CN117349492B

CN117349492B - Course recommendation method and system based on learning data

Info

Publication number: CN117349492B
Application number: CN202311661137.7A
Authority: CN
Inventors: 李建伟
Original assignee: Guoxin Blue Bridge Education Technology Co ltd
Current assignee: Guoxin Blue Bridge Education Technology Co ltd
Priority date: 2023-12-06
Filing date: 2023-12-06
Publication date: 2024-05-31
Anticipated expiration: 2043-12-06
Also published as: CN117349492A

Abstract

The invention discloses a course recommendation method and a course recommendation system based on learning data. And converting the course key tree containing all course relations into a graph, and converting the learned courses with objective learning sequence into the graph. According to the key word black-and-white image and the real black-and-white image, the learning sequence among all courses is adjusted by the learned objective learning sequence through two methods of data judgment and image judgment. Therefore, the course recommendation is carried out according to the sequence, so that the recommended course can not only contain the knowledge containing relation of the course, but also adapt to the user according to the learned course, and the course recommendation accuracy is high.

Description

Course recommendation method and system based on learning data

Technical Field

The invention relates to the technical field of computers, in particular to a course recommendation method and system based on learning data.

Background

At present, most course recommendation methods only recommend courses of users according to habits of the users and big data, and the sequence of course learning is not considered. In general, knowledge intercommunication exists among a plurality of courses, and learning basic knowledge is more convenient when learning higher-order knowledge later. And learning the high-order knowledge first is difficult to fuse. And knowledge relations among courses are complex, and better relation among courses is required to be extracted for course recommendation.

Disclosure of Invention

The invention aims to provide a course recommendation method and system based on learning data, which are used for solving the problems in the prior art.

In a first aspect, an embodiment of the present invention provides a course recommendation method based on learning data, including:

acquiring learning data; the learning data includes a plurality of pre-learning courses and learned course information; the learned lesson information includes a plurality of learned lessons and a learning order between the learned lessons;

Obtaining a plurality of course keyword sets; one course keyword set corresponds to one course which is learned or not learned by the user; elements in the course keyword set represent a summary of knowledge learned by the course;

Constructing a course keyword tree according to the course keyword sets; the course keyword tree represents a tree formed by the sequence of course learning arranged according to keywords among all courses;

copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image;

marking the learning sequence among the learned courses of the user based on the learned course information to obtain a real black-and-white image;

reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the self-adaptive course tree learns the learning sequence from the morning to the evening of all course recommendations for a user;

based on the level of the adaptive course tree corresponding to the plurality of pre-learning courses, recommending the plurality of pre-learning courses to the user.

Optionally, reconstructing a course learning sequence suitable for user learning according to the keyword black-and-white image and the real black-and-white image to obtain an adaptive course tree, including:

Constructing a coordinate axis by taking the lower left corner of the black-and-white keyword image as an origin; the abscissa of the black-and-white keyword image is the hierarchy of the course keyword tree; the ordinate of the key word black-and-white image represents a plurality of learned courses; the key word black-and-white image is a black-and-white image which stores the hierarchical information of the course key word tree;

Constructing a coordinate axis by taking the lower left corner of a real black-and-white image as an origin; the abscissa of the real black-and-white image is the label of the learned sequence; the ordinate of the real black-and-white image represents a plurality of learned courses;

judging the difference of learning sequences of courses according to the key word black-and-white images and the real black-and-white images to obtain a difference value;

inputting the key word black-and-white image and the real black-and-white image into a first convolution network, and adjusting the key word black-and-white image to obtain a first adjustment image;

the learned course part of the first adjustment image is kept black, other areas are set to be white, and the difference between the learned course part and the black-and-white image of the key word is judged to obtain a first convolution difference value;

And if the sum of the difference value and the first convolution difference value is smaller than the difference threshold value, converting the first adjustment image into a self-adaptive course tree according to the hierarchical relation of which the abscissa is taken as a tree and the ordinate is taken as a corresponding course.

Optionally, the obtaining a difference value according to the difference of the learning sequence of the keyword black-and-white image and the real black-and-white image, includes:

Acquiring the abscissa corresponding to the same course in the real black-and-white image in the key word black-and-white image, and sequentially arranging the courses in the real black-and-white image according to the learning sequence of the courses in the real black-and-white image to obtain a level vector;

Subtracting the next element from the previous element to obtain an adjacent vector;

Arranging elements with 0 and negative numbers in adjacent vectors from small to large to obtain positive vectors;

arranging elements with negative numbers in adjacent vectors from large to small to obtain negative vectors;

Inserting the negative vector into the end of the positive vector to obtain a sequence vector;

forming a key learning sequence by two courses corresponding to elements in the sequence vector according to the arranged sequence;

and subtracting the learned courses from the key learning sequence according to the same courses to obtain a gap value.

Optionally, one node of the course keyword tree corresponds to one course keyword set;

The course keyword set is divided into an old keyword and a new keyword; the old keywords are content for using the previous courses; the new keywords are the new learning content of the courses corresponding to the course keyword sets.

Optionally, the constructing a course keyword tree according to the plurality of course keyword sets includes:

Obtaining the same number of new keywords of one keyword set and old keywords of other keyword sets as elements of the set to obtain the same set;

the course keyword set corresponding to the element with the value smaller than that of other elements in the same set is used as a first course keyword set;

obtaining the number of new keywords in a first course keyword set which is the same as the number of old keywords in other course keyword sets, and obtaining a first number set;

Using a course keyword set corresponding to elements larger than a first threshold value in the first number set as a child node of the first course keyword set;

and repeatedly obtaining the child nodes corresponding to each layer of nodes, and sequentially constructing a course keyword tree.

Optionally, the repeatedly obtaining the child nodes corresponding to each layer of nodes sequentially constructs a course keyword tree, including:

acquiring a second course keyword set; the second course keyword set is one course keyword set in the child nodes of the first course keyword set;

Acquiring a second other course keyword set; the second other course keyword sets are all course keyword sets except the second course keyword set and the first course keyword set;

Obtaining the same number of new keywords in a second course keyword set and old keywords in second other course keyword sets to obtain a second number set;

taking a course keyword set corresponding to an element larger than a first threshold value in the second number set as a child node of the second course keyword set;

if the course keyword sets representing the same course exist at different levels in the course keyword tree, saving the course keyword sets with the level larger than other nodes, and deleting other nodes;

Optionally, the abscissa of the real black-and-white image in the real black-and-white image and the area represented by one course are 1 to 1;

the horizontal coordinate of the black-and-white image of the keyword and the area represented by one course are 1-to-many or 1-to-1.

Optionally, the training method of the first convolutional network includes:

dividing the learned courses of the user into two parts according to different dividing points to obtain a plurality of training course sets;

The number of the training course sets is the number of different dividing modes of the learned courses divided into two parts;

Taking the learned courses which do not exist in the training course set and are adjacent to the learning sequence as annotation courses; a plurality of training courses correspondingly obtain a plurality of labeling courses;

The area where the training courses are located in the key word black-and-white image and the real black-and-white image is reserved, and other areas are converted into white, so that a training key word image and a training real image are obtained;

And obtaining loss by inputting the output value and the labeling data of the training keyword image and the training real image into a first convolution network, and training the first convolution network.

Optionally, the first convolutional network is a convolutional neural network.

In a second aspect, an embodiment of the present invention provides a course recommendation system based on learning data, including:

The acquisition module is used for: acquiring learning data; the learning data includes a plurality of pre-learning courses and learned course information; the learned lesson information includes a plurality of learned lessons and a learning order between the learned lessons; obtaining a plurality of course keyword sets; one course keyword set corresponds to one course which is learned or not learned by the user; elements in the course keyword set represent a summary of knowledge learned by the course;

The course keyword tree construction module: constructing a course keyword tree according to the course keyword sets; the course keyword tree represents a tree formed by the sequence of course learning arranged according to keywords among all courses;

And an image conversion module: copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image; marking the learning sequence among the learned courses of the user based on the learned course information to obtain a real black-and-white image;

And a learning sequence adjustment module: reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the self-adaptive course tree learns the learning sequence from the morning to the evening of all course recommendations for a user;

And a recommendation module: based on the level of the adaptive course tree corresponding to the plurality of pre-learning courses, recommending the plurality of pre-learning courses to the user.

Compared with the prior art, the embodiment of the invention achieves the following beneficial effects:

The embodiment of the invention also provides a course recommendation method and a course recommendation system based on the learning data, wherein the method comprises the following steps: learning data is acquired. The learning data includes a plurality of pre-learning courses and learned course information. The learned lesson information includes a plurality of learned lessons and a learning order between the learned lessons. A plurality of course keyword sets are obtained. One course keyword set corresponds to one course that the user has learned or has not learned. The elements in the course keyword set represent a summary of knowledge learned by the course. And constructing a course keyword tree according to the course keyword sets. The course keyword tree represents a tree formed by the sequence of course learning arranged according to keywords among all courses. And copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image. And marking the learning sequence among the learned courses of the user based on the learned course information to obtain a real black-and-white image. Reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the adaptive course tree learns the early to late learning order of all course recommendations for the user. Based on the level of the adaptive course tree corresponding to the plurality of pre-learning courses, recommending the plurality of pre-learning courses to the user.

The invention adopts the method of extracting course keywords to summarize course knowledge, and builds a course keyword tree under the containing condition of the knowledge. And converting the course key tree containing all course relations into a graph, and converting the learned courses with objective learning sequence into the graph. According to the key word black-and-white image and the real black-and-white image, the learning sequence among all courses is adjusted by the learned objective learning sequence through two methods of data judgment and image judgment. The course recommendation is performed according to the sequence, so that the recommended courses can not only contain knowledge containing relations of the courses, but also adapt to users according to the learned courses.

Drawings

Fig. 1 is a flowchart of a course recommendation method based on learning data according to an embodiment of the present invention.

Fig. 2 is a lesson keyword tree represented in a black-and-white image of keywords in a lesson recommendation method based on learning data according to an embodiment of the present invention.

Detailed Description

The present invention will be described in detail with reference to the accompanying drawings.

Example 1

As shown in fig. 1, an embodiment of the present invention provides a course recommendation method based on learning data, where the method includes:

s101: learning data is acquired. The learning data includes a plurality of pre-learning courses and learned course information. The learned lesson information includes a plurality of learned lessons and a learning order between the learned lessons.

The pre-learning data is a course which is expected to be learned by a user but not learned, the learned course is a course which is learned before, and the learning sequence among the learned courses is arranged from far to near according to the time point of learning. The pre-learning prepares courses to be learned and intended to be learned for the user.

S102: a plurality of course keyword sets are obtained. One course keyword set corresponds to one course that the user has learned or has not learned. The elements in the course keyword set represent a summary of knowledge learned by the course.

All courses learned and not learned by the user are stored in the database and are data which can be learned by the user. And the number is greater than the sum of the learned courses and pre-learned courses of the user.

In this embodiment, the elements in the course keyword set, that is, the keywords of the course, are obtained from the catalog of the course.

S103: and constructing a course keyword tree according to the course keyword sets. The course keyword tree represents a tree formed by the sequence of course learning arranged according to keywords among all courses.

S104: and copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image.

The key word black-and-white image constructs the abscissa of the key word black-and-white image in the hierarchy of the course key word tree. In this embodiment, the abscissa of the black-and-white keyword image is established at intervals of 1, and the initial value 0 of the abscissa of the black-and-white keyword image is that if the node is the root node, the node is placed at the position with the initial value of 1, and the child node of the root node is placed at the position with the initial value of 2, so that the course keyword tree is sequentially converted into the black-and-white keyword image.

S105: the learning sequence marks among the learned courses are taken as the abscissa of the real black-and-white image, the learned courses are taken as the ordinate of the real black-and-white image, and the learning sequence structure among the learned courses of the user is copied in the real black-and-white image; the true black-and-white image indicates the presence of a lesson in black, and indicates the absence of a lesson in white.

Wherein, the key word black-and-white image and the real black-and-white image are represented by black to show that courses exist, and white to show that no courses exist.

S106: reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the adaptive course tree learns the early to late learning order of all course recommendations for the user.

If the level corresponding to one course in the adaptive course tree is 1 and the level corresponding to the other course is 2, the course with the level of 2 is recommended after the course with the level of 1 is recommended.

S107: based on the level of the multiple pre-learning courses corresponding to the adaptive course tree, recommending the multiple pre-learning courses to the user.

And recommending a plurality of pre-learning courses to the user from small to large at a level corresponding to the self-adaptive course tree.

And constructing a coordinate axis by taking the lower left corner of the black-and-white keyword image as an origin. The abscissa of the black-and-white image of the keyword is the hierarchy of the course keyword tree. The ordinate of the key black-and-white image represents a plurality of learned courses. The keyword black-and-white image is a black-and-white image which stores hierarchical information of a course keyword tree.

And constructing a coordinate axis by taking the lower left corner of the real black-and-white image as an origin. The abscissa of the real black-and-white image is the label of the learned sequence. The ordinate of the true black-and-white image represents a plurality of learned courses.

And judging the difference of the learning sequence of the lessons according to the key word black-and-white image and the real black-and-white image to obtain a difference value.

And inputting the keyword black-and-white image and the real black-and-white image into a first convolution network, and judging the difference between the keyword black-and-white image and the real black-and-white image to obtain a first adjustment image.

And keeping black in the learned course part of the first adjustment image, setting other areas to be white, and judging the difference between the first adjustment image and the black-and-white image of the key word to obtain a first convolution difference value.

In this embodiment, the gap threshold is 5.

By the method, the corresponding positions of the key word black-and-white image and the real black-and-white image are converted into black-and-white according to whether learning courses are performed. And adjusting the learning sequence of the lessons corresponding to the key black-and-white images according to the specific lessons of the real black-and-white images, and judging the relation of the lessons as the adjusted relation of the lessons according to the difference between the directly calculated difference value and the pixel values of the key black-and-white images and the real black-and-white images. The reconstructed self-adaptive course tree can not only contain knowledge containing relations of courses, but also adapt to users according to the learned courses.

Optionally, the step of judging the difference of the learning sequence of the lesson according to the key word black-and-white image and the real black-and-white image to obtain a difference value includes:

Acquiring an abscissa corresponding to the same course in the real black-and-white image in the key black-and-white image as an element of a coordinate set; and sequentially arranging elements in the coordinate set according to the learning sequence of courses in the real black-and-white image to obtain a hierarchical vector.

In this embodiment, as in the present invention, the learning sequence of the a course is 1, the learning sequence of the b course is 2, the learning sequence of the c course is 3, the learning sequence of the d course is 4, and the learning sequence of the e course is 5 in the real black-and-white image. A lesson keyword tree represented in a black-and-white image of keywords is shown in fig. 2. The hierarchical vector is [4,1,3,2,2].

And subtracting the next element from the previous element to obtain an adjacent vector.

And performing difference operation on two adjacent elements in the hierarchical vector.

Wherein the adjacent vector is [3, -2,1,0]

And arranging elements with 0 and positive numbers in adjacent vectors from small to large to obtain positive vectors.

Wherein the positive vector is [0, -1,3].

And arranging elements with negative numbers in adjacent vectors from large to small to obtain negative vectors.

Wherein the negative vector is [ -2].

And inserting the negative vector into the end of the positive vector to obtain the sequence vector.

Wherein the order vector is [0,1,3, -2].

And forming a keyword learning sequence by two courses corresponding to the elements in the sequence vector according to the arranged sequence.

If two courses corresponding to the elements in the sequence vector appear in the front, only non-appearing courses are filled in. As in the present embodiment, the keyword learning order indicates [ D, E, C, a, B ].

By the method, the learning sequence with the smallest value of the key jump level is used as the self-adaptive course tree, wherein the learning sequence with the smallest difference from the courses in the key black-and-white image is found in the key black-and-white image.

Optionally, one node of the course keyword tree corresponds to one course keyword set.

The course keyword set is divided into an old keyword and a new keyword. The old key is content for using the previous course. The new keywords are the new learning content of the courses corresponding to the course keyword sets.

By the method, the used course keywords are classified. Thereby acquiring the inclusion relation of the course knowledge.

The method comprises the steps of obtaining the same number of new keywords of one keyword set and old keywords of other keyword sets as elements of the set, and obtaining the same set.

And taking the course keyword set corresponding to the element with the value smaller than that of other elements in the same set as the first course keyword set.

And acquiring the number of new keywords in the first course keyword set which is the same as the number of old keywords in other course keyword sets, and obtaining a first number set.

Wherein the elements in the first number set represent the number of keywords that overlap for two different courses.

And taking the course keyword set corresponding to the elements larger than the first threshold value in the first number set as the child node of the first course keyword set.

In this embodiment, the value of the first threshold is 20.

By the method, the inclusion relation of the keywords is obtained, and courses are put into different levels according to the inclusion of the keywords. Thereby obtaining a course keyword tree meeting the course inclusion relation.

And obtaining a second course keyword set. The second course keyword set is one course keyword set in the child nodes of the first course keyword set.

A second set of other course keywords is obtained. The second set of other course keywords is the entire set of course keywords except the second set of course keywords and the first set of course keywords.

And obtaining the number of the new keywords in the second course keyword set which is the same as the number of the old keywords in the second other course keyword sets, and obtaining a second number set.

And taking the course keyword set corresponding to the elements larger than the first threshold value in the second number set as the child node of the second course keyword set.

In this embodiment, the first threshold is 20, which indicates that the second course keyword set overlaps 20 course keywords of other course keyword sets except the root node and the second course keyword set.

If the course keyword sets representing the same course exist at different levels in the course keyword tree, the course keyword sets with the level larger than other nodes are stored, and other nodes are deleted.

Through the method, if the curriculum keyword set of the current level is overlapped with other curriculum keyword sets except the root node and the current curriculum keyword set, the curriculum keyword set with the number larger than the first threshold value is added into the nodes of the current curriculum keyword tree, the level is smaller than the current level, and the corresponding nodes on the curriculum keyword tree are transferred to the child nodes of the current level.

Optionally, the abscissa of the true black-and-white image and the area represented by one course are 1 to 1.

Optionally, the training method of the first convolutional network includes:

dividing the learned courses of the user into two parts according to different dividing points to obtain a plurality of training course sets.

The number of training course sets is the number of different dividing modes of the learned courses divided into two parts.

And taking the learned courses which do not exist in the training course set and are adjacent to the learning sequence as annotation courses. The plurality of training courses correspondingly obtain a plurality of labeling courses.

And reserving the area where the training courses are located in the keyword black-and-white image and the real black-and-white image, and converting other areas into white to obtain the training keyword image and the training real image.

In this embodiment, the loss function for obtaining the loss is a cross entropy loss function.

Optionally, the first convolutional network is a convolutional neural network.

Wherein the first convolutional network is a convolutional neural network (convolutional neural network, CNN).

Example 2

Based on the course recommendation method based on the learning data, the embodiment of the invention also provides a course recommendation system based on the learning data, which comprises the following steps:

The acquisition module is used for: learning data is acquired. The learning data includes a plurality of pre-learning courses and learned course information. The learned lesson information includes a plurality of learned lessons and a learning order between the learned lessons. A plurality of course keyword sets are obtained. One course keyword set corresponds to one course that the user has learned or has not learned. The elements in the course keyword set represent a summary of knowledge learned by the course.

The course keyword tree construction module: and constructing a course keyword tree according to the course keyword sets. The course keyword tree represents a tree formed by the sequence of course learning arranged according to keywords among all courses.

And an image conversion module: and copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image. And marking the learning sequence among the learned courses of the user based on the learned course information to obtain a real black-and-white image.

And a learning sequence adjustment module: reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the adaptive course tree learns the early to late learning order of all course recommendations for the user.

The algorithms and displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, the present invention is not directed to any particular programming language. It will be appreciated that the teachings of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of enablement and best mode of the present invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

Various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functions of some or all of the components in an apparatus according to embodiments of the present invention may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present invention can also be implemented as an apparatus or device program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present invention may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names.

Claims

1. A course recommendation method based on learning data, comprising:

The learning sequence of the learned courses is copied into the real black-and-white image by taking the learning sequence of the learned courses as the abscissa of the real black-and-white image and taking the learned courses as the ordinate of the real black-and-white image; the true black-and-white image indicates that courses exist by black, and white indicates that no courses exist;

Reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree; the hierarchy of the self-adaptive course tree represents a learning sequence from morning to evening in which a user learns all course recommendations;

Sequentially recommending a plurality of pre-learning courses to a user according to the corresponding levels of the self-adaptive course tree;

reconstructing a course learning sequence suitable for user learning according to the key word black-and-white image and the real black-and-white image to obtain a self-adaptive course tree, wherein the self-adaptive course tree comprises:

the learned course part of the first adjustment image is kept black, other areas are set to be white, and the difference between the image and the black-and-white keyword image is judged to obtain a first convolution difference value;

if the sum of the difference value and the first convolution difference value is smaller than a difference threshold value, converting the first adjustment image into a self-adaptive course tree according to the hierarchical relation of which the abscissa is taken as a tree and the ordinate is taken as a corresponding course;

the step of obtaining a difference value according to the difference of the learning sequence of the determined courses according to the key word black-and-white image and the real black-and-white image comprises the following steps:

Acquiring an abscissa corresponding to the same course in the real black-and-white image in the key black-and-white image as an element in the hierarchical vector; sequentially arranging the elements according to the learning sequence of courses in the real black-and-white image to obtain a hierarchical vector;

arranging elements with 0 and positive numbers in adjacent vectors from small to large to obtain positive vectors;

2. The course recommendation method based on learning data according to claim 1, wherein one node of the course keyword tree corresponds to one course keyword set;

3. The course recommendation method based on learning data according to claim 1, wherein said constructing a course keyword tree from said plurality of course keyword sets comprises:

4. The course recommendation method based on learning data according to claim 3, wherein the repeatedly obtaining the child nodes corresponding to each layer of nodes sequentially constructs a course keyword tree, and the method comprises:

5. The lesson recommendation method based on learning data according to claim 1, wherein an abscissa of a real black-and-white image and a region represented by one lesson among the real black-and-white images are 1 to 1;

6. The lesson recommendation method based on learning data as claimed in claim 1, wherein the training method of the first convolutional network comprises:

7. The lesson recommendation method based on learning data as claimed in claim 6, wherein the first convolutional network is a convolutional neural network.

8. A course recommendation system based on learning data, comprising:

And an image conversion module: copying the learning sequence between courses of the course keyword tree in the image to obtain a keyword black-and-white image; the learning sequence of the learned courses is copied into the real black-and-white image by taking the learning sequence of the learned courses as the abscissa of the real black-and-white image and taking the learned courses as the ordinate of the real black-and-white image; the true black-and-white image indicates that courses exist by black, and white indicates that no courses exist;

and a recommendation module: sequentially recommending a plurality of pre-learning courses to a user according to the corresponding levels of the self-adaptive course tree;