CN114912035B - Short video recommendation method and system based on Bayesian probability matrix decomposition - Google Patents
Short video recommendation method and system based on Bayesian probability matrix decomposition Download PDFInfo
- Publication number
- CN114912035B CN114912035B CN202210577571.6A CN202210577571A CN114912035B CN 114912035 B CN114912035 B CN 114912035B CN 202210577571 A CN202210577571 A CN 202210577571A CN 114912035 B CN114912035 B CN 114912035B
- Authority
- CN
- China
- Prior art keywords
- user
- short video
- scoring
- data
- model
- 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.)
- Active
Links
- 239000011159 matrix material Substances 0.000 title claims abstract description 89
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000000342 Monte Carlo simulation Methods 0.000 claims abstract description 10
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 3
- 239000013598 vector Substances 0.000 claims description 39
- 238000005070 sampling Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000013461 design Methods 0.000 description 8
- 238000001914 filtration Methods 0.000 description 6
- 238000013459 approach Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000008570 general process Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—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
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The invention discloses a short video recommendation method and a short video recommendation system based on Bayesian probability matrix decomposition, wherein the short video recommendation method comprises the following steps: acquiring original sparse scoring data of a user through a data acquisition module; preprocessing data through a data processing module, dividing a data set into a training set and a testing set, and processing the training set and the testing set into a scoring matrix; introducing a Bayesian network on the basis of a probability matrix decomposition model through a model training module, adding a constraint matrix and a Logistic function, building a constrained Bayesian probability matrix decomposition model, and carrying out model training by using a Markov chain Monte Carlo method; loading the preprocessed original sparse scoring matrix, predicting the scoring information of the unknown specific user on the short video by using a trained model through a prediction module, and filling the original sparse scoring matrix according to a prediction result; and finally, sorting the scoring data in the scoring matrix after filling in a descending order through a recommendation module, and pushing the corresponding short video for the specific user.
Description
Technical Field
The invention relates to the technical field of Internet video recommendation, in particular to a short video recommendation method and system based on Bayesian probability matrix decomposition.
Background
With the rapid development of information technology, the number of software products on the mobile end and the web end is rapidly increased, and in order to attract and retain users, collaborative filtering algorithms (collaborative filtering algorithm) are often adopted to recommend the users to the favorite commodities, so that the flow of the software is ensured, and more profits are obtained. The collaborative filtering recommendation algorithm can not well solve the problem of low prediction accuracy of scoring sparse users. In order to improve the prediction precision, a recommendation method based on a probability matrix decomposition model is proposed, although the method reduces the prediction error, the regularization parameters are required to be manually adjusted, and the model enables the characteristics of users with less scores to be close to the average value of the prior distribution, so that the scoring prediction of the users with less scores is close to the average score of the object, and the accuracy of recommendation is affected; a recommendation method based on a constrained probability matrix decomposition model is provided, and although the method avoids that the features of users with less scores are close to the average value of prior distribution, the prediction accuracy is still low; a recommendation method for recommending heterogeneous similarity models based on implicit feedback is provided, and although prediction accuracy is improved, algorithm complexity is increased, and prediction errors are still high.
Disclosure of Invention
The invention aims to provide a short video recommendation method and a short video recommendation system based on Bayesian probability matrix decomposition for the problem of high prediction error of users with fewer scores by the collaborative filtering algorithm.
The invention aims at realizing the following technical scheme:
the system mainly comprises five modules, namely a data acquisition module, a data processing module, a model training module, a prediction module and a recommendation module.
A short video recommendation method and system based on Bayesian probability matrix decomposition mainly comprises the steps of obtaining original scoring data of a short video by a user, preprocessing the obtained original sparse scoring data, constructing a constrained Bayesian probability matrix decomposition model (L-CBPMF model) using a Logistic function, training the model by using a Markov chain Monte Carlo method, predicting according to the trained model, and recommending the short video according to a prediction result.
In a further design scheme of the invention, the data acquisition module acquires data information, wherein the data information comprises user information, short video information and scoring information of N users on M short videos; the user information mainly collects user id, the short video information mainly collects short video id, and the scoring information mainly collects praise number of the user on a certain type of short video.
In a further design scheme of the invention, the data processing module preprocesses the acquired sparse scoring data, and the data is processed according to 8:2, wherein 80% of the training sets are training sets, and the remaining 20% of the training sets are testing sets, the training sets are used for training models, and the testing sets are used for evaluating generalization capability of the models. And constructing scoring information of N users on M short videos into a sparse scoring matrix, wherein the existing data in the scoring matrix is the scoring information of the users on the short videos, and the hollow part of the scoring matrix is that the users do not score the short videos.
In a further design scheme of the invention, the model training module uses a Bayesian network on the basis of a traditional probability matrix decomposition model, and adds a constraint matrix W k, and the user characteristics are constrained through the constraint matrix, namely, a constraint vector except for an original characteristic vector is given to each short video, so that the average value of all short video constraint vectors scored by each user influences the characteristic vector of the user; and using the Logistic function to represent the nonlinear relation of the potential factors, and constructing a constrained Bayesian probability matrix decomposition model using the Logistic function. And loading the preprocessed data, and training a model by using a Markov chain Monte Carlo method.
In a further design scheme of the invention, the prediction module predicts the scoring information of the unknown specific user on the short video by using the trained model, and fills the original sparse scoring matrix according to the prediction result, namely fills the scoring data of the specific user on the specific short video to the position corresponding to the sparse matrix.
In a further design scheme of the invention, the recommendation module performs descending order sequencing on the scoring data in the scoring matrix after filling, and recommends the corresponding short video for the specific user, namely pushes the short video of the corresponding category to the corresponding user according to the preference degree of the user.
Further, in the constrained Bayesian probability matrix decomposition model (L-CBPMF) using a Logistic function, N users and M short videos form an N×M-dimensional scoring matrix R ij, and an element R in the matrix R ij represents the score of a user i on a short video j. In the model, the attitude or preference of a default user is determined by a few unobserved factors, and the number of potential features is set to be D, so that a D×N dimensional matrix U represents a potential feature matrix of the user, a D×M dimensional matrix V represents a potential feature matrix of the short video, U i represents a potential feature vector of the user i, and V j represents a potential feature vector of the short video j; w 0,W1,W2 is an identity matrix, v 0,v1,v2 is a degree of freedom, Θ U,ΘV,ΘW,ΘB is a hyper-parameter of the model, μ U,μV,μW,μB is a mean parameter of gaussian distribution, Λ U,ΛV,ΛW,ΛB is a variance matrix of gaussian distribution.
Further, to better express the nonlinear relationship between the potential factors, a gaussian distribution with a variance of α -1 is set where the score R ij of the user i on the short video j obeys the mean of B i g(YTVj). Thereby constructing a probabilistic objective function of the model LCBPMF, the scoring condition distribution is as follows: Wherein I ij is an indication function, where the user has scored the short video, I ij =1, whereas I ij＝0,Bi represents the user I scoring scale parameter, element R in matrix R ij represents the user I scoring the short video j, Representing a gaussian distribution function, g (x) is a Logistic function, used to convey the nonlinear relationship of potential eigenvalues.
Further, the model is based on a Bayesian network structure, a potential similarity constraint matrix W k is added into potential feature vectors of users, namely, constraint vectors except the feature vectors are allocated to each short video, so that the feature values of the users are influenced by the scored constraint vectors, and the feature values of the users are prevented from approaching to the average value of prior distribution.
Further, based on the formulaThe user's feature vector may be extracted and updated to generate a new user feature vector by constraining user U i. The indicator function I ik indicates whether the user I scores the short video k. If scored, I ik = 1, otherwise I ik = 0. After accumulating the scores W k of the user i on all the short videos k, taking an average value, and adding the obtained result with the offset value of the prior distribution average value of the user i to obtain a new user feature vector Y i.
Further, the general process of solving the posterior probability is complex, the joint probability is not easy to obtain, the model adopts a Markov chain Monte Carlo method to extract samples, and the method is based on a formulaApproximately equal to the objective function, where T is the number of samples, samplesIs generated by running a Markov chain whose plateau distribution is the posterior distribution of model parameters and hyper-parameters { U, V, W, B, Θ U,ΘV,ΘW,ΘB }.
In a further design of the present invention, the model training process is as follows: firstly, constructing a Markov chain which is stably distributed as expected joint probability by utilizing conditional probability, then, sampling super parameters T times by utilizing a Gibbs sampling method, then, traversing and updating the eigenvalue of U, V, W, B vectors, training a model on a training set, performing error verification on a testing set to obtain a prediction score, and filling an original sparse scoring matrix.
The invention has the following outstanding beneficial effects:
The short video recommending method and system based on Bayesian probability matrix decomposition can solve the problem of low recommending accuracy caused by less scoring data of users, can more accurately predict the preference degree of users for certain types of short videos, uses Bayesian networks on the basis of a traditional probability matrix decomposition model, adds constraint matrix W k to enable the scored constraint vectors of the users to influence the characteristic values of the users, uses a Logistic function to transfer the nonlinear relation of potential factors, solves the problems of low recommending accuracy and poor recommending quality of the users with sparse scoring by a collaborative filtering algorithm under the condition that the scoring data of the users are insufficient, and reduces scoring predicting errors, improves recommending quality and user satisfaction under the condition that algorithm complexity is not increased.
Drawings
FIG. 1 is a schematic flow chart of one embodiment of the present invention;
FIG. 2 is a schematic diagram of an embodiment of the present invention;
FIG. 3 is a schematic diagram of a model structure according to an embodiment of the present invention;
Detailed Description
The invention will be further described with reference to the accompanying drawings and examples.
The embodiment of the invention provides a short video recommendation method based on Bayesian probability matrix decomposition, which comprises the following steps: the method comprises the steps of obtaining scoring data of a user on a short video, wherein the scoring data comprises user information and short video information, and the scoring information of the user on the short video; and processing the scoring data, dividing the data into a training set and a testing set, and simultaneously converting the scoring data into a sparse scoring matrix of the user and the corresponding short video. An L-CBPMF model is constructed, and parameters of the model comprise: n users and M short videos form an N multiplied by M scoring matrix R ij, and an element R in the matrix R ij represents the score of the user i on the short video j. In the model, the attitudes or preferences of default users are determined by a few unobserved factors, the number of potential features is set to be D, then a D×N dimensional matrix U represents the potential feature matrix of the user, a D×M dimensional matrix V represents the potential feature matrix of the short video, U i represents the potential feature vector of the user i, V j represents the potential feature vector of the short video j, Y i represents the new user feature vector, W k represents the potential similarity constraint matrix, W 0,W1,W2 is the identity matrix, V 0,v1,v2 is the degree of freedom, mu U,μV,μW,μB is the mean parameter of the Gaussian distribution, and Λ U,ΛV,ΛW,ΛB is the variance matrix of the Gaussian distribution. Training by using a Markov chain Monte Carlo method after model construction is completed, predicting unknown user scoring data by the trained model, filling an original sparse scoring matrix according to a prediction result, and finally realizing accurate recommendation according to the prediction scoring data.
Referring to fig. 1 and fig. 2, the embodiment of the invention provides a short video recommendation method and system based on bayesian probability matrix decomposition, which further comprises a data acquisition module 1, a data processing module 2, a model training module 3, a prediction module 4 and a recommendation module 5, wherein data information is acquired through the data acquisition module 1, namely original scoring data of a short video by a user is acquired, the data comprises user information, short video information and scoring information of the short video by the user, the scoring data is preprocessed through the data processing module 2, the scoring data is transformed into a sparse scoring matrix, and the scoring data is processed according to 8:2 dividing the data set into a training set and a testing set; a constrained Bayesian probability matrix decomposition model using a Logistic function is constructed through a model training module 3, a Markov chain Monte Carlo method is used for training the model, unknown user scoring information is predicted through a prediction module 4 by using the trained model, an original user sparse scoring matrix is filled, the filled scoring values are ordered according to descending order through a recommendation module 5, and corresponding short videos are pushed to corresponding users according to the scoring data.
In a further design of the present invention, in the constrained bayesian probability matrix decomposition model (L-CBPMF) using a Logistic function, N users and M short videos form an n×m-dimensional scoring matrix R, and an element R in the matrix R ij represents a score of a user i on a short video j. In the model, the attitudes or preferences of the default users are determined by a few unobserved factors, and the number of potential features is set to be D, so that a D×N-dimensional matrix U represents the potential feature matrix of the users, and a D×M-dimensional matrix V represents the potential feature matrix of the short videos.
Furthermore, the model is based on a Bayesian framework, a potential similarity constraint matrix W k is added into the user feature vector, and a constraint vector outside the feature vector is distributed for each short video, so that the constraint vectors of all short videos scored by the user affect the feature values of the short videos, and the situation that the feature values of the user approach to the average value of prior distribution is effectively avoided.
Further, to better represent the nonlinear relationship of the potential factors, the score R ij for all short videos j for default user i obeys a Gaussian distribution with average B i g(YTVj and variance α -1.
Further, setting the similarity constraint matrix as W, the feature vector of the new user is expressed using the following formula, namelyThe function I ik user is instructed to indicate whether user I scored the short video k by constraining user feature vector U i to generate a new user feature vector. If scored, I ik =1, otherwise I ik =0. After accumulating the scores W k of the user i for all items k, the average is taken and the obtained result is then added to the offset value of the a priori distribution average of the user i to obtain a new user feature vector.
Further, under normal conditions, the solution posterior probability is complex because the joint probability is not easily obtained, so the invention uses the Markov chain Monte Carlo method to extract the sample, and uses the following formula to approach the complex objective function, wherein the formula is as follows: Wherein I ij is an indication function, where the user has scored the short video, I ij =1, whereas I ij＝0,Bi represents the user I scoring scale parameter, element R in matrix R ij represents the user I scoring the short video j, The Gaussian distribution function is represented, g (x) is a Logistic function and is used for transmitting a nonlinear relation of potential characteristic factors, and T is the sampling times.
Further, a sample is generated by running a Markov chainThe plateau distribution of the Markov chain is the posterior distribution of model parameters and superparameters { U, V, W, B, Θ U,ΘV,ΘW,ΘB }. The monte carlo based method progressively yields accurate results.
In a further design scheme of the invention, the specific process of model training is as follows: firstly, a Markov chain with stable distribution being expected to be combined distribution is constructed by using conditional probability, then, a Gibbs sampling method is adopted to sample the super parameter Θ U,ΘV,ΘW,ΘB, and then, the characteristic value of the U, V, W, B vector is updated in an iterative mode. And finally, traversing the data set, and calculating a prediction score and a prediction error, thereby obtaining the prediction precision of the model.
The short video recommending method and system based on Bayesian probability matrix decomposition can solve the problem of low recommending accuracy caused by less scoring data of users, can more accurately predict the preference degree of users for certain types of short videos, uses Bayesian networks on the basis of a traditional probability matrix decomposing model, adds constraint matrix W k to enable the scored constraint vectors of the users to influence the characteristic values of the users, uses a Logistic function to transfer the nonlinear relation of potential factors, solves the problems of low recommending accuracy and poor recommending quality of the users with sparse scoring by a collaborative filtering algorithm under the condition that the scoring data of the users is insufficient, and reduces scoring prediction errors and improves recommending quality under the condition that algorithm complexity is not increased.
The above is a preferred embodiment of the present invention, and all changes made according to the technical solution of the present invention belong to the protection scope of the present invention when the generated functional effects do not exceed the scope of the technical solution of the present invention.
Claims (5)
1. The short video recommendation method and system based on Bayesian probability matrix decomposition are characterized in that the system comprises five modules, namely a data acquisition module, a data processing module, a model training module, a prediction module and a recommendation module; acquiring original sparse scoring data of a user through a data acquisition module; preprocessing data through a data processing module, dividing a data set into a training set and a testing set, and processing the data into a scoring matrix; building a new constrained Bayesian probability matrix decomposition model through a model training module, and training the model by using a Markov chain Monte Carlo method; predicting the scoring information of the unknown specific user on the short video by using a trained model through a prediction module, and filling an original sparse scoring matrix according to a prediction result; sorting the scoring data in the scoring matrix after filling in a descending order through a recommendation module, and pushing the corresponding short video for a specific user; the model training module uses a Bayesian network on the basis of a traditional probability matrix decomposition model, and adds a constraint matrix W k, the constraint matrix is used for constraining user characteristics, a constraint vector except an original feature vector is given to each short video, the average value of all short video constraint vectors scored by each user influences the feature vector of the user, a constraint vector except the feature vector is distributed to each short video, the constraint vector of all short videos scored by the user influences the feature value of the short video, and the feature value of the user is prevented from approaching the average value of prior distribution; using a Logistic function to represent a nonlinear relation of potential factors, and constructing a constrained Bayesian probability matrix decomposition model using the Logistic function; loading the preprocessed data, and training a model by using a Markov chain Monte Carlo method; based on the formulaThe feature vector of the user can be extracted and updated, and a new user feature vector Y i is generated by constraining the user U i; indication function I ik indicates whether user I scored short video k; if scored, then I ik = 1, otherwise I ik = 0; accumulating the scores W k of the user i on all the short videos k, taking the average value of the scores W k, and then adding the obtained result with the offset value of the prior distribution average value of the user i to obtain a new user feature vector Y i; the model adopts a Markov chain Monte Carlo method to extract samples, and is based on a formula Approximately equal to the objective function; wherein: p is the probability objective function of the probability,Vj t,Representing U, V, W, B eigenvalues of the eigenvector generated by a markov chain T times, T representing the number of samples,Representing a predictive scoring matrix, R representing an original scoring matrix, Θ 0 being a hyper-parameter of the model; the specific training process of the model is as follows: firstly, constructing a Markov chain with stable distribution as expected joint distribution by using conditional probability, then adopting a Gibbs sampling method to sample the super parameter Θ U,ΘV,ΘW,ΘB, and then iteratively updating the characteristic value of the U, V, W, B vector; finally, a prediction score and a prediction error are calculated, so that the prediction precision of the model is obtained.
2. The short video recommendation method and system based on bayesian probability matrix decomposition according to claim 1, wherein the data acquisition module acquires data information including user information, short video information, and scoring information of M short videos by N users; the user information is used for collecting user id, the short video information is used for collecting short video id, and the number of praise of the information collecting user on a certain type of short video is scored.
3. The short video recommendation method and system based on bayesian probability matrix decomposition according to claim 1, wherein the data processing module pre-processes the acquired sparse scoring data according to 8:2, dividing a training set and a testing set, wherein 80% of the training set and the remaining 20% of the testing set, and constructing scoring information of N users on M short videos into an original sparse scoring matrix.
4. The short video recommendation method and system based on bayesian probability matrix decomposition according to claim 1, wherein the prediction module predicts the unknown scoring information of the specific user on the short video by using the trained model, and fills the original sparse scoring matrix according to the prediction result, namely fills the scoring data of the specific user on the specific short video to the position corresponding to the sparse matrix.
5. The short video recommendation method and system based on bayesian probability matrix decomposition according to claim 1, wherein the recommendation module performs descending order sorting on the scoring data in the scoring matrix after filling, and recommends corresponding short videos for specific users, namely pushes corresponding short videos to the corresponding users according to user preference degrees.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210577571.6A CN114912035B (en) | 2022-05-25 | 2022-05-25 | Short video recommendation method and system based on Bayesian probability matrix decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210577571.6A CN114912035B (en) | 2022-05-25 | 2022-05-25 | Short video recommendation method and system based on Bayesian probability matrix decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114912035A CN114912035A (en) | 2022-08-16 |
| CN114912035B true CN114912035B (en) | 2024-07-12 |
Family
ID=82769681
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210577571.6A Active CN114912035B (en) | 2022-05-25 | 2022-05-25 | Short video recommendation method and system based on Bayesian probability matrix decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912035B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109840833A (en) * | 2019-02-13 | 2019-06-04 | 苏州大学 | Bayes's collaborative filtering recommending method |
| AU2020101604A4 (en) * | 2020-07-31 | 2020-09-10 | The University of Xinjiang | A Recommendation with Item Cooccurrence based on Metric Factorization |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8880439B2 (en) * | 2012-02-27 | 2014-11-04 | Xerox Corporation | Robust Bayesian matrix factorization and recommender systems using same |
| CN109947987B (en) * | 2019-03-22 | 2022-10-25 | 江西理工大学 | Cross collaborative filtering recommendation method |
-
2022
- 2022-05-25 CN CN202210577571.6A patent/CN114912035B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109840833A (en) * | 2019-02-13 | 2019-06-04 | 苏州大学 | Bayes's collaborative filtering recommending method |
| AU2020101604A4 (en) * | 2020-07-31 | 2020-09-10 | The University of Xinjiang | A Recommendation with Item Cooccurrence based on Metric Factorization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114912035A (en) | 2022-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tharwat et al. | A survey on active learning: State-of-the-art, practical challenges and research directions | |
| US20230229891A1 (en) | Reservoir computing neural networks based on synaptic connectivity graphs | |
| US11568201B2 (en) | Predicting neuron types based on synaptic connectivity graphs | |
| JP2021524099A (en) | Systems and methods for integrating statistical models of different data modality | |
| CN112967088A (en) | Marketing activity prediction model structure and prediction method based on knowledge distillation | |
| JP5308360B2 (en) | Automatic content classification apparatus, automatic content classification method, and automatic content classification program | |
| CN111243736A (en) | Survival risk assessment method and system | |
| CN118132849B (en) | A recommendation method for steel structure welding process quality management based on big data processing | |
| CN108009571A (en) | A kind of semi-supervised data classification method of new direct-push and system | |
| CN113535947B (en) | Multi-label classification method and device for incomplete data with missing labels | |
| Ma et al. | Improving uncertainty calibration of deep neural networks via truth discovery and geometric optimization | |
| CN117891939A (en) | Text classification method combining particle swarm algorithm with CNN convolutional neural network | |
| CN112259157A (en) | Protein interaction prediction method | |
| CN114912035B (en) | Short video recommendation method and system based on Bayesian probability matrix decomposition | |
| Cheng et al. | A Memory Guided Transformer for Time Series Forecasting | |
| Gupta et al. | Machine learning algorithms | |
| CN114692724B (en) | Training method of data classification model, data classification method and device | |
| CN116701873A (en) | Behavior dangerous trend prediction method and system based on subjective and objective information fusion representation | |
| CN115712777A (en) | Ranking method of literature recommendation system based on logistic regression | |
| CN115712855A (en) | Self-learning-based label rule generation method and device | |
| Liu et al. | Implicit generative prior for Bayesian neural networks | |
| Muddana et al. | Python for Machine Learning | |
| CN112949599A (en) | Candidate content pushing method based on big data | |
| CN118070788B (en) | A method and related equipment for standardizing institution names based on knowledge graph | |
| Sachenko et al. | Evaluation of ensemble machine learning models for movie recommendation systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |