Model training method, course recommendation method, device, equipment and medium

Abstract

The application provides a model training method, a course recommendation method, a device, equipment and a medium. The method comprises the following steps: obtaining an initial training sample, the initial training sample being a sample obtained from historical data information of a learning platform, the initial training sample comprising a student and a course learned by the student; determining a target course in the course, learned information of the target course, a target student in the student and learning attribute information of the target student according to the initial training sample; determining the correlation degree of a target professional category and a historical learning course according to the target professional category and a course attribute category; optimizing the initial training sample according to the course quality and the correlation degree to obtain an optimized training sample; and training a to-be-trained model according to the optimized training sample to obtain a target training model. The method improves the training effect of the to-be-trained model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of course recommendation, and in particular to a model training method, a course recommendation method, a device, equipment and a medium. BACKGROUND

[0002] It is a common method in today's information explosion society to help learners find the courses they need to learn more efficiently and reduce information overload by recommending courses that learners are interested in through a learning platform.

[0003] Currently, when recommending courses to learners, the learning platform in the prior art generally collects learner information to recommend courses to learners through an internally set course recommendation model.

[0004] However, the existing recommendation model has the problem of low accuracy of recommended courses due to poor model training effect. SUMMARY

[0005] The present application provides a model training method, a course recommendation method, a device, equipment and a medium to solve the problem of low accuracy of recommended courses due to poor model training effect.

[0006] In a first aspect, the present application provides a model training method applied to a learning platform, comprising:

[0007] obtaining an initial training sample, the initial training sample being a sample obtained from historical data information of the learning platform, the initial training sample including a learner and a course learned by the learner;

[0008] determining a target course in the course, learned information of the target course, a target learner in the learner and learning attribute information of the target learner according to the initial training sample, wherein the learned information includes learned time information of the target course obtained from the historical data information, and the learning attribute information includes a target professional category of the target learner and a course attribute category of a historical learning course learned by the target learner obtained from the historical data information;

[0009] determining a course quality of the target course according to the learned time information;

[0010] determining an association degree of the target professional category and the historical learning course according to the target professional category and the course attribute category;

[0011] optimizing the initial training sample according to the course quality and the association degree to obtain an optimized training sample;

[0012] training a to-be-trained model according to the optimized training sample to obtain a target training model.

[0013] In the present application, the initial training sample is obtained, comprising:

[0014] The historical data information is obtained;

[0015] According to the historical data information, the to-be-trained sample is determined, and the to-be-trained sample includes student information, course information and student professional category;

[0016] According to the student professional category, the first target to-be-trained sample is determined, wherein the student professional category in the first target to-be-trained sample is the professional category;

[0017] According to the first target to-be-trained sample, the initial training sample is obtained.

[0018] In the present application, according to the first target to-be-trained sample, the initial training sample is obtained, comprising:

[0019] According to the first target to-be-trained sample, the second target to-be-trained sample is determined from the to-be-trained sample, wherein the student information in the second target to-be-trained sample meets the requirement of the preset student information similarity with the student information in the first target to-be-trained sample, and the course information in the second target to-be-trained sample meets the requirement of the preset course information similarity with the course information in the first target to-be-trained sample;

[0020] According to the first target to-be-trained sample and the second target to-be-trained sample, the initial training sample is obtained.

[0021] In the present application, according to the learned time information, the course quality of the target course is determined, comprising:

[0022] According to the learned time information, the learned number of times, the average course completion rate and the average learned duration of the target course are determined;

[0023] The historical learned time information of each course is obtained from the historical data information;

[0024] According to the learned number of times, the average course completion rate, the average learned duration and the historical learned time information, the course quality of the target course is determined.

[0025] In the present application, according to the learned number of times, the average course completion rate, the average learned duration and the historical learned time information, the course quality of the target course is determined, comprising:

[0026] According to the historical learning time information, the historical learning times, the historical average course completion rate, the historical average learning duration, the target historical learning times, the target historical average course completion rate and the target historical average learning duration are determined, wherein the target historical learning times are the average value of the historical learning times greater than the preset number of times, the target historical average course completion rate is the average value of the historical average course completion rate greater than the preset completion rate, and the target historical average learning duration is the average value of the historical average learning duration greater than the preset duration;

[0027] According to the learning times, the average course completion rate, the average learning duration, the historical learning times, the historical average course completion rate, the historical average learning duration, the target historical learning times, the target historical average course completion rate and the target historical average learning duration, the course quality of the target course is determined.

[0028] In the present application, according to the learning times, the average course completion rate, the average learning duration, the historical learning times, the historical average course completion rate, the historical average learning duration, the target historical learning times, the target historical average course completion rate and the target historical average learning duration, the course quality of the target course is determined, comprising:

[0029] According to the learning times and the historical average learning times, a first index result is obtained;

[0030] According to the average course completion rate and the historical average course completion rate, a second index result is obtained;

[0031] According to the average learning duration and the historical average learning duration, a third index result is obtained;

[0032] According to the learning times and the target historical learning times, a fourth index result is obtained;

[0033] According to the average course completion rate and the target historical average course completion rate, a fifth index result is obtained;

[0034] According to the average learning duration and the target historical average learning duration, a sixth index result is obtained;

[0035] According to the first index result, the second index result, the third index result, the fourth index result, the fifth index result and the sixth index result, the course quality of the target course is determined.

[0036] In the present application, according to the target professional category and the course attribute category, the correlation degree of the target professional category and the historical learning course is determined, comprising:

[0037] The target professional category and the course attribute category are compared to obtain a comparison result;

[0038] According to the comparison result, the correlation degree of the target professional category and the historical learning course is determined.

[0039] In the present application, according to the comparison result, the correlation degree of the target professional category and the historical learning course is determined, comprising:

[0040] If the comparison result represents that the target professional category and the course attribute category match, then according to the comparison result, the correlation degree of the target professional category and the historical learning course is determined.

[0041] In the present application, according to the comparison result, the correlation degree of the target professional category and the historical learning course is determined, comprising:

[0042] If the comparison result represents that the target professional category and the course attribute category do not match, then according to the comparison result, the number of corresponding groups of the students and the courses corresponding to each other in the initial training sample is determined, and the sample quantity of the target training sample in which the target student and the historical learning course exist in the initial training sample is determined;

[0043] According to the number of corresponding groups and the sample quantity, the correlation degree of the target student and the historical learning course is determined.

[0044] In the present application, according to the number of corresponding groups and the sample quantity, the correlation degree of the target student and the historical learning course is determined, comprising:

[0045] According to the number of corresponding groups and the sample quantity, the average value of the number of groups is obtained;

[0046] According to the average value of the number of groups and the sample quantity, the correlation degree of the target student and the historical learning course is determined.

[0047] In a second aspect, the present application provides a course recommendation method, comprising:

[0048] Obtaining the login student information, the login student course information and the student professional category of the platform login student;

[0049] Inputting the login student information, the login student course information and the student professional category of the platform login student into the target training model to obtain the target recommended course;

[0050] Pushing the target recommended course to the terminal device of the platform login student.

[0051] In a third aspect, the present application provides a model training device, comprising:

[0052] A first obtaining module is configured to obtain an initial training sample, wherein the initial training sample is a sample obtained from historical data information of a learning platform, and the initial training sample comprises students and courses learned by the students;

[0053] The first determining module is configured to determine, according to the initial training sample, a target course in the courses, learned information of the target course, a target learner in the learners, and learning attribute information of the target learner, wherein the learned information comprises learned time information of the target course acquired from historical data information, and the learning attribute information comprises a target professional category of the target learner and a course attribute category of a historical learning course learned by the target learner;

[0054] The second determining module is configured to determine, according to the learned time information, a course quality of the target course.

[0055] The third determining module is configured to determine, according to the target professional category and the course attribute category, an association degree between the target professional category and the historical learning course.

[0056] The optimization module is configured to optimize the initial training sample according to the course quality and the association degree, to obtain an optimized training sample.

[0057] The training module is configured to train the to-be-trained model according to the optimized training sample, to obtain a target training model.

[0058] In a fourth aspect, the present application provides a course recommendation device, comprising:

[0059] The second obtaining module is configured to obtain login learner information, login learner course information, and a learner professional category of a platform login learner.

[0060] The obtaining module is configured to input the login learner information, the login learner course information, and the learner professional category of the platform login learner into the target training model, to obtain a target recommended course.

[0061] The pushing module is configured to push the target recommended course to a terminal device of the platform login learner.

[0062] In a fifth aspect, the present application provides an electronic device, comprising a processor and a memory in communication connection with the processor.

[0063] The memory stores computer execution instructions.

[0064] The processor executes the computer execution instructions stored in the memory, to implement the method described in the present application.

[0065] In a fifth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the method described in the present application.

[0066] The model training method, the course recommendation method, the device, the equipment and the medium provided by the application, by obtaining an initial training sample, the initial training sample is a sample obtained from historical data information of a learning platform, and the initial training sample includes a student and a course learned by the student; according to the initial training sample, a target course in the course, learned information of the target course, a target student in the student and learning attribute information of the target student are determined, wherein the learned information includes learned time information of the target course obtained from the historical data information, and the learning attribute information includes a target professional category of the target student and a course attribute category of a historical learning course learned by the target student; according to the learned time information, the course quality of the target course is determined; according to the target professional category and the course attribute category, the correlation degree between the target professional category and the historical learning course is determined; according to the course quality and the correlation degree, the initial training sample is optimized to obtain an optimized training sample; and according to the optimized training sample, a to-be-trained model is trained to obtain a target training model. According to the learned time information of the target course, the number of clicks and the learning time of the target course can be determined, so as to determine the course quality of the target course, and according to the target professional category of the target student and the course attribute category of the historical learning course learned by the target student, the correlation degree between the target professional category and the historical learning course can be determined. Therefore, by optimizing the initial training sample according to the course quality and the correlation degree, data that does not meet the requirements of the course quality and the correlation degree is deleted, so that the data quality of the optimized training sample is higher, and the training effect of the to-be-trained model is improved. BRIEF DESCRIPTION OF DRAWINGS

[0067] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and serve to explain the principles of the application.

[0068] Figure 1 A flowchart of a model training method provided by an embodiment of the application;

[0069] Figure 2 A flowchart of a course recommendation method provided by an embodiment of the application;

[0070] Figure 3 A flowchart of another model training method provided by an embodiment of the application;

[0071] Figure 4 A structural diagram of a model training device provided by an embodiment of the application;

[0072] Figure 5 A structural diagram of a course recommendation device provided by an embodiment of the application;

[0073] Figure 6A structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0074] The specific embodiments of the present application have been shown through the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application by any means, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0075] Exemplary embodiments will be described in detail herein with reference to the drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments are not meant to represent all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as detailed in the appended claims.

[0076] In the prior art, when a learning platform uses a course recommendation model to recommend courses, the learning platform generally inputs historical data of the learning platform as training samples into a to-be-trained model for training, so as to obtain the course recommendation model. However, when the historical data is input as training samples into the to-be-trained model, the historical data is generally only subjected to simple deduplication processing, and is not specifically cleaned according to actual situations of course learning by students, so that the course recommendation model obtained has low course recommendation accuracy and cannot meet actual needs of students.

[0077] To solve the above problems, the present application provides a model training method, which can determine the number of clicks and the learning duration of a target course according to the learning time information of the target course, so as to determine the course quality of the target course, and determine the correlation degree between the target professional category and the historical learning course according to the target professional category of a target student and the course attribute category of the historical learning course learned by the target student. Thus, the initial training samples are optimized by the course quality and the correlation degree, and data that does not meet the requirements of the course quality and the correlation degree is deleted, so that the data quality of the optimized training samples is higher, the effect of training the to-be-trained model is improved, and the target training model obtained after training can more accurately push relevant courses to students who log in to the platform when recommending courses.

[0078] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the present application will be described below with reference to the drawings.

[0079] The execution subject of the recommendation method provided in the embodiments of the present application can be a server. The server can be a mobile phone, a tablet computer, a computer or the like. The embodiments of the present application do not particularly limit the implementation of the execution subject, as long as the execution subject can obtain an initial training sample, the initial training sample is a sample obtained from historical data information of a learning platform, and the initial training sample includes a learner and a course learned by the learner. According to the initial training sample, a target course in the course, learned information of the target course, a target learner in the learner and learning attribute information of the target learner are determined. The learned information includes learned time information of the target course obtained from the historical data information, and the learning attribute information includes a target professional category of the target learner and a course attribute category of a historical learning course learned by the target learner. According to the learned time information, the course quality of the target course is determined. According to the target professional category and the course attribute category, the correlation degree between the target professional category and the historical learning course is determined. According to the course quality and the correlation degree, the initial training sample is optimized to obtain an optimized training sample. According to the optimized training sample, a to-be-trained model is trained to obtain a target training model.

[0080] Machine learning (ML) is a multi-disciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, computational complexity theory and other subjects. Machine learning is a multi-disciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, computational complexity theory and other subjects. Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent. Machine learning is applied in various fields of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0081] Figure 1 A flowchart of a model training method provided in the embodiments of the present application is shown. The execution subject of the method can be a server or other servers, which are not particularly limited in the embodiments, for example, as shown in the figure, the method can include: Figure 1

[0082] In S101, an initial training sample is obtained, the initial training sample is a sample obtained from historical data information of a learning platform, and the initial training sample includes a learner and a course learned by the learner.

[0083] ​The initial training sample can refer to a sample obtained from historical data information of the learning platform. There can be multiple initial training samples. The initial training sample can be a sample used for model training. The initial training sample can be obtained from the historical data information. The initial training sample can include a learner and a course. The learner can include an ID of the learner. After obtaining the ID of the learner, a learner attribute of the learner and a professional talent category to which the learner belongs can be determined. The course can be a course learned by the learner. In the embodiments of the present application, the course learned by the learner includes at least one lesson.

[0084] The historical data information can be historical data of the learner who logs in to the learning platform for learning. The historical data information can include learning behavior attributes, learner attributes, course attributes, and professional talent categories. In the embodiments of the present application, the learning behavior attributes can include a learner ID (Identity document), a learning time, and a learning course ID. The learner attributes can include a learner ID, a learner attribute 1, a learner attribute 2, a learner attribute 3, and a professional talent category number. The course attributes can include a course ID, a course attribute 1, a course attribute 2, a course attribute 3, and the like. The professional talent category can include a professional talent category number, a professional talent category name, a professional talent category keyword 1, a professional talent category keyword 2, a professional talent category keyword 3, and the like.

[0085] In the embodiments of the present application, the method for obtaining the initial training sample can include:

[0086] Obtaining historical data information;

[0087] According to the historical data information, determining a to-be-trained sample, the to-be-trained sample including learner information, course information, and a learner professional category;

[0088] According to the learner professional category, determining a first target to-be-trained sample, wherein the learner professional category in the first target to-be-trained sample is a professional category;

[0089] According to the first target to-be-trained sample, obtaining the initial training sample.

[0090] The to-be-trained sample can include "student ID, student attribute 1, student attribute 2, student attribute 3,..., learning time, learning course ID, professional talent category number". In some embodiments, after obtaining the historical data information, "learning behavior category attribute, student category attribute, course category attribute and professional talent category" can be filled into "student ID, student attribute 1, student attribute 2, student attribute 3,..., learning time, learning course ID, professional talent category number", thereby obtaining the initial training sample. For example, the initial training sample can be "Zhang San, new employee, communication engineering major, 24 years old, master, wireless communication post, March 30, 2021, 9:00-10:00, cloud computing technology course ID, 100223".

[0091] The student information can refer to "student ID, student attribute 1, student attribute 2, student attribute 3,...".

[0092] The course information can refer to "learning time, learning course ID".

[0093] The student professional category can refer to "professional talent category" of the student, wherein the professional talent category can be a student professional category determined according to the student's major or post. In the embodiments of the present application, the professional talent category can include a professional category, a non-professional category and a potential professional category, wherein the students of the professional category can refer to students with clear major or post; the students of the non-professional category can refer to students whose major or post cannot be determined; the students of the potential professional category can refer to students whose major or post cannot be determined, but whose student information and course information have high similarity by comparing the student information and course information of the students of the professional category.

[0094] The first target to-be-trained sample can be a to-be-trained sample whose student professional category is a professional category.

[0095] In the embodiments of the present application, the method for obtaining the initial training sample according to the first target to-be-trained sample can include:

[0096] According to the first target to-be-trained sample, a second target to-be-trained sample is determined from the to-be-trained sample, wherein the student information in the second target to-be-trained sample meets the preset student information similarity requirement with the student information in the first target to-be-trained sample, and the course information in the second target to-be-trained sample meets the preset course information similarity requirement with the course information in the first target to-be-trained sample;

[0097] According to the first target to-be-trained sample and the second target to-be-trained sample, the initial training sample is obtained.

[0098] The second target training sample is a training sample in the initial training sample whose learner professional category is the potential professional category. In the embodiment of the present application, after the first target training sample is determined, the second target training sample in the initial training sample can be determined according to the learner information and the course information in the first target training sample. The preset similarity requirement can refer to the similarity of the learner information and the similarity of the course information. For example, if the text similarity of the learner information in the initial training sample and the learner information in the first target training sample meets the preset similarity requirement, and the text similarity of the course information in the initial training sample and the course information in the first target training sample meets the preset similarity requirement, the initial training sample can be determined as the second target training sample. The method for determining the text similarity can include determining by a natural language model.

[0099] After the first target training sample and the second target training sample are obtained, the initial training sample can be obtained.

[0100] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the embodiments of the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation portal for user to choose authorization or refusal.

[0101] S102, determining a target course in the course, learned information of the target course, a target learner in the learners, and learning attribute information of the target learner according to the initial training sample, wherein the learned information includes learned time information of the target course acquired from historical data information, and the learning attribute information includes a target professional category of the target learner and a course attribute category of a historical learning course learned by the target learner.

[0102] The target course can be any course selected from the initial training sample. The learned information of the target course can be acquired from the historical data information according to the ID of the target course.

[0103] The target learner can be any learner selected from the initial training sample. The learning attribute information of the target learner can be acquired from the historical data information according to the ID of the target learner.

[0104] The target professional category can refer to a category corresponding to the professional category of the student. In the embodiments of the present application, the target professional category can be obtained by extracting the keywords in the information base corresponding to the professional category of the student. For example, if the student is a major of information and communication engineering, the target professional category can include "communication" and "multimedia".

[0105] The historical learning course can be all courses learned by the target student. The course attribute category can refer to the classification of the historical learning course. For example, if the classification of the historical learning course is the communication category, the learning attribute information can be "communication". In the embodiments of the present application, the course attribute category can be obtained by extracting the keywords in the course information database corresponding to the course attribute category. The database can be stored in the historical data information of the learning platform.

[0106] S103, determining the course quality of the target course according to the learned time information.

[0107] The learned time information can refer to the learned time of the target course.

[0108] The course quality can represent the length and frequency of learning the target course by the user. The longer the length and the higher the frequency, the better the course quality. In the embodiments of the present application, the course quality of the target course can be determined by the learned frequency and the average learned length.

[0109] In the embodiments of the present application, the method for determining the course quality of the target course according to the learned time information can include:

[0110] determining the learned frequency, the average course completion rate and the average learned length of the target course according to the learned time information;

[0111] obtaining the historical learned time information of each course from the historical data information;

[0112] determining the course quality of the target course according to the learned frequency, the average course completion rate, the average learned length and the historical learned time information.

[0113] The average course completion rate can refer to the progress of learning the target course. For example, if the learning length of the target course is 60 minutes, and the average learned length of the target course is 30 minutes, the average course completion rate is 50%.

[0114] The historical learned time information can refer to the learned time of each course.

[0115] In the embodiments of the present application, the method for determining the course quality of the target course according to the learned number, the average course completion rate, the average learned duration and the historical learned time information can include:

[0116] According to the historical learned time information, the historical learned number, the historical average course completion rate, the historical average learned duration, the target historical learned number, the target historical average course completion rate and the target historical average learned duration are determined, wherein the target historical learned number is an average value of the historical learned numbers greater than the preset number among the historical learned numbers, the target historical average course completion rate is an average value of the historical average course completion rates greater than the preset completion rate among the historical average course completion rates, and the target historical average learned duration is an average value of the historical average learned durations greater than the preset duration among the historical average learned durations.

[0117] According to the learned number, the average course completion rate, the average learned duration, the historical learned number, the historical average course completion rate, the historical average learned duration, the target historical learned number, the target historical average course completion rate and the target historical average learned duration, the course quality of the target course is determined.

[0118] The historical average course completion rate can refer to the progress of learning the course.

[0119] In the embodiments of the present application, the method for determining the course quality of the target course according to the learned number, the average course completion rate, the average learned duration, the historical learned number, the historical average course completion rate, the historical average learned duration, the target historical learned number, the target historical average course completion rate and the target historical average learned duration can include:

[0120] A first index result is obtained according to the learned number and the historical average learned number;

[0121] A second index result is obtained according to the average course completion rate and the historical average course completion rate;

[0122] A third index result is obtained according to the average learned duration and the historical average learned duration;

[0123] A fourth index result is obtained according to the learned number and the target historical learned number;

[0124] A fifth index result is obtained according to the average course completion rate and the target historical average course completion rate;

[0125] A sixth index result is obtained according to the average learned duration and the target historical average learned duration;

[0126] According to the first index result, the second index result, the third index result, the fourth index result, the fifth index result and the sixth index result, the course quality of the target course is determined.

[0127] The method for determining the course quality of the target course can be determined by the formula:

[0128]

[0129] determined.

[0130] ST j The number of times of learning the target course can be the number of times of learning the target course, The historical average number of times of learning all courses can be the historical average number of times of learning all courses, The first index result can be 1 when the number of times of learning the target course is greater than the historical average number of times of learning all courses;

[0131] The fourth index result can be the target historical number of times of learning greater than the preset number of times threshold in the historical number of times of learning all courses, wherein the preset number of times threshold can be the top 30% of the number of times of learning all historical courses, and the fourth index result is 1 when the number of times of learning the target course is greater than the historical average number of times of learning the target course;

[0132] CST j The average number of times of learning the target course can be the average number of times of learning the target course, CT j The course duration of the target course can be the course duration of the target course, The average course completion rate of the target course can be the average course completion rate of the target course, The historical average course completion rate of all courses can be the historical average course completion rate of all courses, The second index result can be 1 when the average course completion rate of the target course is greater than the historical average course completion rate of all courses;

[0133] The fourth index result can be the target historical number of times of learning greater than the preset completion rate threshold in the historical course completion rate, wherein the preset completion rate threshold can be the top 20% of the number of times of learning all historical courses, and the fourth index result is 1 when the average course completion rate of the target course is greater than the historical average course completion rate of the target course;

[0134] CTST j The average learning duration of the target course can be the average learning duration of the target course, The historical average learning duration of all courses can be the historical average learning duration of all courses, and the third index result is 1 when the average learning duration of the target course is greater than the historical average learning duration of all courses;

[0135] The target historical average learning duration can refer to a historical average learning duration greater than a preset duration threshold in all historical average learning durations. The preset duration threshold can be a preset time of the top 40% of all historical average learning durations. When the average learning duration of the target course is greater than the target historical average learning duration, the fifth index result is 1.

[0136] a, β and γ are constant coefficients, which can take default values, for example, a, β and γ can all be 1.

[0137] In the embodiments of the present application, when f1 j ≥ 3, the course quality of the target course meets the quality requirement, and when f1 j < 3, it indicates that the course quality of the target course does not meet the quality requirement.

[0138] In S104, the association degree between the target major category and the historical learning course is determined according to the target major category and the course attribute category.

[0139] The association degree can represent whether the target major category of the student is associated with the classification of the historical learning course. In the embodiments of the present application, the method of determining the association degree between the target major category and the historical learning course according to the target major category and the course attribute category can include:

[0140] The target major category and the course attribute category are compared to obtain a comparison result.

[0141] The association degree between the target major category and the historical learning course is determined according to the comparison result.

[0142] The comparison result can represent whether the target major category and the course attribute category match. In the embodiments of the present application, when the target major category and the course attribute category have the same keyword, it can be determined that the comparison result is a match. For example, when the target major category and the course attribute category are both communication, the target major category and the course attribute category match.

[0143] In the embodiments of the present application, the method of determining the association degree between the target major category and the historical learning course according to the comparison result can include:

[0144] If the comparison result indicates that the target major category and the course attribute category match, the association degree between the target major category and the historical learning course is determined according to the comparison result.

[0145] In the embodiments of the present application, the method of determining the association degree between the target major category and the historical learning course according to the comparison result can include:

[0146] If the comparison result represents that the target professional category and the course attribute category are not matched, according to the comparison result, a corresponding group number of the student and the course corresponding to each other in the initial training sample and a sample number of the target training sample in which the target student and the historical learning course exist in the initial training sample are determined;

[0147] According to the corresponding group number and the sample number, the correlation degree of the target student and the historical learning course is determined.

[0148] The corresponding group number can refer to a group number of one student and one course corresponding to the student appearing in the initial training sample.

[0149] In the embodiments of the present application, the method for determining the correlation degree of the target student and the historical learning course according to the corresponding group number and the sample number can include:

[0150] According to the corresponding group number and the sample number, a group number average value is obtained.

[0151] According to the group number average value and the sample number, the correlation degree of the target student and the historical learning course is determined.

[0152] When the comparison result is that the target professional category and the course attribute category are not matched, the method for determining the correlation degree of the target student and the historical learning course can be through the formula:

[0153] determination.

[0154] Y k,j may be a historical learning course representing the course attribute category. When the historical learning course appears in the initial training sample k, Y k,j is 1, and when the historical learning course does not appear in the initial training sample k, Y k,j is 0.

[0155] Z k,r may be a target professional category of the target student. When the target professional category appears in the initial training sample k, Z k,r is 1, and when the target professional category does not appear in the initial training sample k, Z k,r is 0.

[0156] may refer to the sample number of the target training sample.

[0157] T1may be a pre-set number threshold. When the sample number of the target training sample is greater than T1,

[0158] The target professional category and the average number of groups of historical learning courses in the course representing the course attribute category can be referred to as the professional category of the student and the course attribute category. A constant coefficient δ is used to adjust the proportion size, and in the embodiment of the present application, δ can be set to 1.

[0159] In the embodiment of the present application, when f2 j,r When f2 j,r When f2≥1, the correlation degree of the target student and the historical learning course meets the requirement.

[0160] In the embodiment of the present application, the target student in one initial training sample can correspond to multiple historical learning courses. When there is at least one group of target students and historical learning courses with a correlation degree meeting the requirement in one initial training sample, the initial training sample is represented as a training sample with a correlation degree meeting the requirement.

[0161] S105, optimizing the initial training sample according to the course quality and the correlation degree to obtain an optimized training sample.

[0162] The optimization can refer to a processing process of retaining or deleting the target course and the target student according to the course quality and the correlation degree. For example, when the course quality of the target course does not meet the quality requirement, the initial training sample with the target course can be deleted, or all courses in the initial training sample set can be determined for course quality, and all target courses with course quality not meeting the quality requirement can be deleted. When the correlation degree of the target professional category and the historical learning course does not meet the correlation degree requirement, the initial training sample set with the target professional category and the historical learning course can be deleted, or the correlation degree of all professional categories and historical learning courses in the initial training sample set can be determined, so that all initial training samples with the target professional category and the historical learning course can be deleted. In this way, the optimized training sample can be obtained.

[0163] S106, training the to-be-trained model according to the optimized training sample to obtain a target training model.

[0164] The to-be-trained model can be a neural network model. The student, student attribute, learning time and professional category in the optimized training sample are used as input items, and the course is used as a label to train the to-be-trained model. The to-be-trained model continuously adjusts the weights of each input item until the output result of the to-be-trained model meets the expected requirement. In this way, the target training model can be obtained.

[0165] In the embodiment of the present application, the target training model can be trained by inputting the student attributes and learning courses into the data mining software IBM SPSS Statistics.

[0166] The model training method provided in the present application can determine the number of clicks and the learning duration of the target course according to the learning time information of the target course, thereby determining the course quality of the target course, and determining the correlation degree of the target professional category and the historical learning courses according to the target professional category of the target student and the course attribute category of the historical learning courses learned by the target student. In this way, the initial training sample is optimized by the course quality and the correlation degree, and the data that does not meet the requirements of the course quality and the correlation degree is deleted, so that the data quality of the optimized training sample is higher, and the training effect of the to-be-trained model is improved.

[0167] Figure 2 The flowchart of the course recommendation method provided in the embodiment of the present application. The execution subject of the method can be a server or other server, and the embodiment does not make special limitation here, such as Figure 2 The method can include:

[0168] S201, obtaining login student information, login student course information and student professional category of a platform login student;

[0169] S202, inputting the login student information, the login student course information and the student professional category of the platform login student into a target training model to obtain a target recommended course;

[0170] S203, pushing the target recommended course to a terminal device of the platform login student.

[0171] The platform login student can refer to a student who logs in a learning platform. After the student logs in the platform, the learning platform can obtain the student information, the login student course information and the student professional category of the platform login student after authorization of the student, input the student information, the login student course information and the student professional category into the target training model, obtain the target recommended course through the target training model, and push the target recommended course to the terminal device of the platform login student.

[0172] Figure 3 The flowchart of another model training method provided in the embodiment of the present application. The execution subject of the method can be a computer, and the embodiment does not make special limitation here, such as Figure 3 The method can include:

[0173] S301, obtaining historical learning data of a known professional talent list in a learning platform.

[0174] wherein, the historical learning data comprises: learning behavior attribute R1={learner unique identifier ID, learning time, learning course unique identifier}; learner attribute R2={learner unique identifier ID, learner attribute 1, learner attribute 2, learner attribute 3, …, professional talent category number}; course attribute R3={course unique identifier ID, course attribute 1, course attribute 2, course attribute 3, …}; and professional talent category R4={professional talent category number, professional talent category name, professional talent category keyword 1, professional talent category keyword 2, professional talent category keyword 3, …}.

[0175] In the embodiment of the present application, the professional talent category R4 can be determined according to the registration information of the learner in the learning platform, and the historical learning data of the known professional talent list in the learning platform can be obtained by screening the registration information of the learner, determining the known professional talent, and obtaining the corresponding historical learning data through the known professional talent.

[0176] S302, obtaining the initial training data set of the known professional talent according to the historical learning data.

[0177] wherein, the initial training data of the known professional talent can be obtained according to the learning behavior attribute R1, the learner attribute R2, the course attribute R3 and the professional talent category R4.

[0178] In the embodiment of the present application, the initial training data R0={learner ID, learner attribute 1, learner attribute 2, learner attribute 3, …, learning time, learning course ID, professional talent category number}. Wherein, the independent variables in the initial training data R0 can be set as: learner attribute 1, learner attribute 2, learner attribute 3, …, learning time, learning course ID, and the dependent variable can be set as: professional talent category number. Wherein, the professional talent category number can represent the category of the professional talent, for example, when the professional talent category is the known professional talent, the first digit of the professional talent category number is “1”; when the professional talent category is the potential professional talent, the first digit of the professional talent category number is “2”; when the professional talent category is the non-professional talent, the professional talent category number is empty.

[0179] Filling the learning behavior attribute R1, the learner attribute R2, the course attribute R3 and the professional talent category R4 into R0 can obtain the initial training data X1 of the known professional talent, and the initial training data set of the known professional talent can be obtained by a plurality of initial training data, therefore, the initial training data set X of the known professional talent={X1, X3, X3, …, Xi, …, XI}.

[0180] S303, determining the potential professional talent and the initial training data set of the potential professional talent according to the initial training data set of the known professional talent.

[0181] S304. Based on the initial training dataset of known professionals and the initial training dataset of potential professionals, obtain the initial training samples.

[0182] The independent variables of the initial training samples can be: student attribute 1, student attribute 2, student attribute 3, ..., learning time, professional talent category number, and the dependent variables can be: learning course ID1, learning course ID2, ..., learning course IDN.

[0183] S305. Based on the initial training samples, obtain the initial target course set, which includes one or more initial target courses.

[0184] The initial target course set can be a set of initial target courses after filtering the learning courses extracted from the initial training samples.

[0185] Filtering learning courses can refer to the process of eliminating duplicate learning courses.

[0186] S306. Identify the target courses that meet the requirements in the initial target courses.

[0187] Among them, "meeting the requirements" can refer to courses that meet both quality and relevance requirements.

[0188] Methods for determining whether an initial target course meets quality requirements include determining the course's learning rate and learning duration indicators. In this embodiment, this can be achieved using the formula:

[0189]

[0190] Among them, the initial target course is selected as F1. j Courses with ≥3 hours of instruction.

[0191] One method to determine whether an initial target course meets the relevance requirement is the relevance between the course and the major category. In this embodiment, this can be achieved using the formula:

[0192] Sure.

[0193] Among them, f2 is selected as the initial target course. j,r Courses with a minimum score of 1.

[0194] In this embodiment of the application, when selecting the initial target course f1 j Courses with ≥3 and initial target courses f2 j,r After identifying courses with a score of ≥1, you can determine the target courses as those that simultaneously meet both the quality and relevance requirements, or you can select courses that meet both the quality and relevance requirements separately.

[0195] S307, determine a target training sample according to the target course.

[0196] The target training sample includes the target course. After the target course is determined, the target training sample including the target course can be determined in the initial training data set of the known professional talent and the initial training data set of the potential professional talent.

[0197] S308, input the target training sample into a model to be trained to perform model training, and obtain a course recommendation model.

[0198] In the embodiment of the present application, "student attribute 1, student attribute 2, student attribute 3, …, learning time, professional talent category number" can be used as "independent variables", and "target course" can be used as "label" to input into the model for training, and the course recommendation model is obtained.

[0199] In the embodiment of the present application, the target training sample can also be input into the data mining software to obtain the course recommendation model, and the data mining software can be IBM SPSS Statistics.

[0200] After the course recommendation model is obtained, the learning platform can recommend courses to the student according to the student attributes of the student.

[0201] Another model training method provided by the embodiment of the present application can optimize the data based on the characteristics of the historical learning data, combine the data recording time, duration and professional talent type of the student, and eliminate invalid data and low-quality data in the historical learning data, so that the data quality for model training is higher, thereby improving the effect of model training.

[0202] Figure 4 The structure diagram of the model training device provided by the embodiment of the present application is shown in FIG. 4. As shown in FIG. 4, the model training device 40 includes a first acquisition module 401, a first determination module 402, a second determination module 403, a third determination module 404, an optimization module 405 and a training module 406. Wherein: Figure 4

[0203] The first acquisition module 401 is configured to acquire an initial training sample. The initial training sample is a sample acquired from the historical data information of the learning platform, and the initial training sample includes a student and a course learned by the student.

[0204] ​The first determining module 402 is configured to determine, according to the initial training sample, a target course in the courses, learned information of the target course, a target student in the students, and learning attribute information of the target student, wherein the learned information comprises learned time information of the target course acquired from historical data information, and the learning attribute information comprises a target professional category of the target student and a course attribute category of a historical learning course learned by the target student.

[0205] The second determining module 403 is configured to determine, according to the learned time information, a course quality of the target course.

[0206] The third determining module 404 is configured to determine, according to the target professional category and the course attribute category, an association degree between the target professional category and the historical learning course.

[0207] The optimization module 405 is configured to optimize the initial training sample according to the course quality and the association degree, to obtain an optimized training sample.

[0208] The training module 406 is configured to train the to-be-trained model according to the optimized training sample, to obtain a target training model.

[0209] In the embodiments of the present application, the first obtaining module 401 can be specifically configured to:

[0210] obtain historical data information;

[0211] determine, according to the historical data information, a to-be-trained sample, wherein the to-be-trained sample comprises student information, course information and a professional category of a student;

[0212] determine a first target to-be-trained sample according to the professional category of the student, wherein the professional category of the student in the first target to-be-trained sample is the professional category;

[0213] obtain the initial training sample according to the first target to-be-trained sample.

[0214] In the embodiments of the present application, the first obtaining module 401 can be specifically configured to:

[0215] determine, according to the first target to-be-trained sample, a second target to-be-trained sample from the to-be-trained sample, wherein the student information in the second target to-be-trained sample meets a preset student information similarity requirement with the student information in the first target to-be-trained sample, and the course information in the second target to-be-trained sample meets a preset course information similarity requirement with the course information in the first target to-be-trained sample;

[0216] obtain the initial training sample according to the first target to-be-trained sample and the second target to-be-trained sample.

[0217] In the embodiments of the present application, the second determining module 403 can be specifically used for:

[0218] According to the learned time information, the learned number of times, the average course completion rate, and the average learned duration of the target course are determined.

[0219] The historical learned time information of each course is obtained from the historical data information.

[0220] According to the learned number of times, the average course completion rate, the average learned duration, and the historical learned time information, the course quality of the target course is determined.

[0221] In the embodiments of the present application, the second determining module 403 can be specifically used for:

[0222] According to the historical learned time information, the historical learned number of times, the historical average course completion rate, the historical average learned duration, the target historical learned number of times, the target historical average course completion rate, and the target historical average learned duration are determined, wherein the target historical learned number of times is the average value of the historical learned number of times greater than the preset number of times in the historical learned number of times, the target historical average course completion rate is the average value of the historical average course completion rate greater than the preset completion rate in the historical average course completion rate, and the target historical average learned duration is the average value of the historical average learned duration greater than the preset duration in the historical average learned duration.

[0223] According to the learned number of times, the average course completion rate, the average learned duration, the historical learned number of times, the historical average course completion rate, the historical average learned duration, the target historical learned number of times, the target historical average course completion rate, and the target historical average learned duration, the course quality of the target course is determined.

[0224] In the embodiments of the present application, the second determining module 403 can be specifically used for:

[0225] According to the learned number of times and the historical average learned number of times, a first index result is obtained.

[0226] According to the average course completion rate and the historical average course completion rate, a second index result is obtained.

[0227] According to the average learned duration and the historical average learned duration, a third index result is obtained.

[0228] According to the learned number of times and the target historical learned number of times, a fourth index result is obtained.

[0229] According to the average course completion rate and the target historical average course completion rate, a fifth index result is obtained.

[0230] According to the average learned duration and the target historical average learned duration, a sixth index result is obtained.

[0231] According to the first index result, the second index result, the third index result, the fourth index result, the fifth index result and the sixth index result, the course quality of the target course is determined.

[0232] In the embodiments of the present application, the third determining module 404 can be specifically used for:

[0233] comparing the target major category and the course attribute category to obtain a comparison result;

[0234] According to the comparison result, the association degree between the target major category and the historical learning course is determined.

[0235] In the embodiments of the present application, the third determining module 404 can be specifically used for:

[0236] If the comparison result represents that the target major category and the course attribute category match, then according to the comparison result, the association degree between the target major category and the historical learning course is determined.

[0237] In the embodiments of the present application, the third determining module 404 can be specifically used for:

[0238] If the comparison result represents that the target major category and the course attribute category do not match, then according to the comparison result, the number of corresponding groups of the students and the courses corresponding to each other in the initial training sample and the number of samples of the target training sample in which the target student and the historical learning course exist in the initial training sample are determined.

[0239] According to the number of corresponding groups and the number of samples, the association degree between the target student and the historical learning course is determined.

[0240] In the embodiments of the present application, the third determining module 404 can be specifically used for:

[0241] According to the number of corresponding groups and the number of samples, a group number average value is obtained.

[0242] According to the group number average value and the number of samples, the association degree between the target student and the historical learning course is determined.

[0243] From the above, the model training device of the embodiment comprises a first obtaining module 401, which is configured to obtain initial training samples, the initial training samples being samples obtained from historical data information of a learning platform, and the initial training samples comprising students and courses learned by the students; a first determining module 402, which is configured to determine, according to the initial training samples, a target course in the courses, learned information of the target course, a target student in the students, and learning attribute information of the target student, wherein the learned information comprises learned time information of the target course obtained from the historical data information, and the learning attribute information comprises a target professional category of the target student and a course attribute category of a historical learning course learned by the target student; a second determining module 403, which is configured to determine, according to the learned time information, a course quality of the target course; a third determining module 404, which is configured to determine, according to the target professional category and the course attribute category, an association degree between the target professional category and the historical learning course; an optimization module 405, which is configured to optimize, according to the course quality and the association degree, the initial training samples to obtain optimized training samples; and a training module 406, which is configured to train, according to the optimized training samples, a to-be-trained model to obtain a target training model. In this way, the effect of model training is improved.

[0244] Figure 5 A structural schematic diagram of a course recommendation device provided by an embodiment of the present application is shown in FIG. 5. As shown in FIG. 5, the course recommendation device 50 comprises a second obtaining module 501, a obtaining module 502, and a pushing module 503. Figure 5

[0245] Among them:

[0246] The second obtaining module 501 is configured to obtain login student information, login student course information, and student professional categories of platform login students.

[0247] The obtaining module 502 is configured to input the login student information, the login student course information, and the student professional categories of the platform login students into the target training model to obtain a target recommended course.

[0248] The pushing module 503 is configured to push the target recommended course to a terminal device of the platform login student.

[0249] From the above, the course recommendation device of the embodiment comprises the second obtaining module 501, which is configured to obtain login student information, login student course information, and student professional categories of platform login students; the obtaining module 502, which is configured to input the login student information, the login student course information, and the student professional categories of the platform login students into the target training model to obtain a target recommended course; and the pushing module 503, which is configured to push the target recommended course to a terminal device of the platform login student. In this way, the recommended course is more accurate. ​

[0250] Figure 6 A structural schematic diagram of an electronic device provided in an embodiment of the present application is shown in FIG. 6. As shown in the figure, the electronic device 60 includes: Figure 6

[0251] The electronic device 60 can include a processor 601 with one or more processing cores, a memory 602 with one or more computer-readable storage media, a communication component 603, and the like. The processor 601, the memory 602, and the communication component 603 are connected through a bus 604.

[0252] In the implementation process, the at least one processor 601 executes the computer-executable instructions stored in the memory 602, so that the at least one processor 601 performs the model training method and the course recommendation method as described above.

[0253] The specific implementation process of the processor 601 can refer to the method embodiments described above, which have similar implementation principles and technical effects. Therefore, the details are not described here again.

[0254] In the above-described Figure 6 In the embodiments shown in the above-described

[0255] The memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory.

[0256] ​The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application does not limit to only one bus or one type of bus.

[0257] In some embodiments, a computer program product is also provided, including a computer program or instructions, which, when executed by a processor, implement the steps of any of the model training method or the course recommendation method described above.

[0258] The specific implementation of each operation above can refer to the previous embodiments, which will not be repeated here.

[0259] Those of ordinary skill in the art can understand that all or part of the steps of the various methods of the above embodiments can be completed by instructions, or by relevant hardware controlled by the instructions, which can be stored in a computer readable storage medium and loaded and executed by a processor.

[0260] To this end, an embodiment of the present application provides a computer readable storage medium, which stores a plurality of instructions. The instructions can be loaded by a processor to execute the steps of any of the model training method or the course recommendation method provided by the embodiments of the present application.

[0261] The storage medium can include a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, etc.

[0262] According to an aspect of the present application, a computer program product or a computer program is provided, which includes computer instructions stored in a computer readable storage medium.

[0263] Since the instructions stored in the storage medium can execute the steps of any of the model training method or the course recommendation method provided by the embodiments of the present application, the beneficial effects that can be achieved by any of the model training method or the course recommendation method provided by the embodiments of the present application can be achieved. Details are described in the previous embodiments, which will not be repeated here.

[0264] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0265] It is to be understood that the application is not limited to the precise construction herein disclosed and shown in the drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is limited only by the claims that follow.

Claims

1. A model training method, characterized in that, Applied to a learning platform, the method includes: Obtain initial training samples, which are samples obtained from the historical data information of the learning platform. The initial training samples include students and courses that the students have studied. Based on the initial training samples, the target courses in the course, the learning information of the target courses, the target students among the students, and the learning attribute information of the target students are determined. The learning information includes the learning time information of the target courses obtained from the historical data information, and the learning attribute information includes the target major category of the target students and the course attribute categories of the historical courses studied by the target students, obtained from the historical data information. Based on the learning time information, the course quality of the target courses is determined. Based on the target major category and the course attribute category, determine the correlation between the target major category and the history learning course; Based on the course quality and the relevance, the initial training samples are optimized to obtain optimized training samples; Based on the optimized training samples, the model to be trained is trained to obtain the target training model, which is used for course recommendation. The step of determining the correlation between the target major category and the history learning course based on the target major category and the course attribute category includes: The comparison results are obtained by comparing the target major category and the course attribute category; If the comparison result indicates that the target major category and the course attribute category match, then the correlation between the target major category and the history learning course is determined based on the comparison result. If the comparison result indicates that the target major category and the course attribute category do not match, then based on the comparison result, determine the number of corresponding groups of students and courses in the initial training sample, and the number of target training samples in the initial training sample that contain the target students and the historical learning courses; The average number of groups is obtained based on the corresponding group number and the sample size; The correlation between the target student and the historical learning course is determined based on the average number of groups and the sample size. When the comparison result shows that the target major category and the course attribute category do not match, the method to determine the correlation between the target students and the historical learning courses can be expressed by the formula: Sure; Among them, Y k,j The function represents whether a historical learning course, which is a category of course attributes, appears in the initial training sample k. If a historical learning course appears in the initial training sample k, then Y... k,j If Y is 1, then Y is 1 when the history learning course does not appear in the initial training sample k. k,j =0; Z k,r Z represents whether the target major category of the target learner appears in the initial training sample k. If the target major category appears in the initial training sample k, then Z... k,r Z is 1 when the target specialty category does not appear in the initial training sample k. k,r =0; The number of samples in the target training dataset; T1 is a pre-set threshold value, meaning that when the number of target training samples exceeds T1, =1; The average number of groups of courses representing the target major category and the history learning courses in the courses representing the course attributes of the students, where δ is a constant coefficient used to adjust the proportion; According to the preset settings, when f2 j,r When f2 < δ, the relevance between the target learners and the history course does not meet the requirements. j,r When the value is ≥δ, the relevance between the target students and the history learning course meets the requirements.

2. The method according to claim 1, characterized in that, The process of obtaining the initial training samples includes: Obtain historical data information; Based on the historical data, a training sample is determined, which includes student information, course information, and student major category. Based on the student's major category, a first target training sample is determined, wherein the student's major category in the first target training sample is the major category. Based on the first target training sample, the initial training sample is obtained.

3. The method according to claim 2, characterized in that, The step of obtaining the initial training samples based on the first target training samples includes: Based on the first target training sample, a second target training sample is determined from the training sample, wherein the student information in the second target training sample and the student information in the first target training sample meet the preset student information similarity requirement, and the course information in the second target training sample and the course information in the first target training sample meet the preset course information similarity requirement. The initial training samples are obtained based on the first target training sample and the second target training sample.

4. The method according to claim 1, characterized in that, Determining the course quality of the target course based on the learned time information includes: Based on the learning time information, determine the number of times the target course is learned, the average course completion rate, and the average learning duration. Obtain the historical learning time information for each course from the historical data information; The course quality of the target course is determined based on the number of times it is studied, the average course completion rate, the average study duration, and the historical study time information.

5. The method according to claim 4, characterized in that, Determining the course quality of the target course based on the number of times it has been studied, the average course completion rate, the average study duration, and the historical study time information includes: Based on the historical learning time information, the following are determined: historical learning count, historical average course completion rate, historical average learning duration, target historical learning count, target historical average course completion rate, and target historical average learning duration. The target historical learning count is the average of the historical learning counts that are greater than a preset count; the target historical average course completion rate is the average of the historical average course completion rates that are greater than a preset completion rate; and the target historical average learning duration is the average of the historical average learning durations that are greater than a preset duration. The course quality of the target course is determined based on the number of times it is studied, the average course completion rate, the average study duration, the historical number of times it is studied, the historical average course completion rate, the historical average study duration, the target historical number of times it is studied, the target historical average course completion rate, and the target historical average study duration.

6. The method according to claim 5, characterized in that, The step of determining the course quality of the target course based on the number of times the course is studied, the average course completion rate, the average study duration, the historical number of times the course is studied, the historical average course completion rate, the historical average study duration, the target historical number of times the course is studied, the target historical average course completion rate, and the target historical average study duration includes: The first indicator result is obtained based on the number of times it has been learned and the historical average number of times it has been learned. The second indicator result is obtained based on the average course completion rate and the historical average course completion rate; The third indicator result is obtained based on the average learning time and the historical average learning time. The fourth indicator result is obtained based on the number of times the target has been learned and the number of times the target has been learned in history; The fifth indicator result is obtained based on the average course completion rate and the target historical average course completion rate; The sixth indicator result is obtained based on the average learning time and the historical average learning time of the target. The course quality of the target course is determined based on the results of the first indicator, the second indicator, the third indicator, the fourth indicator, the fifth indicator, and the sixth indicator.

7. A course recommendation method, characterized in that, The method includes: Obtain the login student information, login student course information, and student major category of the logged-in students on the platform; The login student information, login student course information, and student major category of the logged-in students on the platform are input into the target training model trained by the method described in claims 1 to 6 to obtain the target recommended courses. The target recommended courses are pushed to the terminal devices of students logged into the platform.

8. A model training device, characterized in that, include: The first acquisition module is used to acquire initial training samples, which are samples obtained from historical data information of the learning platform. The initial training samples include students and courses that the students have studied. The first determining module is used to determine, based on the initial training samples, the target course in the course, the learning information of the target course, the target student in the students, and the learning attribute information of the target student, wherein the learning information includes the learning time information of the target course obtained from the historical data information, and the learning attribute information includes the target major category of the target student and the course attribute category of the historical learning courses studied by the target student obtained from the historical data information. The second determining module is used to determine the course quality of the target course based on the learning time information; The third determining module is used to determine the correlation between the target major category and the history learning course based on the target major category and the course attribute category; An optimization module is used to optimize the initial training samples based on the course quality and the relevance to obtain optimized training samples. The training module is used to train the model to be trained based on the optimized training samples to obtain a target training model, which is used for course recommendation. The step of determining the correlation between the target major category and the history learning course based on the target major category and the course attribute category includes: The comparison results are obtained by comparing the target major category and the course attribute category; If the comparison result indicates that the target major category and the course attribute category match, then the correlation between the target major category and the history learning course is determined based on the comparison result. If the comparison result indicates that the target major category and the course attribute category do not match, then based on the comparison result, determine the number of corresponding groups of students and courses in the initial training sample, and the number of target training samples in the initial training sample that contain the target students and the historical learning courses; The average number of groups is obtained based on the corresponding group number and the sample size; The correlation between the target student and the historical learning course is determined based on the average number of groups and the sample size. When the comparison result shows that the target major category and the course attribute category do not match, the method to determine the correlation between the target students and the historical learning courses can be expressed by the formula: Sure; Among them, Y k,j The function represents whether a historical learning course, which is a category of course attributes, appears in the initial training sample k. If a historical learning course appears in the initial training sample k, then Y... k,j If Y is 1, then Y is 1 when the history learning course does not appear in the initial training sample k. k,j =0; Z k,r Z represents whether the target major category of the target learner appears in the initial training sample k. If the target major category appears in the initial training sample k, then Z... k,r Z is 1 when the target specialty category does not appear in the initial training sample k. k,r =0; The number of samples in the target training dataset; T1 is a pre-set threshold value, meaning that when the number of target training samples exceeds T1, =1; The average number of groups of courses representing the target major category and the history learning courses in the courses representing the course attributes of the students, where δ is a constant coefficient used to adjust the proportion; According to the preset settings, when f2 j,r When f2 < δ, the relevance between the target learners and the history course does not meet the requirements. j,r When the value is ≥δ, the relevance between the target students and the history learning course meets the requirements.

9. A course recommendation device, characterized in that, include: The second acquisition module is used to acquire the login student information, login student course information, and student major category of the logged-in students on the platform; The module is used to input the login student information, login student course information and student major category of the logged-in student of the platform into the target training model trained by the method described in claims 1 to 6, so as to obtain the target recommended courses; The push module is used to push the target recommended courses to the terminal devices of students logged into the platform.

10. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 7.

Images