audience selection |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 100070, No. 101-8, building 1, 31, zone 188, South Fourth Ring Road, Beijing, Fengtai District Applicant after: Guoxin Youyi Data Co., Ltd Address before: 100070, No. 188, building 31, headquarters square, South Fourth Ring Road West, Fengtai District, Beijing Applicant before: SIC YOUE DATA Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |