CN111401525A - Adaptive learning system and method based on deep learning - Google Patents
Adaptive learning system and method based on deep learning Download PDFInfo
- Publication number
- CN111401525A CN111401525A CN202010199603.4A CN202010199603A CN111401525A CN 111401525 A CN111401525 A CN 111401525A CN 202010199603 A CN202010199603 A CN 202010199603A CN 111401525 A CN111401525 A CN 111401525A
- Authority
- CN
- China
- Prior art keywords
- learning
- resource
- learner
- resources
- function
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/40—Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses an adaptive learning system and method based on deep learning, which comprises the following steps: aiming at the increase of the number of line learning platforms, the learning resource forms are more and more diversified, an adaptive learning system and method based on deep learning are designed and developed, output visual description under a learner-resource bipartite graph association model is established by designing a deep neural network input optimization strategy based on a mutual information feature selection Model (MIFS), and then a resource recommendation model is obtained by utilizing deep neural network training to realize personalized learning resource recommendation. The characteristics of strong data adaptability, good processing performance and the like of the deep learning algorithm are utilized for analytical study on the aspects of learner preference, learner characteristic types and the like, the accuracy of predicting the learner preference is high, and the experience and effect of online learning of the learner are improved.
Description
Technical Field
The invention belongs to the technical field of online learning, and particularly relates to an adaptive learning system and method based on deep learning.
Background
With the advent of the big data era, the education big data generated by combining the education field can find the teaching rule from massive education data, so as to optimize the teaching mode and develop more efficient teaching work. The essence of the education big data is the big data in the education field, the data sources of the education big data not only comprise the learning data of learners, but also comprise all behavior data of people in daily education activities, and the education big data has the characteristics of multiple subjects, multiple dimensions, multiple forms and the like. Based on the personalized learning idea, the constructed adaptive online learning system can collect learning behavior data of students, analyze and construct student abilities and recommend proper exclusive learning resources. The adaptive learning has better effect and advantage in the aspect of guiding the adaptive learning according to individual differences of students. The patent provides a new adaptive learning resource recommendation method based on deep learning, an adaptive learning system is designed and developed, and the method has good accuracy, stability and effectiveness when learning resources are recommended.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides an adaptive learning system and method based on deep learning.
The invention provides an adaptive learning system based on deep learning, which is characterized by comprising a front-end module, a rear-end module and a storage module. The front-end module is an operation interface of users such as students, teachers, administrators and the like, is responsible for processing input and output of user information and displaying teaching and test question contents, and is connected with the rear-end module through a network. The back-end module comprises a basic service part and a business service part and is responsible for processing, modifying and inquiring user, teaching, test question data and the like. The back end module is connected with the storage module to store the user and the teaching data.
The adaptive learning method based on deep learning mainly comprises the following steps:
step 1, a user logs in and learns through a front-end module, and a rear-end module analyzes and processes a data set of the user in a storage module to obtain data of students and learning resources;
step 2, searching the association between the learner and the resource in a plurality of characteristics of the learning resource data and establishing a characteristic selection model so as to complete the input process of the recommendation method;
step 3, establishing a learner-resource bipartite graph association model to obtain the output of a recommendation method;
step 4, judging whether the learner learns a certain learning resource and the attention degree of the learning resource based on the deep neural network model;
and 5, determining a final learning resource recommendation list, and displaying the recommended learning resources to the user through the front-end module.
Preferably, the method for establishing the feature selection model in step 2 is as follows:
a Mutual Information Feature Selection (MIFS) -based method is selected, an information metric evaluation function is crucial to the MIFS-based feature selection method, and the information metric evaluation function can be expressed as follows:
where S is the selected feature, S is the single feature, f is the candidate feature, C is the class, β is the adjustment coefficient function g (C, f) is the amount of information between C, f, g (S, f) is the amount of information between S, f.
The selected features represent some features that determine to influence the learner in selecting resources, including knowledge content, learning duration, resource presentation form of the resources.
Preferably, the method for establishing the bipartite graph association model in step 3 is as follows:
define the learner set as L ═ l1,l2,l3,…lm}; the resource set is as follows: r ═ R1,r2,r3,…rm}. Obtaining a binary relation matrix X composed of a learner set and a resource setm×n:
Wherein the row vector represents the learner, the column vector represents the learner having learned the resource, if Xm×nIf 1, it means that the learner has learned the resource, and if Xm×nIf 0, then there is no learning, and this is simpleThe normalization process of (a) does not objectively reflect the degree of interest of the learner in the learning resource, and thus the frequency of learning a certain resource cannot be ignored. The learner's learning frequency for learning resources may reflect different degrees of preference, and the average frequency of resource learning may be defined as:
whereinRepresents liAverage number of times of using learning resources, n (l)i) Represents liThe amount of the resource that has been learned,represents liTotal number of learning resources. Because the frequency is greatly related to whether the learner is interested in learning the resource, the learner has more times to learn a certain learning resource, and can consider that the learner has higher attention to the resource or the resource.
At the same time, the concrete learning frequency is predicted, compared with actual learning frequency, and its error is judged, i.e. regression analysis of recommendation model, and further in order to determine that a certain learning is recommended to learner or not, the average using frequency of learning resource is usedAs a critical value, the classification analysis of the recommendation model is performed, and the determination conditions are as follows:
preferably, the method for establishing the deep neural network model in step 4 is as follows:
the deep neural network model designed aiming at the learning resource recommendation problem consists of an input layer, a plurality of hidden layers and an output layer, and is specifically designed as follows according to the construction of the two models.
Hidden layer design: the hidden layer adopts a Sigmoid activation function, wherein x is a value of an abscissa, f (x) is a value of an ordinate, and a function expression of the function is as follows:
designing a cost function: the method uses a standard quadratic cost function, wherein C in the function represents cost, y represents actual value, and a represents output value, and the function formula is as follows:
designing an output layer: the learning resource recommendation problem solved by the method is finally converted into a recommendation problem or an unrerecommendation problem, so that the output layer adopts a classical logistic regression model, and the Sigmoid function is the probability function of the output layer.
Compared with the prior art, the invention has the beneficial technical effects that: the system can build a user model according to the user behavior, so that learning resources which are possibly liked by the user can be recommended to the user in a targeted manner, the learning enthusiasm of the user is stimulated, and the learning enthusiasm of the user is improved.
The attached drawings show
FIG. 1 is a flow chart of the steps of the present invention.
Fig. 2 is a structural frame diagram of the present invention.
Detailed Description
The invention provides an adaptive learning system based on deep learning, which is characterized by comprising a front-end module, a rear-end module and a storage module. The front-end module is an operation interface of users such as students, teachers, administrators and the like, is responsible for processing input and output of user information and displaying teaching and test question contents, and is connected with the rear-end module through a network. The back-end module comprises a basic service part and a business service part and is responsible for processing, modifying and inquiring user, teaching, test question data and the like. The back end module is connected with the storage module to store the user and the teaching data.
The adaptive learning method based on deep learning mainly comprises the following steps:
step 1, a user logs in and learns through a front-end module, and a rear-end module analyzes and processes a data set of the user in a storage module to obtain data of students and learning resources;
step 2, searching the association between the learner and the resource in a plurality of characteristics of the learning resource data and establishing a characteristic selection model so as to complete the input process of the recommendation method;
specifically, a Mutual Information Feature Selection (MIFS) -based method is selected, and in the MIFS-based feature selection method, an information metric evaluation function is crucial to the method, and the information metric evaluation function can be expressed as:
where S is the selected feature, S is the single feature, f is the candidate feature, C is the class, β is the adjustment coefficient function g (C, f) is the amount of information between C, f, g (S, f) is the amount of information between S, f.
The selected features represent some features that determine to influence the learner in selecting resources, including knowledge content, learning duration, resource presentation form of the resources.
Step 3, establishing a learner-resource bipartite graph association model to obtain the output of a recommendation method;
specifically, a model training dataset is generated. And carrying out unique hot coding on the data set to form m-binary-characteristic sparse data, operating interaction records of students and learning resources item by item, inputting the unique hot coding of a current item in each time step, outputting a full connection layer of each learning resource neuron in a catalogue, and taking the learning resource with the most activated neuron as recommendation.
Step 3, constructing a prediction model based on a project reaction theory, and judging whether the step of using the knowledge points on the recommended test questions by the students can be right or not;
specifically, the learner set is defined as L ═ l1,l2,l3,…lm}; the resource set is as follows: r ═ R1,r2,r3,…rm}. Obtaining a binary relation matrix X composed of a learner set and a resource setm×n:
Wherein the row vector represents the learner, the column vector represents the learner having learned the resource, if Xm×nIf 1, it means that the learner has learned the resource, and if Xm×nIf the value is 0, the learner has not learned, and thus the simple normalization process cannot objectively reflect the attention of the learner on the learning resource, so that the frequency of learning a certain resource cannot be ignored. The learner's learning frequency for learning resources may reflect different degrees of preference, and the average frequency of resource learning may be defined as:
whereinRepresents liAverage number of times of using learning resources, n (l)i) Represents liThe amount of the resource that has been learned,represents liTotal number of learning resources. Because the frequency is greatly related to whether the learner is interested in learning the resource, the learner can be considered to learn the resource or not because the learning frequency of a learning resource is more, and the learner can be considered to learn the resource or notThe resources have higher attention.
At the same time, the concrete learning frequency is predicted, compared with actual learning frequency, and its error is judged, i.e. regression analysis of recommendation model, and further in order to determine that a certain learning is recommended to learner or not, the average using frequency of learning resource is usedAs a critical value, the classification analysis of the recommendation model is performed, and the determination conditions are as follows:
step 4, judging whether the learner learns a certain learning resource and the attention degree of the learning resource based on the deep neural network model;
specifically, the deep neural network model designed for the learning resource recommendation problem is composed of an input layer, a plurality of hidden layers and an output layer, and is specifically designed as follows according to the construction of the two models.
Hidden layer design: the hidden layer adopts a Sigmoid activation function, wherein x is a value of an abscissa, f (x) is a value of an ordinate, and a function expression of the function is as follows:
designing a cost function: the method uses a standard quadratic cost function, wherein C in the function represents cost, y represents actual value, and a represents output value, and the function formula is as follows:
designing an output layer: the learning resource recommendation problem solved by the method is finally converted into a recommendation problem or an unrerecommendation problem, so that the output layer adopts a classical logistic regression model, and the Sigmoid function is the probability function of the output layer.
And 5, determining a final learning resource recommendation list, and displaying the recommended learning resources to the user through the front-end module.
The present invention has been described in detail, and the principle and embodiments of the present invention are explained by using specific examples, which are only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present disclosure should not be construed as a limitation to the present invention.
Claims (5)
1. The invention provides an adaptive learning system based on deep learning, which is characterized by comprising a front-end module, a rear-end module and a storage module, wherein the front-end module is used for storing a learning result; the front-end module is an operation interface of a student user, a teacher user and an administrator user, is responsible for processing input and output of user information and displaying teaching and test question contents, and is connected with the rear-end module through a network; the back-end module comprises a basic service part and a business service part and is responsible for processing, modifying and inquiring user and teaching and test question data; the back end module is connected with the storage module to store the user and the teaching data.
2. The adaptive learning method based on deep learning mainly comprises the following steps:
step 1, a user logs in and learns through a front-end module, and a rear-end module analyzes and processes a data set of the user in a storage module to obtain data of students and learning resources thereof;
step 2, searching the association between the learner and the resource in a plurality of characteristics of the learning resource data and establishing a characteristic selection model so as to complete the input process of the recommendation method;
step 3, establishing a learner-resource bipartite graph association model to obtain the output of a recommendation method;
step 4, judging whether the learner learns a certain learning resource and the attention degree of the learning resource based on the deep neural network model;
and 5, determining a final learning resource recommendation list, and displaying the recommended learning resources to the user through the front-end module.
3. The adaptive learning method based on deep learning of claim 2,
the method for establishing the feature selection model in the step 2 comprises the following steps: selecting a Mutual Information Feature Selection (MIFS) based method to obtain an information measurement evaluation function, wherein the information measurement evaluation function can be expressed as:
where S is the selected feature, S is the individual feature, f is the candidate feature, C is the class, β is the adjustment coefficient function g (C, f) is the amount of information between C, f, g (S, f) is the amount of information between S, f, the selected feature represents some features determined to affect the learner' S selection of resources, including knowledge content, learning duration, resource presentation form of the resources.
4. The adaptive learning method based on deep learning of claim 2, wherein:
the association model of the bipartite graph in step 3 is established by defining the learner set as L ═ l1,l2,l3,…lm}; the resource set is as follows: r ═ R1,r2,r3,…rmObtaining a binary relation matrix X consisting of a learner set and a resource setm×n:
Wherein the row vector represents the learner, the column vector represents the learner having learned the resource, if Xm×nIf 1, it means that the learner has learned the resource, and if Xm×nWhen the value is 0, the learning is not performed,the simple normalization processing cannot objectively reflect the attention degree of the learner on the learning resources, so that the frequency of learning a certain resource cannot be ignored; the learner's learning frequency for learning resources may reflect different degrees of preference, and the average frequency of resource learning may be defined as:
whereinRepresents liAverage number of times of using learning resources, n (l)i) Represents liThe amount of the resource that has been learned,represents liThe total times of learning resources, because the frequency is greatly related to whether the learner is interested in learning the resources, the learner has more times of learning a certain learning resource, and can think that the learner has higher attention to the resource or the resources;
at the same time, the concrete learning frequency is predicted, compared with actual learning frequency, and its error is judged, i.e. regression analysis of recommendation model, and further in order to determine that a certain learning is recommended to learner or not, the average using frequency of learning resource is usedAs a critical value, the classification analysis of the recommendation model is performed, and the determination conditions are as follows:
5. the adaptive learning method based on deep learning of claim 2, wherein:
the method for establishing the deep neural network model in the step 4 comprises the following steps:
the deep neural network model designed aiming at the learning resource recommendation problem consists of an input layer, a plurality of hidden layers and an output layer, and is specifically designed as follows according to the construction of the two models:
hidden layer design: the hidden layer adopts a Sigmoid activation function, wherein x is a value of an abscissa, f (x) is a value of an ordinate, and a function expression of the function is as follows:
designing a cost function: the method uses a standard quadratic cost function, wherein C in the function represents cost, y represents actual value, and a represents output value, and the function formula is as follows:
designing an output layer: the learning resource recommendation problem solved by the method is finally converted into a recommendation problem or an unrerecommendation problem, so that the output layer adopts a classical logistic regression model, and the Sigmoid function is the probability function of the output layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010199603.4A CN111401525A (en) | 2020-03-20 | 2020-03-20 | Adaptive learning system and method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010199603.4A CN111401525A (en) | 2020-03-20 | 2020-03-20 | Adaptive learning system and method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111401525A true CN111401525A (en) | 2020-07-10 |
Family
ID=71436340
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010199603.4A Pending CN111401525A (en) | 2020-03-20 | 2020-03-20 | Adaptive learning system and method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111401525A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112099866A (en) * | 2020-07-30 | 2020-12-18 | 福建天泉教育科技有限公司 | Implementation method and terminal for learning active plug-in |
CN113378076A (en) * | 2021-06-29 | 2021-09-10 | 哈尔滨工业大学 | Online education-oriented learner collaborative learning social relationship construction method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106528693A (en) * | 2016-10-25 | 2017-03-22 | 广东科海信息科技股份有限公司 | Individualized learning-oriented educational resource recommendation method and system |
CN108615423A (en) * | 2018-06-21 | 2018-10-02 | 中山大学新华学院 | Instructional management system (IMS) on a kind of line based on deep learning |
CN108765228A (en) * | 2018-06-20 | 2018-11-06 | 陆柒(北京)科技有限公司 | A kind of adaptive private teaching learning method of computer |
-
2020
- 2020-03-20 CN CN202010199603.4A patent/CN111401525A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106528693A (en) * | 2016-10-25 | 2017-03-22 | 广东科海信息科技股份有限公司 | Individualized learning-oriented educational resource recommendation method and system |
CN108765228A (en) * | 2018-06-20 | 2018-11-06 | 陆柒(北京)科技有限公司 | A kind of adaptive private teaching learning method of computer |
CN108615423A (en) * | 2018-06-21 | 2018-10-02 | 中山大学新华学院 | Instructional management system (IMS) on a kind of line based on deep learning |
Non-Patent Citations (1)
Title |
---|
李浩君等: "深度学习视角下的个性化学习资源推荐方法", 《现代远程教育研究》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112099866A (en) * | 2020-07-30 | 2020-12-18 | 福建天泉教育科技有限公司 | Implementation method and terminal for learning active plug-in |
CN113378076A (en) * | 2021-06-29 | 2021-09-10 | 哈尔滨工业大学 | Online education-oriented learner collaborative learning social relationship construction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108021616B (en) | Community question-answer expert recommendation method based on recurrent neural network | |
Yukselturk et al. | Predicting dropout student: an application of data mining methods in an online education program | |
Peña-Ayala | Educational data mining: A survey and a data mining-based analysis of recent works | |
Özpolat et al. | Automatic detection of learning styles for an e-learning system | |
CN111241311A (en) | Media information recommendation method and device, electronic equipment and storage medium | |
CN111242310B (en) | Feature validity evaluation method and device, electronic equipment and storage medium | |
CN111831905B (en) | Recommendation method and device based on team scientific research influence and sustainability modeling | |
Saito et al. | Learning path recommender system based on recurrent neural network | |
Wang et al. | Attention-based CNN for personalized course recommendations for MOOC learners | |
CN111368082A (en) | Emotion analysis method for domain adaptive word embedding based on hierarchical network | |
CN111400592A (en) | Personalized course recommendation method and system based on eye movement technology and deep learning | |
Zhong et al. | Design of a personalized recommendation system for learning resources based on collaborative filtering | |
CN110751289A (en) | Online learning behavior analysis method based on Bagging-BP algorithm | |
CN115146161A (en) | Personalized learning resource recommendation method and system based on content recommendation | |
CN111401525A (en) | Adaptive learning system and method based on deep learning | |
Hamim et al. | Student profile modeling using boosting algorithms | |
CN112529750A (en) | Learning event recommendation method and system based on graph neural network model | |
Zhu | Network course recommendation system based on double-layer attention mechanism | |
He et al. | Exercise recommendation method based on knowledge tracing and concept prerequisite relations | |
CN114239730A (en) | Cross-modal retrieval method based on neighbor sorting relation | |
KR20220053982A (en) | Method for recommanding educational institute based on artificial intelligence | |
CN113742586B (en) | Learning resource recommendation method and system based on knowledge graph embedding | |
Zeng | Application of conditional random field model based on machine learning in online and offline integrated educational resource recommendation | |
Scalise et al. | A Taxonomy of Critical Dimensions at the Intersection of Learning Analytics and Educational Measurement | |
CN112699933A (en) | Automatic identification method and system for processing capacity of user teaching material |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200710 |
|
RJ01 | Rejection of invention patent application after publication |