CN108874960A - Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study - Google Patents
Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study Download PDFInfo
- Publication number
- CN108874960A CN108874960A CN201810575724.7A CN201810575724A CN108874960A CN 108874960 A CN108874960 A CN 108874960A CN 201810575724 A CN201810575724 A CN 201810575724A CN 108874960 A CN108874960 A CN 108874960A
- Authority
- CN
- China
- Prior art keywords
- user
- curriculum video
- noise reduction
- encoding encoder
- reduction self
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention discloses the curriculum video proposed algorithms based on noise reduction self-encoding encoder mixed model in a kind of on-line study.This method includes two parts of score in predicting algorithm of user characteristics extraction algorithm and curriculum video feature extraction algorithm, adaptive matrix decomposition model based on noise reduction self-encoding encoder.Provide the specific steps of user characteristics extraction algorithm and curriculum video feature extraction algorithm based on noise reduction self-encoding encoder.Provide the specific steps of the score in predicting algorithm of adaptive matrix decomposition model.Compared with existing course video recommendations algorithm, the present invention can carry out Automatic Feature Extraction from rating matrix and content information using noise reduction self-encoding encoder；Meanwhile the feature extracted being dissolved among matrix decomposition, and establish content information and the non-linear interactive relation of user-curriculum video rating matrix, higher recommendation accuracy can be reached.
Description
Technical field
The present invention relates to the technical fields that curriculum video in on-line study is recommended, and in particular to is based in a kind of on-line study
The curriculum video proposed algorithm of noise reduction self-encoding encoder mixed model.
Background technique
In recent years, with the fast development of on-line study, the recommendation of curriculum video is helping people to overcome information overload side
Face plays an increasingly important role, and how to carry out accurate modeling to user preference and recommend interested curriculum video to it
It is a hot spot technology problem.
Mixed recommendation is a kind of important method that curriculum video is recommended in on-line study, still, most of mixed recommendation
Algorithm depends on manual feature extraction, and the correlativity between attribute and attribute for causing some complicated is difficult to handle, so not
Only the waste plenty of time is in feature extraction, and the feature extracted is likely to inaccuracy, this is existing on-line study
One deficiency of middle curriculum video mixing proposed algorithm.
In addition, most of mixing proposed algorithm is all to carry out simple linear weighted function to the result of a variety of proposed algorithms to ask
With, cause the accuracy recommended not high enough, this be in existing on-line study curriculum video mixing proposed algorithm another is insufficient.
Summary of the invention
For the deficiency of curriculum video proposed algorithm in existing on-line study, the present invention provides bases in a kind of on-line study
In the curriculum video proposed algorithm of noise reduction self-encoding encoder mixed model, which includes the user characteristics based on noise reduction self-encoding encoder
Two parts of score in predicting algorithm of extraction algorithm and curriculum video feature extraction algorithm, adaptive matrix decomposition model, wherein
Feature extraction algorithm provided by the invention based on noise reduction self-encoding encoder can be extracted from history scoring and attribute information automatically
The validity feature of user and curriculum video out avoids the cumbersome process of manual extraction feature；Adaptive square provided by the invention
The score in predicting algorithm user that extracts noise reduction self-encoding encoder of battle array decomposition model and curriculum video feature incorporate matrix decomposition it
In, content information and the non-linear interactive relation of user-curriculum video rating matrix are established, is recommended with existing curriculum video
Algorithm, which is compared, can obtain higher recommendation accuracy.
The present invention is characterized in that including the following contents：
1, a kind of curriculum video proposed algorithm of the mixed model based on noise reduction self-encoding encoder
The algorithm include user characteristics extraction algorithm based on noise reduction self-encoding encoder and curriculum video feature extraction algorithm, from
Two parts of score in predicting algorithm for adapting to matrix decomposition model, see Fig. 1.
2, user characteristics extraction algorithm and curriculum video feature extraction algorithm based on noise reduction self-encoding encoder
In order to solve the problems, such as that manual extraction feature is cumbersome, the present invention provides the features based on noise reduction self-encoding encoder to mention
Algorithm is taken, this feature extraction algorithm is divided into the extraction algorithm to user characteristics and the extraction algorithm to curriculum video feature again.When
When extracting user characteristics, Fig. 5 is seen, the input of noise reduction self-encoding encoder corresponds to the history scoring record and customer attribute information of user
Noise version can get the hidden layer character representation of user by training；When extracting curriculum video feature, Fig. 6 is seen, drop
The input for self-encoding encoder of making an uproar corresponds to the noise version of history the scoring record and curriculum video attribute information of curriculum video, finally
It can get the hidden layer character representation of curriculum video.
3, the score in predicting algorithm of adaptive matrix decomposition model
After the user characteristics and curriculum video feature for obtaining the extraction of noise reduction self-encoding encoder, the present invention provides adaptive squares
The score in predicting algorithm of battle array decomposition model, is shown in Fig. 4.User characteristics and the course view that the algorithm first extracts noise reduction self-encoding encoder
Frequency feature is embedded into the low-dimensional matrix of matrix decomposition, establishes new score in predicting formula, and expression is shown in formula (1)；
Loss function is established based on the formula, expression is shown in formula (2)；Then ladder is asked to parameters using the loss function
Degree, expression are shown in formula (3), formula (4), formula (5), formula (6), formula (7) and formula (8)；Using stochastic gradient
Descent algorithm undated parameter, expression are shown in formula (9), formula (10), formula (11), formula (12), formula (13) and public affairs
Formula (14)；The last prediction according to curriculum video is scored, and provides a curriculum video recommendation list for each user.
Detailed description of the invention
Fig. 1 is that the curriculum video in a kind of on-line study provided by the invention based on noise reduction self-encoding encoder mixed model is recommended
The flow chart of algorithm.
Fig. 2 is the flow chart of S1 in Fig. 1.
Fig. 3 is the flow chart of S2 in Fig. 1.
Fig. 4 is the flow chart of S3 in Fig. 1.
Fig. 5 is the neural network structure figure that noise reduction self-encoding encoder extracts user characteristics.
Fig. 6 is the neural network structure figure that noise reduction self-encoding encoder extracts curriculum video feature.
Symbol description used in the present invention：
N- number of users
M- curriculum video number
The dimension of the hidden feature vector of k-
γ-learning rate
λ-regular terms
τ-iteration threshold
The dimension of p- user property feature
The dimension of q- curriculum video attributive character
Specific embodiment
With reference to the accompanying drawing, specific embodiments of the present invention will be described in detail.
The invention discloses the curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study,
The algorithm includes user characteristics extraction algorithm and curriculum video feature extraction algorithm, adaptive matrix based on noise reduction self-encoding encoder
Two parts of score in predicting algorithm of decomposition model.
The curriculum video proposed algorithm overall flow figure of mixed model based on noise reduction self-encoding encoder is as shown in Figure 1.One, just
Beginningization
S1 in this part corresponding diagram 1, detail flowchart are shown in Fig. 2.
S1：Initialization
S1.1：Initiation parameter collection
1) assignment is carried out to number of users n and curriculum video number m according to real data collection.
2) user's prescribed coding feature vector dimension k, learning rate γ, regularization coefficient λ and the iteration threshold of the algorithm
τ。
3) random initializtion zoom factor αuAnd αi, offset vector ∈uAnd ∈i, by bias term buAnd biAll it is set to 0.
S1.2：Initialization scoring list and attribute list
Matrix R=(the r arranged with n row mui) indicate u-th of user to the score value of i-th curriculum video, wherein u ∈ 1,
2,3 ..., n }, i ∈ { 1,2,3 ..., m }.With the behavior master of matrix R, the scoring list of user is obtained Based on the column of matrix R, the scoring list of curriculum video is obtained User's
Attribute list withIndicate, the attribute list of curriculum video withIt indicates.
Wherein p indicates the dimension of user property, and q indicates the dimension of curriculum video attribute.
S1.3：Initialising subscriber-curriculum video scoring triple
With triple T=<U, I, R>Indicate user to the score information of curriculum video；U indicates user's collection, and | U |=n, I
Indicate curriculum video collection, and | I |=m, R indicate score information matrix.
Two, feature extraction
S2 in this part corresponding diagram 1, specific flow chart are shown in Fig. 3.
S2：Feature extraction algorithm based on noise reduction self-encoding encoder
S2.1：Extract user characteristics
1) all users are traversed and obtains the scoring list of each userAnd attribute listAs the input of noise reduction self-encoding encoder, wherein u ∈ { 1,2,3 ..., n }, n indicate number of users, m table
Show that curriculum video number, p indicate the dimension of user property feature.
2) random noise is added to the input data of noise reduction self-encoding encoder, obtains damage versionWithAnd input network
It is trained.The mode that noise is added can be one lesser random value of addition on the input data, and the random value is obeyed
Gaussian Profile obtains clean output by noise reduction self-encoding encoder.
3) function is utilizedIt is encoded, then decoding obtains
OutputCalculate square error： With under gradient
Drop method is trained noise reduction self-encoding encoder, minimizes above-mentioned square error.
4) after self-encoding encoder training, by coding function udae*(su,xu)=fu(Wu·su+Vu·xu+bu) obtain
The hidden layer character representation of user.
S2.2：Extract curriculum video feature
1) all curriculum videos are traversed and obtains the scoring list of every curriculum videoAnd category
Property listAs the input of self-encoding encoder, wherein i ∈ { 1,2,3 ..., m }, q indicate curriculum video
The dimension of attributive character.
2) random noise is added to the input data of self-encoding encoder, obtains damage versionWithAnd it inputs network and is instructed
Practice.
3) function is utilizedIt is encoded, then decoding is exported
Data：Calculate square error Declined with gradient
Algorithm is trained, and minimizes above-mentioned square error.
4) after self-encoding encoder training, by coding function idea*(si,yi)=fi(Wi·si+Vi·yi+bi) obtain
The hidden layer character representation of curriculum video.
Three, adaptive matrix decomposes
S3 in this part corresponding diagram 1, specific flow chart are shown in Fig. 4.
S3：The score in predicting algorithm that adaptive matrix decomposes
S3.1：Establish score in predicting formula：
In formula (1),Indicate user u to the score in predicting of curriculum video i；αuAnd αiRespectively indicate user and curriculum video
Hidden feature zoom factor；∈uAnd ∈iRespectively indicate the hidden characteristic offset vector of user and curriculum video；buAnd biTable respectively
Show the biasing coefficient of user and curriculum video；μ indicates the average value of all scorings.
S3.2：Establish the loss function L of score in predicting：
In formula (2), ruiIt is true scoring of the user u to curriculum video i；IuiIt is an indicator function, works as ruiWhen=0, i.e.,
User u does not score to curriculum video i, then Iui=0；Otherwise Iui=1；λ indicates regularization coefficient；freg=αu 2+αi 2+bu 2+bi 2
+‖∈u‖2+‖∈i‖2, the purpose that regular terms is added here is over-fitting in order to prevent.
S3.3：Initialize the number of iterations Z=1
S3.4：The gradient value of 6 parameters of the Z times iteration is obtained using following 6 formula：
S3.5：Parameter update is carried out using stochastic gradient descent algorithm (SGD)
For formula (9) in formula (14), γ indicates learning rate.
S3.6：Z+1 is assigned to Z, judges whether Z≤τ is true；If so, then go to step 4 execution；Otherwise, step is gone to
Rapid 7 execute；Wherein τ indicates the threshold value of the number of iterations.
S3.7：The highest preceding portion the K curriculum video of each user's score in predicting is obtained, the curriculum video for constituting user recommends column
Table.
1) u-th of user is calculated to the score in predicting of i-th curriculum video (i=1,2 ..., m) according to the formula of formula (1)The highest preceding portion the K curriculum video of score in predicting is selected in the curriculum video that user u did not score, and becomes the class of user u
Journey video recommendations list.
2) to each user u (u=1,2 ..., n), column are recommended using the curriculum video that method 1) obtains all users
Table.
Claims (3)
1. the curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study, it is characterised in that：Packet
Containing user characteristics extraction algorithm and curriculum video feature extraction algorithm, adaptive matrix decomposition model based on noise reduction self-encoding encoder
Score in predicting algorithm.
2. the curriculum video in a kind of on-line study according to claim 1 based on noise reduction self-encoding encoder mixed model is recommended
Algorithm, which is characterized in that the user characteristics extraction algorithm and curriculum video feature extraction based on noise reduction self-encoding encoder is calculated
Method is as follows：
1) specific step is as follows for the user characteristics extraction algorithm based on noise reduction self-encoding encoder：
It scores and records for the history of user uWith the attribute information of user u By adding
Enter the damage version that noise respectively obtains history scoring recordWith the damage version of attribute informationBe added noise method be
A lesser random value, and the random value Normal Distribution are added in data；Then the volume Jing Guo noise reduction self-encoding encoder
Code function fu() coding obtains the hidden layer character representation of user：
Then pass through decoding functions gu() obtains the output data of noise reduction self-encoding encoder：So
After calculate square error：Noise reduction self-encoding encoder is trained with gradient descent method, is made
Above-mentioned square error minimizes, the coding udae that hidden layer obtains at this time*(su,xu) it is namely based on the use of noise reduction self-encoding encoder
The output data of family feature extraction algorithm；In described above, m is the number of curriculum video；P is the dimension of user property；For
User u indicates that user u did not watch curriculum video j if the score value is 0 to the score value of curriculum video j；WuFor user
Scoring record suTo the weight of hidden layer；VuFor customer attribute information xuTo the weight of hidden layer；
2) specific step is as follows for the curriculum video feature extraction algorithm based on noise reduction self-encoding encoder：
It scores and records for the history of curriculum video iWith the attribute information of curriculum video iThe damage version of history scoring record is respectively obtained by the way that noise is addedWith the damage of attribute information
Bad versionThen the coding function f Jing Guo noise reduction self-encoding encoderi() coding obtains the hidden layer mark sheet of curriculum video
Show：Then pass through decoding functions gi() obtains noise reduction self-encoding encoder
Output data： Calculate square errorUse gradient
Descent algorithm is trained, and minimizes above-mentioned square error, the coding idea that hidden layer obtains at this time*(si,yi) it is exactly base
In the output data of the curriculum video feature extraction algorithm of noise reduction self-encoding encoder；In described above, n is the number of user；Q is class
The dimension of journey video attribute；It is user j to the score value of curriculum video i, if the score value is that 0 expression user j is not watched
Cross curriculum video i；WiThe weight of hidden layer is recorded for the scoring of curriculum video history；ViFor curriculum video attribute information yiTo hidden
Weight containing layer.
3. the curriculum video in a kind of on-line study according to claim 1 based on noise reduction self-encoding encoder mixed model is recommended
Algorithm, which is characterized in that the score in predicting algorithm of the adaptive matrix decomposition model, specific step is as follows：
Step 1：Establish score in predicting formula：
In formula,It is user u to the score in predicting of curriculum video i；αuAnd αiThe hidden feature of respectively user u and curriculum video i
Zoom factor；∈uAnd ∈iThe offset vector of the hidden feature of respectively user u and curriculum video i；buAnd biRespectively user u and
The biasing coefficient of curriculum video i；μ indicates the average value of all scorings；
Step 2：Construct the loss function L of score in predicting:
In formula, ruiIt is true scoring of the user u to curriculum video i；IuiIt is an indicator function, works as ruiWhen=0, i.e. u couples of user
When curriculum video i does not score, then Iui=0, otherwise Iui=1；λ is a constant；Regular terms freg=αu 2+αi 2+bu 2+bi 2+‖
∈u‖2+‖∈i‖2, the purpose that regular terms is added is over-fitting in order to prevent；
Step 3：Initialize the number of iterations：Z=1
Step 4：It is utilized respectively the gradient value that following 6 formula obtain 6 parameters of the Z times iteration：
Step 5：Using following 6 formula, parameter update is carried out using stochastic gradient descent method, wherein γ indicates learning rate：
Step 6：The value of Z is added 1, judges whether Z≤τ is true；If so, then go to step 4 execution；Otherwise step 7 is gone to hold
Row；Wherein τ indicates the threshold value of the number of iterations；
Step 7：The highest preceding portion the K curriculum video of each user's score in predicting is obtained, the curriculum video recommendation list of user is constituted；
Step 7.1：It is pre- to the scoring of i-th curriculum video (i=1,2 ..., m) that u-th of user is calculated according to the formula of step 1
It surveysThe highest preceding portion the K curriculum video of score in predicting is selected in the curriculum video that user u did not score, and becomes user u's
Curriculum video recommendation list；
Step 7.2：To each user u (u=1,2 ..., n), pushed away using the curriculum video that the method for step 7.1 obtains the user
Recommend list.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810575724.7A CN108874960A (en) | 2018-06-06 | 2018-06-06 | Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810575724.7A CN108874960A (en) | 2018-06-06 | 2018-06-06 | Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108874960A true CN108874960A (en) | 2018-11-23 |
Family
ID=64337222
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810575724.7A Pending CN108874960A (en) | 2018-06-06 | 2018-06-06 | Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108874960A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321483A (en) * | 2019-06-18 | 2019-10-11 | 深圳职业技术学院 | A kind of online course content of platform recommended method, device, system and storage medium based on user's sequence sexual behaviour |
| CN110717103A (en) * | 2019-10-09 | 2020-01-21 | 东北大学 | Improved collaborative filtering method based on stack noise reduction encoder |
| CN110990715A (en) * | 2019-11-14 | 2020-04-10 | 北京交通大学 | Multi-source user attribute inference method based on layer self-encoder |
| CN111310029A (en) * | 2020-01-20 | 2020-06-19 | 哈尔滨理工大学 | Mixed recommendation method based on user commodity portrait and potential factor feature extraction |
| CN111311324A (en) * | 2020-02-18 | 2020-06-19 | 电子科技大学 | User-commodity preference prediction system and method based on stable neural collaborative filtering |
| CN111784619A (en) * | 2020-07-03 | 2020-10-16 | 电子科技大学 | Fusion method of infrared and visible light images |
| CN112528074A (en) * | 2020-12-14 | 2021-03-19 | 扬州大学 | Film recommendation method of knowledge graph combined with self-coding machine |
| CN112749805A (en) * | 2021-01-15 | 2021-05-04 | 浙江工业大学 | Online course recommendation method based on multiple entity relationships |
| CN113205388A (en) * | 2021-05-12 | 2021-08-03 | 中国地质大学（武汉） | Collaborative filtering recommendation optimization algorithm and device based on matrix decomposition |
| CN117112832A (en) * | 2023-09-12 | 2023-11-24 | 社培科技(广东)有限公司 | Teaching method and system based on combination of online monitoring and offline guidance |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107273438A (en) * | 2017-05-24 | 2017-10-20 | 深圳大学 | A kind of recommendation method, device, equipment and storage medium |
| CN107943897A (en) * | 2017-11-17 | 2018-04-20 | 东北师范大学 | A kind of user recommends method |
-
2018
- 2018-06-06 CN CN201810575724.7A patent/CN108874960A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107273438A (en) * | 2017-05-24 | 2017-10-20 | 深圳大学 | A kind of recommendation method, device, equipment and storage medium |
| CN107943897A (en) * | 2017-11-17 | 2018-04-20 | 东北师范大学 | A kind of user recommends method |
Non-Patent Citations (3)
| Title |
|---|
| F.ZHUANG: "Representation learning via Dual-Autoencoder for recommendation", 《NEURAL NETWORKS》 * |
| XIN DONG: "A Hybrid Collaborative Filtering Model with Deep Structure for Recommender Systems", 《PROCEEDINGS OF THE THIRTY-FIRST AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE(AAAI-17)》 * |
| Y.KOREN: "Matrix factorization techniques for recommender systems", 《COMPUTER》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321483A (en) * | 2019-06-18 | 2019-10-11 | 深圳职业技术学院 | A kind of online course content of platform recommended method, device, system and storage medium based on user's sequence sexual behaviour |
| CN110717103B (en) * | 2019-10-09 | 2021-08-31 | 东北大学 | Improved collaborative filtering method based on stack noise reduction encoder |
| CN110717103A (en) * | 2019-10-09 | 2020-01-21 | 东北大学 | Improved collaborative filtering method based on stack noise reduction encoder |
| CN110990715A (en) * | 2019-11-14 | 2020-04-10 | 北京交通大学 | Multi-source user attribute inference method based on layer self-encoder |
| CN110990715B (en) * | 2019-11-14 | 2023-06-06 | 北京交通大学 | Multi-source user attribute deducing method based on layer self-encoder |
| CN111310029A (en) * | 2020-01-20 | 2020-06-19 | 哈尔滨理工大学 | Mixed recommendation method based on user commodity portrait and potential factor feature extraction |
| CN111310029B (en) * | 2020-01-20 | 2022-11-01 | 哈尔滨理工大学 | Mixed recommendation method based on user commodity portrait and potential factor feature extraction |
| CN111311324A (en) * | 2020-02-18 | 2020-06-19 | 电子科技大学 | User-commodity preference prediction system and method based on stable neural collaborative filtering |
| CN111311324B (en) * | 2020-02-18 | 2022-05-20 | 电子科技大学 | User-commodity preference prediction system and method based on stable neural collaborative filtering |
| CN111784619B (en) * | 2020-07-03 | 2023-04-28 | 电子科技大学 | Fusion method of infrared and visible light images |
| CN111784619A (en) * | 2020-07-03 | 2020-10-16 | 电子科技大学 | Fusion method of infrared and visible light images |
| CN112528074A (en) * | 2020-12-14 | 2021-03-19 | 扬州大学 | Film recommendation method of knowledge graph combined with self-coding machine |
| CN112528074B (en) * | 2020-12-14 | 2023-06-16 | 扬州大学 | Movie recommendation method combining knowledge graph with self-encoder |
| CN112749805A (en) * | 2021-01-15 | 2021-05-04 | 浙江工业大学 | Online course recommendation method based on multiple entity relationships |
| CN112749805B (en) * | 2021-01-15 | 2024-02-13 | 浙江工业大学 | Online course recommendation method based on various entity relations |
| CN113205388A (en) * | 2021-05-12 | 2021-08-03 | 中国地质大学（武汉） | Collaborative filtering recommendation optimization algorithm and device based on matrix decomposition |
| CN117112832A (en) * | 2023-09-12 | 2023-11-24 | 社培科技(广东)有限公司 | Teaching method and system based on combination of online monitoring and offline guidance |
| CN117112832B (en) * | 2023-09-12 | 2024-03-19 | 社培科技(广东)有限公司 | Teaching method and system based on combination of online monitoring and offline guidance |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108874960A (en) | Curriculum video proposed algorithm based on noise reduction self-encoding encoder mixed model in a kind of on-line study | |
| CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
| Neidhardt et al. | Eliciting the users' unknown preferences | |
| CN108363790A (en) | For the method, apparatus, equipment and storage medium to being assessed | |
| US9223775B2 (en) | User question processing method and system | |
| CN109816265B (en) | Knowledge characteristic mastery degree evaluation method, question recommendation method and electronic equipment | |
| CN108090229A (en) | A kind of method and apparatus that rating matrix is determined based on convolutional neural networks | |
| CN105335157B (en) | A kind of demand classes sort method for integrating subjective and objective evaluation and system | |
| CN109299380A (en) | Exercise personalized recommendation method in online education platform based on multidimensional characteristic | |
| CN106168980A (en) | Multimedia resource recommends sort method and device | |
| CN109635083A (en) | It is a kind of for search for TED speech in topic formula inquiry document retrieval method | |
| CN110110899B (en) | Knowledge mastery degree prediction method, adaptive learning method and electronic equipment | |
| WO2020135642A1 (en) | Model training method and apparatus employing generative adversarial network | |
| CN107609651A (en) | A kind of design item appraisal procedure based on learner model | |
| CN104008204A (en) | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof | |
| CN104572915B (en) | One kind is based on the enhanced customer incident relatedness computation method of content environment | |
| CN110399547A (en) | For updating the method, apparatus, equipment and storage medium of model parameter | |
| CN116308685A (en) | Product recommendation method and system based on aspect emotion prediction and collaborative filtering | |
| CN116541607A (en) | Intelligent recommendation method based on commodity retrieval data analysis | |
| TWI403911B (en) | Chinese dictionary constructing apparatus and methods, and storage media | |
| CN111639485A (en) | Course recommendation method based on text similarity and related equipment | |
| CN108960954B (en) | Content recommendation method and system based on user group behavior feedback | |
| CN108363709A (en) | A kind of chart commending system and method using principal component based on user | |
| CN115687657A (en) | Knowledge graph-assisted test question recommendation method | |
| CN110569444A (en) | Improved slope one algorithm fusing similarity and time attenuation |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181123 |