CN115272015A - Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism - Google Patents
Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism Download PDFInfo
- Publication number
- CN115272015A CN115272015A CN202210675138.6A CN202210675138A CN115272015A CN 115272015 A CN115272015 A CN 115272015A CN 202210675138 A CN202210675138 A CN 202210675138A CN 115272015 A CN115272015 A CN 115272015A
- Authority
- CN
- China
- Prior art keywords
- course
- user
- courses
- knowledge
- graph
- 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
- 238000000034 method Methods 0.000 title claims abstract description 84
- 230000007246 mechanism Effects 0.000 title claims abstract description 30
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 16
- 230000008569 process Effects 0.000 claims abstract description 33
- 238000000605 extraction Methods 0.000 claims abstract description 31
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000011156 evaluation Methods 0.000 claims description 16
- 230000003993 interaction Effects 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 5
- 230000004931 aggregating effect Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000002547 anomalous effect Effects 0.000 claims 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 230000000052 comparative effect Effects 0.000 description 6
- 238000012552 review Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000002679 ablation Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000007596 consolidation process Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000012733 comparative method Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000003116 impacting effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000035755 proliferation Effects 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Strategic Management (AREA)
- Computational Linguistics (AREA)
- Educational Technology (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a course recommendation method and system based on an abnormal picture and cooperative attenuation attention mechanism, wherein the method comprises the following steps: acquiring historical course record information and candidate recommended course information of a target user on a data preprocessing layer, constructing a heterogeneous graph of courses, and further constructing input characteristics of an input model; in the knowledge extraction layer, according to a heterogeneous graph between a user and a course, the high-order features of the user and the course are obtained by adopting the knowledge extraction layer based on graph convolution, and the input features are enriched; in a knowledge evolution layer, a cooperative attenuation attention mechanism is adopted to model the learning process of a target user, so that input features can be combined with the learning process of the target user; in the prediction layer, the output result of the knowledge evolution layer is input into the full-connection network to obtain the predicted course registration rate; and pushing the target recommended course to the target user according to the predicted course registration rate. The method has high recommendation accuracy and comprehensiveness, and can be widely applied to the technical field of artificial intelligence.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a course recommendation method and system based on an abnormal picture and cooperative attenuation attention mechanism.
Background
Today, with the popularity of internet technology, online platforms such as Coursera, edX, and Udacity are well known to most users, and many users have the opportunity to learn courses for well-known universities through these platforms at the lowest cost. However, the proliferation of online curriculum quantities places a higher demand on more accurate and targeted recommendations for users.
Unlike other recommendation tasks, course recommendation faces three important challenges, as it is highly relevant to learning input and learning processes. (1) According to the input hypothesis in the field of education, the input for the learner should be "i +1", where "i" refers to the knowledge of the current stage and "i +1" refers to the learning of the next stage. In other words, learners should be exposed to knowledge just beyond their current abilities to understand much of their content, but still be challenged to progress. Given that most courses require some prior knowledge, users also tend to learn courses that they already possess prior knowledge to ensure understandable input. For example, if the user has knowledge of two courses, namely "linear algebra" and "probability theory", the course of "machine learning" will be more suitable as the recommended course. Meanwhile, the course contains a plurality of knowledge points. Thus, course recommendations should take into account the precedence and knowledge points of the course. (2) According to the forgetting curve, forgetting starts immediately after learning and the degree of memory decreases exponentially. While most users register courses during the learning process showing an increasing trend in difficulty. This indicates that lessons learned in the far past should not have as much information as lessons learned recently when the present invention recommends their lessons. (3) The interpretable recommendation can guide students to do targeted review and consolidation, and the enthusiasm of the users for the recommended course approval and subsequent learning is improved. (4) Popular recommendation systems are evaluated by click rate, NDCG (normalized breaking cumulative gain), and the like, but they all generally have a problem: there is no way to fully measure the relevance of the recommended course set as a whole to the target course.
The traditional course recommendation method does not focus on modeling the learning process of students, does not capture the sequence relation and knowledge point information of courses, and cannot comprehensively evaluate the overall recommendation performance of a course set.
Disclosure of Invention
In view of this, the embodiment of the present invention provides a course recommendation method and system based on an abnormal graph and a cooperative attentive power reduction mechanism, which have high recommendation accuracy and are comprehensive.
One aspect of the embodiments of the present invention provides a course recommendation method based on an abnormal picture and a cooperative attentive power attenuation mechanism, including:
acquiring historical course record information and candidate recommended course information of a target user on a data preprocessing layer, constructing a heterogeneous graph of courses, and further constructing input characteristics of an input model;
in a knowledge extraction layer, according to a heterogeneous graph between the user and the course, acquiring high-order characteristics of the user and the course by adopting a graph convolution-based knowledge extraction layer, and enriching the input characteristics;
in a knowledge advancement layer, modeling a learning process of the target user using a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
inputting the output result of the knowledge evolution layer into a full-connection network at a prediction layer to obtain a predicted course registration rate;
and pushing target recommended courses to the target user according to the predicted course registration rate.
Optionally, the heterogeneous graph includes interaction data of the user, course first-next relations and course knowledge points;
the method for constructing the heterogeneous course graph and further constructing the input characteristics of the input model comprises the following steps of:
constructing a bipartite graph of the courses corresponding to the user according to the registration history of the user in the training set;
constructing a sequence relationship chart of the courses;
determining knowledge points contained in the course according to the catalog and outline of the course;
connecting knowledge points contained in the courses with all the courses, and merging the bipartite graph and the sequence relation graph to obtain a heterogeneous graph between the user and the courses;
and obtaining the high-order characteristics of each node in the abnormal graph by using a graph convolution network according to the abnormal graph.
Optionally, the obtaining, according to the heterogeneous graph between the user and the course, the high-order features of the user and the course by using a knowledge extraction layer based on graph convolution, and enriching the input features includes:
acquiring initial characteristics of a user and a course through an embedded layer;
for a knowledge point, obtaining semantic features of the knowledge point through a pre-trained w2v model;
each time the node passes through a knowledge extraction layer, the characteristics of the neighbor nodes are summed and normalized by aggregating the node characteristics of the neighbors corresponding to the users and the courses on the graph, and the high-order characteristics of each node are obtained; wherein the user aggregated neighbors are courses and the course aggregated neighbors include: semantic information of knowledge points contained in the course, the course which is followed by the current course and the user;
and averaging the obtained characteristics of each user node and the course node in different layers through a plurality of knowledge extraction layers, and adding the characteristics of the course and the class characteristics corresponding to the course to obtain the final characteristics of the node.
Optionally, the modeling the learning process of the target user with a cooperative attentive power mechanism such that the input features can be combined with the learning process of the target user includes:
for each course, calculating the monotonous attention weight of each course; the monotone is used for representing the course query characteristic, and the monotone is multiplied by the key characteristic points of the course and the courses learned before the course to calculate the attention score;
calculating the attention weight after cooperative attenuation according to the exponential attenuation rate and the monotonous attention weight;
according to the attention weight after cooperative attenuation, carrying out weighted summation on the attention weight and each characteristic to obtain the characteristic of the current course;
and acquiring the final characteristics of each course as the input characteristics.
Optionally, the method further comprises: and determining an evaluation index of the course recommendation result by combining the knowledge points.
Optionally, the determining, in combination with the knowledge point, an evaluation index of the course recommendation result includes:
calculating the similarity of courses between any two courses by calculating the same number of knowledge points contained in the courses according to the covering condition of the knowledge points of the courses;
for each user, calculating a course set which is most likely to be registered by the user, then calculating the mean value of the course similarity of each course actually selected by the user and a recommended target course according to the course similarity, obtaining the overall similarity of the set of recommended courses and the target course, and further obtaining the evaluation index of the course recommendation result.
Another aspect of the embodiments of the present invention further provides a course recommendation system based on an abnormal picture and a cooperative attentive power attenuation mechanism, including:
the data preprocessing layer is used for acquiring historical course record information and candidate recommended course information of a target user, constructing a heterogeneous graph of courses and further constructing input characteristics of an input model;
the knowledge extraction layer is used for acquiring high-order characteristics of the user and the course by adopting the knowledge extraction layer based on graph convolution according to the heterogeneous graph between the user and the course, and enriching the input characteristics;
a knowledge progression layer to model a learning process of the target user with a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
the prediction layer is used for inputting the output result of the knowledge evolution layer into a full-connection network to obtain the predicted course registration rate;
and the pushing module is used for pushing the target recommended course to the target user according to the predicted course registration rate.
Another aspect of the embodiments of the present invention further provides an electronic device, including a processor and a memory;
the memory is used for storing programs;
the processor executes the program to implement the method as described above.
Yet another aspect of the embodiments of the present invention provides a computer-readable storage medium, which stores a program, which is executed by a processor to implement the method as described above.
The embodiment of the invention also discloses a computer program product or a computer program, which comprises computer instructions, and the computer instructions are stored in a computer readable storage medium. The computer instructions may be read by a processor of a computer device from a computer-readable storage medium, and the computer instructions executed by the processor cause the computer device to perform the foregoing method.
In the embodiment of the invention, historical course record information and candidate recommended course information of a target user are obtained on a data preprocessing layer, a heterogeneous graph of courses is constructed, and input characteristics of an input model are further constructed; in a knowledge extraction layer, according to a heterogeneous graph between the user and the course, acquiring high-order characteristics of the user and the course by adopting a graph convolution-based knowledge extraction layer, and enriching the input characteristics; at a knowledge progression layer, modeling a learning process of the target user using a collaborative attentional mechanism such that the input features can be combined with the learning process of the target user; inputting the output result of the knowledge evolution layer into a full-connection network at a prediction layer to obtain a predicted course registration rate; and pushing target recommended courses to the target user according to the predicted course registration rate. The method has high recommendation accuracy and comprehensiveness.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flowchart illustrating the overall steps provided by an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a heteromorphic image provided in an embodiment of the invention;
FIG. 3 is a flow chart of a knowledge extraction process based on graph convolution according to an embodiment of the present invention;
FIG. 4 is a flow chart of a process of a knowledge progression layer provided by an embodiment of the invention;
fig. 5 is an attention weight thermodynamic diagram of a knowledge progression layer provided by an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
In view of the problems in the prior art, an aspect of the embodiments of the present invention provides a course recommendation method based on an abnormal graph and a cooperative attentiveness reduction mechanism, as shown in fig. 1, the method of the present invention includes the following steps:
acquiring historical course record information and candidate recommended course information of a target user on a data preprocessing layer, constructing a heterogeneous graph of courses, and further constructing input characteristics of an input model;
in a knowledge extraction layer, according to a heterogeneous graph between the user and the course, acquiring high-order characteristics of the user and the course by adopting a graph convolution-based knowledge extraction layer, and enriching the input characteristics;
in a knowledge advancement layer, modeling a learning process of the target user using a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
inputting the output result of the knowledge evolution layer into a full-connection network at a prediction layer to obtain a predicted course registration rate;
and pushing target recommended courses to the target user according to the predicted course registration rate.
Optionally, the heterogeneous graph includes interaction data of the user, course first-next relations and course knowledge points;
the method for constructing the heterogeneous course graph and further constructing the input characteristics of the input model comprises the following steps of:
constructing a bipartite graph of the courses corresponding to the user according to the registration history of the user in the training set;
constructing a sequence relationship chart of the courses;
determining knowledge points contained in the course according to the catalog and outline of the course;
connecting knowledge points contained in the courses with all the courses, and merging the bipartite graph and the sequence relation graph to obtain a heterogeneous graph between the user and the courses;
and obtaining the high-order characteristics of each node in the abnormal graph by using a graph convolution network according to the abnormal graph.
Optionally, the obtaining, according to the heterogeneous graph between the user and the course, the high-order features of the user and the course by using a knowledge extraction layer based on graph convolution, and enriching the input features includes:
acquiring initial characteristics of a user and a course through an embedded layer;
for a knowledge point, obtaining semantic features of the knowledge point through a pre-trained w2v model;
each time the node passes through a knowledge extraction layer, the characteristics of the neighbor nodes are summed and normalized by aggregating the node characteristics of the neighbors corresponding to the users and the courses on the graph, and the high-order characteristics of each node are obtained; wherein the user aggregated neighbors are courses and the course aggregated neighbors include: semantic information of knowledge points contained in the course, the course which is followed by the current course and the user;
and averaging the obtained characteristics of each user node and the course node in different layers through a plurality of knowledge extraction layers, and adding the characteristics of the course and the class characteristics corresponding to the course to obtain the final characteristics of the node.
Optionally, the modeling the learning process of the target user with a cooperative attentive power mechanism such that the input features can be combined with the learning process of the target user includes:
for each course, calculating the monotonous attention weight of each course; the monotone is used for representing the course query characteristic, and the monotone is multiplied by the key characteristic points of the course and the courses learned before the course to calculate the attention score;
calculating the attention weight after cooperative attenuation according to the exponential attenuation rate and the monotonous attention weight;
according to the attention weight after cooperative attenuation, carrying out weighted summation on the attention weight and each characteristic to obtain the characteristic of the current course;
and acquiring the final characteristics of each course as the input characteristics.
Optionally, the method further comprises: and determining the evaluation index of the course recommendation result by combining the knowledge points.
Optionally, the determining, in combination with the knowledge point, an evaluation index of the course recommendation result includes:
calculating the similarity of courses between any two courses by calculating the same number of knowledge points contained in the courses according to the covering condition of the knowledge points of the courses;
for each user, calculating a course set which is most likely to be registered by the user, then calculating the mean value of the course similarity of each course actually selected by the user and a recommended target course according to the course similarity, obtaining the overall similarity of the set of recommended courses and the target course, and further obtaining the evaluation index of the course recommendation result.
Another aspect of the embodiments of the present invention further provides a course recommendation system based on an abnormal picture and a cooperative attentive power attenuation mechanism, including:
the data preprocessing layer is used for acquiring historical course record information and candidate recommended course information of a target user, constructing a heterogeneous graph of courses and further constructing input characteristics of an input model;
the knowledge extraction layer is used for acquiring high-order characteristics of the user and the course by adopting the knowledge extraction layer based on graph convolution according to the heterogeneous graph between the user and the course, and enriching the input characteristics;
a knowledge progression layer to model a learning process of the target user with a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
the prediction layer is used for inputting the output result of the knowledge evolution layer into a fully-connected network to obtain the predicted course registration rate;
and the pushing module is used for pushing the target recommended course to the target user according to the predicted course registration rate.
Another aspect of the embodiments of the present invention further provides an electronic device, including a processor and a memory;
the memory is used for storing programs;
the processor executes the program to implement the method as described above.
Yet another aspect of the embodiments of the present invention provides a computer-readable storage medium, which stores a program, which is executed by a processor to implement the method as described above.
The embodiment of the invention also discloses a computer program product or a computer program, which comprises computer instructions, and the computer instructions are stored in a computer readable storage medium. The computer instructions may be read by a processor of a computer device from a computer-readable storage medium, and the computer instructions executed by the processor cause the computer device to perform the foregoing method.
The following detailed description of the specific implementation principles of the present invention is made with reference to the accompanying drawings:
the invention introduces three key factors: the method comprises the steps that firstly, a special composition graph is formed, the graph comprises interaction records of user courses, sequence relations of the courses and knowledge points contained in the courses, and high-order signals and semantic information of the knowledge points of the special composition graph are captured by a knowledge extraction layer to enrich the representation of the users and the courses; and the second is a knowledge evolution layer, which models the learning process of the user by using a context-dependent attention mechanism, thereby capturing the knowledge level of the user better. And thirdly, guidance is provided for the student to review the course through visualization of the model. Therefore, a better recommendation effect is achieved, and review guidance is provided for continuous learning of students.
The traditional evaluation index generally focuses on whether the course set recommended by the recommendation system comprises the target course of the user, and the overall recommendation performance of the set is not comprehensively evaluated, so that the recommendation system is evaluated by using the course similarity evaluation index related to the knowledge point, and the recommendation system is helped to achieve a better recommendation effect.
For the data set used by the invention, the interactive data of the users are firstly subjected to deduplication and cleaning, and the last course of 20% of the users is taken out to be used as a test set. In the case where each instance of the training set and the test set is a user and its corresponding history registration course, the present invention selects the last door of the history record as the target recommended course. The training set corresponding to each target class will randomly sample 4 gates as negative examples, and the test set will randomly sample 99 gates as negative examples.
Specifically, the method of the present invention comprises the steps of:
the first step is as follows: data pre-processing
For the user characteristics, the invention fixes the historical course record quantity of the user to be 20 doors, and the historical course record quantity, the user and the target recommended course record are jointly used as a characteristic input model.
The invention constructs a heterogeneous graph about courses, which includes user interaction data, course first and second relations and course knowledge points. Specifically, the user registration history records exist in the training set, and the invention forms a bipartite graph of the corresponding courses of the user through the history records, that is, if the user learns a certain lesson, the relationship between the user and the lesson exists in the interactive graph. Then, the invention constructs a class continuation relationship graph, for example, if the first-time course required by one class is another, the two classes have an edge in the class continuation relationship graph. The invention determines which knowledge points are contained in each course according to the category and outline of the course, and connects the relevant knowledge points with the course. The above figures are finally merged, and the heterogeneous diagram is shown in fig. 2.
For a heterogeneous graph, the invention can then use the graph convolution network to obtain the high-order features of each node in the graph.
The second step is that: knowledge extraction layer based on graph convolution
The positions of the user and the course in the heterogeneous graph are different, so the structure of the graph has certain influence on the characteristics of the user and the graph contains the sequence information and knowledge points of the course and the cooperative signal of course selection of the user. The present invention utilizes graph convolution to capture high-level features of users and classes using heterogeneous graphs to enrich their own features.
As shown in FIG. 3, the present invention first obtains the original features of the user and the lesson through the embedding layer. Aiming at the knowledge points, the semantic features of the knowledge points are obtained by utilizing a pre-trained model w2v (words are mapped to an embedded pre-trained model). After passing through a knowledge extraction layer every time, the method obtains the high-level characteristics of the node by aggregating the node characteristics of the neighbors corresponding to the node characteristics on the graph, namely summing and normalizing the characteristics of the neighbor nodes. It should be noted here that the aggregated neighborhood of the user is the course, and the course aggregated neighborhood includes three parts: the self contains the semantic information of the knowledge points, which is the course and the user that follow the course first. Finally, through a plurality of knowledge extraction layers, the obtained characteristics of different layers of each node (user and course) are averaged to obtain the final characteristics of the node, and in addition, the course characteristics are added with the category characteristics corresponding to the course. The method not only captures the structure information and the cooperative signals of the heterogeneous graph, but also captures the knowledge points and the sequence information of the courses, thereby greatly enriching the characteristics of the users and the courses.
The third step: knowledge progression layer-user learning process modeling
The historical learning courses constitute the knowledge level of the user. However, according to the forgetting curve, the user's historical lessons should provide different amounts of information corresponding to the target recommended lessons. The relevance and time interval between the historical learning course and the target recommended course jointly determine the contribution degree of the historical course. Therefore, for the obtained user and historical learning courses and target recommended courses, the learning process of the user is modeled by a monotone cooperative attenuation attention mechanism, and the cooperation refers to the trend cooperation of a forgetting curve.
Wherein, each course can get the same inquiry of corresponding dimension, key and value characteristic through 3 different full connection levels. As shown in FIG. 4, taking the feature of one course obtained by calculation as an example, the present invention first obtains a monotonic attention weight. Monotonic refers to the fact that the course query feature is multiplied by the key feature points of the course and the courses learned before the course to obtain a calculated attention score. Cooperative attenuation refers to multiplying the obtained attention weight by an exponential attenuation rate, wherein the attenuation rate is determined by the past course and the registration time interval of the course, and the longer the time interval is, the more the attenuation is, the process that a user learns and forgets is simulated. After the attenuated attention weight is obtained, the method obtains the characteristics of the current course through the weighted summation of the attention weight and each characteristic. Similarly, the characteristics of each course can be obtained in such a way, and the characteristics are spliced and output.
Such derived course characteristics model the process of user learning. Eventually, when a course is recommended, a history course similar to the recommended course provides a larger amount of information, and a course that is a longer time provides a smaller amount of information.
The fourth step: prediction layer
The output of the knowledge evolution is a feature with a larger dimension, which comprises context information and a high-order feature of a collaborative sequence diagram, and what the invention needs to predict is the registration rate of the user for the target course. Therefore, the invention predicts the target course registration rate of the user through a full-connection network, and the predicted course registration rate can be obtained by inputting the output of the knowledge evolution layer into the full-connection network.
The fifth step: recommendation system evaluation index combined with knowledge points
Besides popular evaluation indexes such as AUC (area under susceptibility curve), HR (click rate), NDCG (normalized breaking cumulative gain) and MRR (average rank reciprocal), the invention also provides a course recommendation evaluation index based on knowledge points. Specifically, the method comprises the following steps: according to the covering condition of the course knowledge points, the method calculates the similarity of the courses by calculating the same number of the knowledge points contained in the course. For each user, the method can calculate the course set which is most likely to be registered by the user through the model, and then calculate the average value of the course similarity of the courses and the recommended target course according to the course similarity. The evaluation can more comprehensively evaluate whether the recommended course set meets the intention and interest of the user, and can help to improve the overall recommendation effect.
And a sixth step: comparative experiment
In order to prove the effectiveness of the model provided by the invention, a series of ablation experiments and comparative experiments are carried out.
In the comparative experiment, the comparative method of the invention comprises the following steps:
MLP it employs multi-layer awareness to represent users and corresponding courses, predicting the probability of a course being recommended to a user.
The NeuMF combines the traditional matrix decomposition method and the multilayer perceptron, can simultaneously extract low-dimensional and high-dimensional characteristics, and has a good recommendation effect.
CKE. This is a knowledge-graph-based representation regularization method that utilizes semantic embedding to enhance matrix factorization.
LightGCN. It learns the embedding of users and items by propagating linearly on a user-item interaction graph. The weighted sum of the user and item embeddings obtained at each level is then calculated as the final prediction score.
KGAT, it integrates knowledge map and interactive map into a unified map space; by recursive neighbor propagation target node embedding, a higher order relational model is explicitly established, and knowledge-dependent attention is used to distinguish the importance of neighbor embedding.
HRL, it utilizes a hierarchical reinforcement model to modify the user's profile, which is helpful to predict scores.
DIN it uses attention to capture the user's interest in historical behavior. The user's embedding varies with the candidate product, which effectively improves the performance of the model.
BST introduces a Transformer layer to model the historical behavior sequence of the user, thereby capturing the dynamic interest of the user.
Table 1 comparative experiment results table
| Models | AUC | HR@5 | HR@10 | NDCG@5 | NDCG@10 | MRR |
| MLP | 0.8507 | 0.4427 | 0.618 | 0.3059 | 0.3624 | 0.3026 |
| NeuMF | 0.882 | 0.4944 | 0.6534 | 0.3518 | 0.4033 | 0.3429 |
| CKE | 0.8643 | 0.4472 | 0.6055 | 0.3126 | 0.3623 | 0.305 |
| lightgcn | 0.8621 | 0.4834 | 0.6434 | 0.3457 | 0.3973 | 0.3385 |
| KGAT | 0.8888 | 0.4871 | 0.6533 | 0.3477 | 0.4013 | 0.3403 |
| HRL | 0.8853 | 0.4854 | 0.6567 | 0.343 | 0.3984 | 0.3367 |
| BST | 0.9026 | 0.5376 | 0.699 | 0.3853 | 0.4374 | 0.3718 |
| DIN | 0.8958 | 0.5331 | 0.6923 | 0.3795 | 0.431 | 0.3654 |
| Our Model | 0.901 | 0.5444 | 0.7094 | 0.391 | 0.4446 | 0.3773 |
As can be seen from the results in table 1, the proposed method is superior to the comparative baseline in all cases. MLP and NeuMF are not able to effectively extract the user's potential interests from the user's historical behavior, negatively impacting their performance. With respect to graph-convolutional network or knowledgegraph-based methods, the present invention organizes a collaborative sequence diagram into a knowledgegraph for KGAT and CKE. They have proven to be unable to achieve good recommendations due to the same deficiencies as MLP and NeuMF. Sequence-based methods are excellent in comparison of all methods because they overcome the problem of user preference bias. But HRL is less effective because the user's interaction data in the dataset is less, and it is more difficult to capture the user's interests after screening the user's historical lessons. DIN and BST are also poor because no additional information is captured (e.g., sequence information between courses) or modeling the user's learning process, as two already industrialized approaches.
The seventh step: interpretable recommendations
Fig. 5 provides an explanation of a monotonically cooperative decreasing attention mechanism showing an attention weight thermodynamic diagram for one user. The vertical axis represents the current course index and the horizontal axis represents the attention-focused course index. The present invention registers the contribution of the course to the recommended course from his 20 doors by attention weighting. This observation indicates that past related courses contain highly predictive information. User preferences for the current lesson may effectively be captured by attention with a certain rate of decay. While the attention weight of a distant past learning lesson is not as informative as a lesson learned in the near future. Similarity and time interval are key factors in controlling attention weight.
These observations suggest that the monotonic attention decay mechanism of the present invention can provide user feedback by linking the user's target lessons with lessons that they have learned in the past. This information may allow the user to select a particular course to review and consolidate, and then continue learning.
In summary, compared with the prior art, the invention has the following advantages:
1. the invention constructs a heterogeneous graph about courses, which comprises interaction data of users, the sequence information of the courses and knowledge points of the courses. For each user and each course, the invention obtains characteristic embedding of the user and the course through an embedding layer, and then sends the characteristic embedding into a knowledge extraction layer based on a graph volume network to obtain a high-order cooperative signal of the user and the course in a heterogeneous graph, wherein the course is first followed by information and information of knowledge points.
2. The invention combines a knowledge evolution layer and a cooperative attenuation attention mechanism to model the learning process of a user. Synergistic decay means that the weight of attention decays as the user's memory curve decays. In general, in the knowledge progression layer, in addition to similarity between courses, the present invention uses exponential decay and absolute distance measures to calculate attention weights.
3. The method of the invention can provide explanation through visual attention weight, so that the user can select specific relevant courses for review and consolidation before continuing learning.
4. The recommendation system evaluation index related to the knowledge points can capture the similarity between courses comprehensively according to the knowledge points, and the index is applied to the measurement recommendation system, so that the recommendation system can be evaluated more comprehensively, and the recommendation effect is improved.
The effectiveness of the proposed recommendation algorithm is confirmed by a series of ablation experiments and comparative experiments with real world data collected online from the academic hall. The algorithm can be deployed to an online course platform to provide personalized course recommendation service and review guidance for learners.
In alternative embodiments, the functions/acts noted in the block diagrams may occur out of the order noted in the operational illustrations. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality/acts involved. Furthermore, the embodiments presented and described in the flow charts of the present invention are provided by way of example in order to provide a more thorough understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and in which sub-operations described as part of larger operations are performed independently.
Furthermore, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the described functions and/or features may be integrated in a single physical device and/or software module, or one or more functions and/or features may be implemented in a separate physical device or software module. It will also be appreciated that a detailed discussion of the actual implementation of each module is not necessary for an understanding of the present invention. Rather, the actual implementation of the various functional modules in the apparatus disclosed herein will be understood within the ordinary skill of an engineer, given the nature, function, and internal relationship of the modules. Accordingly, those skilled in the art can, using ordinary skill, practice the invention as set forth in the claims without undue experimentation. It is also to be understood that the specific concepts disclosed are merely illustrative of and not intended to limit the scope of the invention, which is defined by the appended claims and their full scope of equivalents.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (10)
1. The course recommendation method based on the heteromorphic graph and the cooperative attentive power attenuation mechanism is characterized by comprising the following steps of:
acquiring historical course record information and candidate recommended course information of a target user at a data preprocessing layer, constructing a heterogeneous graph of courses and further constructing input characteristics of an input model;
in a knowledge extraction layer, according to a heterogeneous graph between the user and the course, acquiring high-order characteristics of the user and the course by adopting a graph convolution-based knowledge extraction layer, and enriching the input characteristics;
in a knowledge advancement layer, modeling a learning process of the target user using a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
inputting the output result of the knowledge evolution layer into a full-connection network at a prediction layer to obtain a predicted course registration rate;
and pushing target recommended courses to the target user according to the predicted course registration rate.
2. The method for course recommendation based on an anomaly map and a cooperative attentive power reduction mechanism as claimed in claim 1, wherein the anomaly map comprises interaction data of a user, course first-next relations and course knowledge points;
the method for constructing the heterogeneous course graph and further constructing the input characteristics of the input model comprises the following steps of:
constructing a bipartite graph of the courses corresponding to the user according to the registration history of the user in the training set;
constructing a sequence relationship chart of the courses;
determining knowledge points contained in the course according to the catalog and outline of the course;
connecting knowledge points contained in the courses with all the courses, and merging the bipartite graph and the sequence relation graph to obtain a heterogeneous graph between the user and the courses;
and obtaining the high-order characteristics of each node in the abnormal graph by using a graph convolution network according to the abnormal graph.
3. The method for recommending lessons based on an anomalous map and a cooperative attentive system according to claim 1, wherein said step of obtaining high-order features of the user and the lessons by using a knowledge extraction layer based on a graph convolution according to an anomalous map between the user and the lessons, and enriching said input features comprises:
acquiring initial characteristics of a user and a course through an embedded layer;
for a knowledge point, obtaining semantic features of the knowledge point through a pre-trained w2v model;
each time the node passes through a knowledge extraction layer, the characteristics of the neighbor nodes are summed and normalized by aggregating the node characteristics of the neighbors corresponding to the users and the courses on the graph, and the high-order characteristics of each node are obtained; wherein the user aggregated neighbors are courses and the course aggregated neighbors include: semantic information of knowledge points contained in the course, the course which is followed by the current course and the user;
and averaging the obtained characteristics of each user node and the course node in different layers through a plurality of knowledge extraction layers, and adding the characteristics of the course and the class characteristics corresponding to the course to obtain the final characteristics of the node.
4. The method for curriculum recommendation based on an anomaly map and a cooperative attentiveness reduction mechanism according to claim 1, wherein said modeling the learning process of the target user with the cooperative attentiveness reduction mechanism such that the input features can be combined with the learning process of the target user comprises:
for each course, calculating the monotonous attention weight of each course; the monotone is used for representing the course query characteristic, and the monotone is multiplied by the key characteristic points of the course and the courses learned before the course to calculate the attention score;
calculating the attention weight after cooperative attenuation according to the exponential attenuation rate and the monotonous attention weight;
according to the attention weight after cooperative attenuation, carrying out weighted summation on the attention weight and each characteristic to obtain the characteristic of the current course;
and acquiring final characteristics of each course as the input characteristics.
5. The method for curriculum recommendation based on an anomaly map and a cooperative attentive mechanism as recited in claim 1, further comprising: and determining the evaluation index of the course recommendation result by combining the knowledge points.
6. The lesson recommendation method based on the heteromorphic graph and the cooperative attentive power mechanism as claimed in claim 5, wherein said determining an evaluation index of the lesson recommendation result by combining knowledge points comprises:
calculating the similarity of courses between any two courses by calculating the same number of knowledge points contained in the courses according to the covering condition of the knowledge points of the courses;
for each user, calculating a course set which is most likely to be registered by the user, then calculating the mean value of the course similarity of each course actually selected by the user and a recommended target course according to the course similarity, obtaining the overall similarity of the set of recommended courses and the target course, and further obtaining the evaluation index of the course recommendation result.
7. A course recommendation system based on an abnormal picture and a cooperative attentive system is characterized by comprising:
the data preprocessing layer is used for acquiring historical course record information and candidate recommended course information of a target user, constructing a heterogeneous graph of courses and further constructing input characteristics of an input model;
the knowledge extraction layer is used for acquiring high-order characteristics of the user and the course by adopting the knowledge extraction layer based on graph convolution according to the heterogeneous graph between the user and the course, and enriching the input characteristics;
a knowledge progression layer to model a learning process of the target user with a cooperative attentive mechanism such that the input features can be combined with the learning process of the target user;
the prediction layer is used for inputting the output result of the knowledge evolution layer into a full-connection network to obtain the predicted course registration rate;
and the pushing module is used for pushing the target recommended course to the target user according to the predicted course registration rate.
8. An electronic device comprising a processor and a memory;
the memory is used for storing programs;
the processor executing the program realizes the method of any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that the storage medium stores a program, which is executed by a processor to implement the method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program realizes the method according to any of claims 1 to 6 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210675138.6A CN115272015A (en) | 2022-06-15 | 2022-06-15 | Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210675138.6A CN115272015A (en) | 2022-06-15 | 2022-06-15 | Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115272015A true CN115272015A (en) | 2022-11-01 |
Family
ID=83760793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210675138.6A Pending CN115272015A (en) | 2022-06-15 | 2022-06-15 | Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115272015A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070034A (en) * | 2023-03-03 | 2023-05-05 | 江西财经大学 | Graph convolution network recommendation method combining self-adaptive period and interest quantity factor |
-
2022
- 2022-06-15 CN CN202210675138.6A patent/CN115272015A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070034A (en) * | 2023-03-03 | 2023-05-05 | 江西财经大学 | Graph convolution network recommendation method combining self-adaptive period and interest quantity factor |
| CN116070034B (en) * | 2023-03-03 | 2023-11-03 | 江西财经大学 | Graph convolution network recommendation method combining self-adaptive period and interest quantity factor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dejaeger et al. | Data mining techniques for software effort estimation: a comparative study | |
| CN111813921B (en) | Topic recommendation method, electronic device and computer-readable storage medium | |
| CN110717098B (en) | Meta-path-based context-aware user modeling method and sequence recommendation method | |
| CN111460249A (en) | Personalized learning resource recommendation method based on learner preference modeling | |
| CN111242310B (en) | Feature validity evaluation method and device, electronic equipment and storage medium | |
| Pang et al. | Adaptive recommendation for MOOC with collaborative filtering and time series | |
| CN113268611B (en) | Learning path optimization method based on deep knowledge tracking and reinforcement learning | |
| CN112597392A (en) | Recommendation system based on dynamic attention and hierarchical reinforcement learning | |
| CN115272015A (en) | Course recommendation method and system based on abnormal picture and cooperative attenuation attention mechanism | |
| CN115618101A (en) | Streaming media content recommendation method and device based on negative feedback and electronic equipment | |
| Zhu | Network course recommendation system based on double-layer attention mechanism | |
| CN113449182B (en) | Knowledge information personalized recommendation method and system | |
| CN112765339B (en) | Personalized book recommendation method and system based on reinforcement learning | |
| CN113449176A (en) | Recommendation method and device based on knowledge graph | |
| CN116012627A (en) | Causal time sequence dual-enhancement knowledge tracking method based on hypergraph clustering | |
| Han | Research on personalized recommendation method of educational resources based on learner behavior analysis | |
| Li et al. | Quantification and prediction of engagement: Applied to personalized course recommendation to reduce dropout in MOOCs | |
| Lee et al. | Prescribing deep attentive score prediction attracts improved student engagement | |
| Fu et al. | Predicting programming performance by using process behavior in a block-based programming environment | |
| Vaidhehi et al. | DESIGN OF A CONTEXT-AWARE RECOMMENDER SYSTEMS FOR UNDERGRADUATE PROGRAM RECOMMENDATIONS | |
| Wyatt | Reversing decision support systems to reveal differences in human strategizing behaviour | |
| CN115329883B (en) | Semantic similarity processing method, device and system and storage medium | |
| CN117633371B (en) | Recommendation method, device and readable storage medium based on multi-attention mechanism | |
| Muhammad et al. | Comparison of user-based and item-based collaborative filtering methods in recommender system | |
| CN114528482B (en) | Method and device for determining recommended object, electronic equipment and storage medium |
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 |