Teaching course recommendation method and system based on big data

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a teaching course recommendation method and system based on big data, and the method comprises the steps: obtaining the first recommendability of a target course according to the user number of the target course and the coordinate distribution range of a learning record; obtaining second recommendability of the target course according to the linear change of the user number of the target course included in the learning record along with time and the correlation between the course title of the target course and the news title; according to the activeness of the user and the learning state of the user embodied by the learning record of each user, obtaining a recommendation coefficient of each user; and performing global recommendation and personalized recommendation on the user based on the numerical range of the recommendation coefficient of the user. According to the invention, the accuracy and real-time performance of course recommendation of the education platform are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for recommending teaching courses based on big data. Background Technology

[0002] With the development of online education, various teaching platforms have emerged, such as university course platforms, musical instrument learning platforms, language learning platforms, and corporate internal course platforms. These platforms provide convenient teaching methods, allowing learners to access courses anytime, anywhere. However, different platforms target different user groups. When users are on platforms with a large variety of courses, they may not be able to find suitable courses, which can affect their motivation. On the other hand, platforms with fewer users have fewer courses, requiring users to make precise searches. Therefore, to reduce the search costs for learners and enable them to quickly locate the courses they need, teaching platforms often recommend courses based on user characteristics.

[0003] In related technologies, for example, Chinese patent document CN113407831B discloses a course recommendation method and device. This method involves comprehensively analyzing the learning attribute information of the user to determine the courses that the user is interested in and pushing them to the user, eliminating the need for manual searching and improving user experience. Chinese patent document CN113918806B discloses a method and related device for automatically recommending training courses. This method collects user and course information through various means, analyzes it, and then accurately recommends training courses that better meet the requirements of the enterprise and the user's needs based on the user's learning progress.

[0004] However, the current course recommendation method suffers from insufficient user information. When user information is inadequate and users haven't completed enough courses, a cold start problem arises, rendering the method ineffective. Furthermore, for courses with high time sensitivity and strong location-based correlation, the current method cannot accurately push them, such as online courses. Therefore, a new method for recommending educational courses is urgently needed. Summary of the Invention

[0005] To address the technical problem of insufficient accuracy in the recommended teaching courses, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a teaching course recommendation method based on big data, comprising: The system acquires personal information and learning records of several users. The personal information includes the user's registration duration and coordinates. The learning records include several courses, course titles, course release times, total course duration, learning start times, and learning durations. It also acquires several news headlines within a preset time range. Any course is selected as the target course. Based on the number of users whose learning records contain the target course and their coordinate distribution range, the system obtains the first recommendability of the target course. Based on the linear change in the number of users whose learning records contain the target course over time and the correlation between the target course title and the news headlines, the system obtains the second recommendability of the target course. Based on the user's activity level and learning status reflected in their learning records, the system obtains the recommendation coefficient for each user. The activity level describes the user's usage of the platform, and the learning status represents the effectiveness of the user's learning of the courses through learning duration. Based on the numerical range of the user's recommendation coefficient, the system performs global and personalized recommendations for the user. The global recommendation is based on the first and second recommendability, while the personalized recommendation is based on a collaborative filtering algorithm and other users with similar personal information and similar recommendation coefficients.

[0007] This invention fully considers the reasons why users learn courses, integrating factors such as the real-time nature of the courses, their popularity among users, and users' personal information into course recommendations, thereby improving the accuracy and real-time nature of course recommendations. Based on the comprehensiveness of user information, this invention performs both global and personalized recommendations, ensuring that recommended courses meet users' learning preferences while also recommending courses that users might like even when user information is insufficient, further improving the accuracy of course recommendations.

[0008] Preferably, obtaining the first recommendability of the target course includes: Construct a convex hull for the coordinates of all target users of the target course, denoted as the first convex hull; construct a convex hull for the coordinates of all users, denoted as the second convex hull; obtain the area of ​​the first convex hull and the area of ​​the second convex hull; obtain the global recommendability and regional recommendability of the target course. The first recommendationability expression of the target course is: ; In the formula, Indicates the primary recommendation of the target course; Indicates the overall recommendability of the target course; This indicates the regional recommendability of the target course.

[0009] This invention obtains the global and regional recommendability of courses based on the geographical range of the course users, avoiding intrusive recommendations of popular courses that are irrelevant to the users.

[0010] Preferably, obtaining the global recommendability and regional recommendability of the target course includes: The global recommendability of the target course satisfies the expression: ; The regional recommendability of the target course satisfies the expression: ; In the formula, Indicates the overall recommendability of the target course; Indicates the regional recommendability of the target course; This indicates the number of target users for the target course; , This represents the area of ​​the first and second convex hulls; This represents the average distance among all target users of the target course. Represents the normalization function; This represents the natural exponential function.

[0011] Preferably, the acquisition of the second recommendability of the target course includes: Obtain the user number trend line for the target course; based on the word segmentation algorithm, classify the words contained in the course title of the target course as Class I words and the words contained in the news title as Class II words. ; In the formula, This indicates the second most recommended aspect of the target course; The number of trend lines representing the number of users of the target course; This represents the average distance between the trend line for the number of users of the i-th target course and the time axis; , This represents the slope of the trend line indicating the number of users for the (i+1)th and ithth users of the target course; Indicates the number of words in the intersection of Category 1 and Category 2 vocabulary; The frequency of words representing the r-th intersection of words from category one and category two; Represents the absolute value function; This represents the normalization function.

[0012] This invention determines the recommendability of a course based on its real-time popularity and user count, ensuring long-term recommendations of high-quality courses with a strong long-tail effect. It also pays close attention to current events-related courses, improving the timeliness of course recommendations.

[0013] Preferably, the method for obtaining the trend line of the number of users for the target course includes: Using the course release time of the target course as the starting point of the timeline, the number of users on the corresponding day is obtained in units of days, and a time-user number scatter plot is constructed. A time window is preset, and the time window is shifted from the starting point of the timeline in steps of days to obtain several time-user number scatter plots within the time window. Linear fitting is performed on each of these time windows to obtain several user number trend lines for the target course.

[0014] Preferably, the recommendation coefficients of each user satisfy the expression: ; In the formula, This represents the recommendation coefficient for the u-th user; This represents the registration duration of the u-th user; This represents the number of courses included in the learning records of the u-th user; This represents the learning duration of the u-th user for the v-th course included in their learning record; This represents the total course duration for the u-th user in the v-th course included in their learning record; Represents the natural exponential function; This represents the normalization function.

[0015] This invention obtains a user's recommendation coefficient based on the user's learning frequency and learning effect, which can fully reflect the user's effective use of the platform, thereby determining the degree to which the user needs personalized recommendations, and providing a foundation for accurate course recommendations.

[0016] Preferably, the step of providing global and personalized recommendations to users includes: Multiply the recommendation coefficient of user a by the number of first-category recommended courses for user a, and round up to obtain the number of final recommended courses of category a for user a; subtract the recommendation coefficient of user a from 1, multiply by the number of second-category recommended courses for user a, and round up to obtain the number of final recommended courses of category a for user a; obtain the courses with the highest recommendability corresponding to the number of final recommended courses of category a and the number of final recommended courses of category a, and use the final recommended courses for recommending courses to user a.

[0017] This invention combines global recommendation with personalized recommendation, avoiding the cold start problem and insufficient recommendation accuracy of traditional recommendation algorithms.

[0018] Preferably, obtaining the first recommended course for the a-th user includes: Based on the distance between users, the similarity of their learning records, and the difference in their recommendation coefficients, obtain the similarity between any two users; then select the user with the highest similarity to the a-th user. Let user a be the reference user of user a. The courses included in the learning records of user a's reference user, but not included in user a's learning records, are designated as user a's first courses to be recommended.

[0019] Preferably, obtaining the second recommended course for the a-th user includes: Multiply the first recommendation and second recommendation of the target course to obtain the overall recommendation of the target course. This overall recommendation is divided into global overall recommendation and regional overall recommendation. Obtain the average distance between the target user of all courses and the a-th user, and select the course with the closest average distance. The course with the highest overall regional recommendation among all courses. This course is the one with the highest overall recommendation value. These courses are collectively referred to as the second recommended course for user a.

[0020] Secondly, the present invention provides a teaching course recommendation system based on big data, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned teaching course recommendation method based on big data is implemented.

[0021] By adopting the above technical solution, a computer program is generated from the above-mentioned big data-based teaching course recommendation method and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0022] The beneficial effects of this invention are as follows: (1) By constructing a multi-dimensional course recommendability model, this invention can provide high-quality “global recommendations” based on the global value of courses, regional popularity, time-series trends and news relevance, even for new users with zero history, effectively solving the cold start problem; (2) This invention accurately assesses user activity and learning effectiveness through user recommendation coefficients, and dynamically adjusts the weights of global recommendation and personalized recommendation based on these coefficients, thus achieving a combination of universality and personalization. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a teaching course recommendation method based on big data in this invention. Detailed Implementation

[0024] This invention discloses a teaching course recommendation method based on big data, referring to... Figure 1 This includes steps S1-S4: S1: Obtain personal information and learning records of several users; obtain several news headlines within a preset time range.

[0025] It should be noted that some course platforms are highly time-sensitive and geographically specific. Taking online courses as an example, time-sensitive courses need to be disseminated in a short period of time, and courses vary from region to region. Therefore, when recommending courses, factors such as time and geographical location need to be considered to ensure that the course recommendations meet the requirements of high timeliness and high geographical specificity, thereby improving the effectiveness of the course recommendations.

[0026] Specifically, the system collects personal information and learning records from several users. The personal information includes at least a user ID, registration duration, and coordinates. The user ID is unique and used to distinguish users. The coordinates are the coordinates of the user's most frequently displayed IP address on the platform in a global coordinate system. The learning records include several courses, along with course titles, course release times, total course duration, start time, and duration of study. It should be noted that, for accurate course recommendations, the data collected should include all IP addresses and different numbers of courses studied.

[0027] The system retrieves several news headlines within a preset time range, which can be set to 7 days or customized by the implementers based on actual implementation needs. These news headlines are obtained by crawling news websites. It should be noted that these news headlines are used to measure the real-time popularity of the course, ensuring the timeliness of course recommendations.

[0028] At this point, the user's personal information and learning records, as well as news headlines within a preset time range, have been obtained.

[0029] S2: Select any course as the target course; based on the number of users whose learning records contain the target course and their coordinate distribution range, obtain the first recommendability of the target course; based on the linear change of the number of users whose learning records contain the target course over time and the correlation between the course title of the target course and the news title, obtain the second recommendability of the target course.

[0030] It should be noted that for some important courses, a large number of users will study them within a short period of time. When the coordinates of these users are concentrated, it indicates that the course is geared towards users in a specific region; when the coordinates of these users are dispersed, it indicates that the course is geared towards a wider range of users. Therefore, this invention combines the distribution of user coordinates and the number of users to determine the primary recommendability of each course.

[0031] Specifically, any course is taken as the target course, users whose learning records contain the target course are identified as the target users of the target course, and the straight-line distance between the coordinates of any two target users is identified as the distance between the two target users.

[0032] It should be noted that the more target users a target course has, and the smaller the distance between all target users, the stronger the regional recommendability of the target course; conversely, the more target users a target course has, and the greater the distance between all target users, the stronger the global recommendability of the target course.

[0033] Preferably, the primary recommendability of the target course is determined based on the number of users whose learning records include the target course and their coordinate distribution range, including: Construct a convex hull for the coordinates of all target users of the target course, denoted as the first convex hull; construct a convex hull for the coordinates of all users, denoted as the second convex hull; obtain the area of ​​the first convex hull and the area of ​​the second convex hull.

[0034] ; ; ; In the formula, Indicates the primary recommendation of the target course; Indicates the overall recommendability of the target course; Indicates the regional recommendability of the target course; This indicates the number of target users for the target course; , This represents the area of ​​the first and second convex hulls; This represents the average distance among all target users of the target course. Represents the normalization function; This represents the natural exponential function.

[0035] In the formula, This represents the percentage of the target course's target users' coordinate range that is included in the coordinate range of all users' coordinate ranges. A larger value indicates a wider distribution of target users, thus a broader audience and greater overall recommendability. Furthermore, a larger average distance and a larger number of target users indicate a larger and less concentrated target user base, further suggesting higher overall and primary recommendability. Similarly, The smaller the value, the smaller the average distance of the target users of the target course and the more concentrated their distribution. This indicates that the target course is geared towards regional users, and therefore the regional recommendability of the target course is higher.

[0036] At this point, the primary recommendation for each course has been determined.

[0037] It should be noted that the first recommendability is quantified based on the number of course users and the scope of those users, reflecting the course's popularity and its range. However, the popularity of these courses is often short-lived. Some current affairs courses will lose attention after the relevant events have passed, while some basic courses will continue to be studied over time, and new users will also need to learn them, thus maintaining a certain level of popularity. These courses still possess recommendability. Therefore, this invention combines changes in course popularity to calculate the second recommendability of each course. Considering that course attention is often related to current events, this invention uses news data from news websites as a reference to obtain course popularity, and combines this with the number of users of the course itself to calculate the second recommendability of the target course.

[0038] Preferably, based on the linear change in the number of users whose learning records contain the target course over time and the relevance between the target course title and the news headline, a second recommendability of the target course is obtained, including: Using the course release time as the starting point of the timeline, the number of users on the corresponding day of the start of learning is obtained, and a time-user count scatter plot is constructed. A preset time window is used, and the time window is shifted from the starting point of the timeline in days, resulting in several time-user count scatter plots within each time window. Linear fitting is then performed on each time window to obtain several user count trend lines for the target course. It should be noted that as the number of users of the target course gradually decreases, the slope of the user count trend lines will gradually flatten and approach the timeline. However, if the slope of the user count trend lines fluctuates and moves away from the timeline, it indicates that the target course exhibits a long-tail effect, meaning that the target course remains highly recommendable for a long period after its release.

[0039] The course titles of the target course and all acquired news titles were segmented using a word segmentation algorithm. Words contained in the course titles were categorized into Category 1 words, and words contained in the news titles were categorized into Category 2 words. It should be noted that the word segmentation algorithm can use the Jieba model. The higher the frequency of Category 1 words within Category 2 words, the higher the real-time popularity of the target course, thus indicating a higher real-time recommendability for the target course.

[0040] The second recommendationability expression of the target course is satisfied: ; In the formula, This indicates the second most recommended aspect of the target course; The number of trend lines representing the number of users of the target course; This represents the average distance between the trend line for the number of users of the i-th target course and the time axis; , This represents the slope of the trend line indicating the number of users for the (i+1)th and ithth users of the target course; Indicates the number of words in the intersection of Category 1 and Category 2 vocabulary; The frequency of words representing the r-th intersection of words from category one and category two; Represents the absolute value function; This represents the normalization function.

[0041] In the formula, This represents the difference in the slope of the trend lines for the (i+1)th and ithth user numbers of the target course, multiplied by the average distance between the trend line for the ith user number of the target course and the time axis. The larger this value, the greater the change in the number of users during the time period corresponding to the (i+1)th and ithth user number trend lines of the target course, and the larger the number of users, thus indicating that the target course is more popular at the corresponding time. This indicates the overall change in user interest after the target course is released. The larger this value, the higher the user interest in the target course after its release, and therefore the greater the second recommendability of the target course.

[0042] At this point, the second recommendation for each course was obtained.

[0043] S3: Based on each user's learning records, which reflect their activity level and learning status, obtain the recommendation coefficient for each user.

[0044] It should be noted that as users utilize the course platform, their learning habits and personal preferences gradually emerge. Personalized recommendations can be made based on these habits. Combining this with secondary recommendability of courses based on real-time popularity and course popularity allows for accurate recommendations while ensuring comprehensiveness. However, with new users constantly joining the platform and increasing user privacy concerns, the analysis of new users' habits and preferences may become inaccurate, hindering personalized recommendations. Instead, recommendations should be based on the user's location and primary recommendability of courses based on factors such as region. Therefore, this invention obtains a user recommendation coefficient. A lower recommendation coefficient indicates a greater reliance on global recommendations, while a higher coefficient suggests a greater emphasis on personalized recommendations.

[0045] It should be further explained that the more courses a user studies and the higher the frequency of study, the stronger the user's self-study habits are, which indicates that the user has a higher level of activity. On the other hand, the number of courses studied does not represent the quality of learning, and some clicks based on curiosity do not represent real learning. Therefore, the learning status reflected by the duration of study can be combined to obtain a more accurate recommendation coefficient for users.

[0046] Specifically, based on each user's learning records, which reflect their activity level and learning status, a recommendation coefficient is obtained for each user, including: The recommendation coefficient of any user satisfies the expression: ; In the formula, This represents the recommendation coefficient for the u-th user; This represents the registration duration of the u-th user; This represents the number of courses included in the learning records of the u-th user; This represents the learning duration of the u-th user for the v-th course included in their learning record; This represents the total course duration for the u-th user in the v-th course included in their learning record; Represents the natural exponential function; This represents the normalization function.

[0047] In the formula, This represents the average interval between the registration of the u-th user and the learning of each course. The larger this value is, the lower the user's learning frequency, and the lower the recommendation coefficient of the u-th user. This represents the ratio of the learning time of the u-th user for the v-th course included in their learning record to the total course time. The larger this value, the more comprehensive the u-th user's learning of the course, and thus the better the user's learning status for the course. This represents the sum of the learning status of the u-th user across all the courses they have studied. The larger this value is, the better the overall learning status of the u-th user, and thus the higher the recommendation coefficient for the u-th user.

[0048] At this point, the recommendation scores for all users have been obtained.

[0049] S4: Based on the numerical range of the user's recommendation coefficient, perform global and personalized recommendations for the user.

[0050] It's important to note that the lower a user's recommendation coefficient, the more appropriate global recommendations should be made; conversely, the higher the recommendation coefficient, the more appropriate personalized recommendations should be made. For global recommendations, the first and second recommendability metrics analyze a course's recommendability based on its geographical location, changes in user numbers, and real-time popularity, respectively. Recommending highly recommendable courses to users with low recommendation coefficients ensures that users can access high-quality courses promptly after joining the platform. For personalized recommendations, user-based collaborative filtering algorithms can be used to recommend courses studied by other users with similar personal information and recommendation coefficients to users with high recommendation coefficients. Simultaneously, recommending some highly recommendable courses to users with high recommendation coefficients prevents them from getting trapped in information silos and missing important courses.

[0051] Specifically, the first recommended course for any user is obtained based on the collaborative filtering algorithm, including: The straight-line distance between the coordinates of any two users is denoted as the distance between the two users.

[0052] The similarity between any two users satisfies the expression: ; In the formula, This represents the similarity between user a and user b. This represents the distance between user a and user b. , Let represent the recommendation coefficients of user a and user b. This represents the number of courses in the intersection of the learning records of user a and user b. This represents the number of courses in the union of the learning records of user a and user b. Represents the absolute value function; This represents the natural exponential function.

[0053] In the formula, This represents the difference in recommendation coefficients between user a and user b. The larger the value, the greater the distance between user a and user b. The larger the value, the greater the difference in geographical location, platform usage time, learning habits, etc. between user a and user b, and the lower the similarity between user a and user b. This value represents the similarity of the courses in the learning records of user a and user b. The larger the value, the more similar the learning records of user a and user b are, and the higher the similarity between user a and user b.

[0054] The user with the highest similarity to the a-th user Let user a be the reference user of user a. Courses included in the learning records of user a's reference user, but not included in user a's learning records, will be designated as user a's first recommended courses. It should be noted that... The value is set by the implementers based on the actual implementation situation, for example The value can be set to 10.

[0055] At this point, the first course to be recommended for any user has been obtained.

[0056] Preferably, based on the first and second recommendability, obtaining the second recommended course for any user includes: multiplying the first recommendability and the second recommendability of the target course to obtain the overall recommendability of the target course, which is divided into global overall recommendability and regional overall recommendability; obtaining the average distance between the target user of all courses and the a-th user, and selecting the course with the closest average distance... The course with the highest overall regional recommendation among all courses. This course is the one with the highest overall recommendation value. These courses are collectively referred to as the second recommended course for user a. It should be noted that... The settings are configured by the implementers based on the actual implementation situation, for example... It can be set to 10.

[0057] Preferably, based on the recommendation coefficient of each user, the final recommended courses for each user are obtained, including: multiplying the recommendation coefficient of the a-th user by the number of first-category recommended courses of the a-th user, and rounding up to obtain the number of first-category final recommended courses of the a-th user; subtracting the recommendation coefficient of the a-th user from 1, multiplying by the number of second-category recommended courses of the a-th user, and rounding up to obtain the number of second-category final recommended courses of the a-th user; obtaining the number of first-category final recommended courses and the number of second-category final recommended courses, and the courses with the highest recommendability corresponding to the number of first-category final recommended courses together constitute the final recommended courses for the a-th user.

[0058] After a user enters the platform, the final recommended courses are transmitted to the user's terminal to complete the course recommendation.

[0059] This completes the big data-based course recommendation process.

[0060] This invention also discloses a teaching course recommendation system based on big data, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a teaching course recommendation method based on big data according to this invention.

[0061] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0062] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A teaching course recommendation method based on big data, characterized in that, include: Obtain personal information and learning records of several users. The personal information includes the user's registration duration and coordinates. The learning records include several courses as well as course titles, course release times, total course durations, learning start times, and learning durations. Obtain several news headlines within a preset time range. Use any course as the target course; based on the number of users whose learning records contain the target course and their coordinate distribution range, determine the first recommendation value of the target course; The second recommendability of the target course is obtained based on the linear change in the number of users whose learning records contain the target course over time and the correlation between the course title of the target course and the news title. Based on each user's learning records, which reflect their activity level and learning status, a recommendation coefficient is obtained for each user. The activity level describes the amount of time a user uses the platform, and the learning status represents the effectiveness of the user's learning of the course through the duration of learning. Based on the numerical range of a user's recommendation coefficient, global and personalized recommendations are made to the user; the global recommendation is based on the first recommendability and the second recommendability, and the personalized recommendation is based on the collaborative filtering algorithm and other users with similar personal information and similar recommendation coefficients.

2. The teaching course recommendation method based on big data according to claim 1, characterized in that, The first recommendability of the target course includes: Construct a convex hull for the coordinates of all target users of the target course, denoted as the first convex hull; construct a convex hull for the coordinates of all users, denoted as the second convex hull; obtain the area of ​​the first convex hull and the area of ​​the second convex hull; obtain the global recommendability and regional recommendability of the target course. The first recommendationability expression of the target course is: ； In the formula, Indicates the primary recommendation of the target course; Indicates the overall recommendability of the target course; This indicates the regional recommendability of the target course.

3. The teaching course recommendation method based on big data according to claim 2, characterized in that, The acquisition of the global recommendability and regional recommendability of the target course includes: The global recommendability of the target course satisfies the expression: ; The regional recommendability of the target course satisfies the expression: ; In the formula, Indicates the overall recommendability of the target course; Indicates the regional recommendability of the target course; This indicates the number of target users for the target course; , This represents the area of ​​the first and second convex hulls; This represents the average distance among all target users of the target course. Represents the normalization function; This represents the natural exponential function.

4. The teaching course recommendation method based on big data according to claim 1, characterized in that, The second recommendability of the target course includes: Obtain the user number trend line for the target course; based on the word segmentation algorithm, classify the words contained in the course title of the target course as Class I words and the words contained in the news title as Class II words. ； In the formula, This indicates the second most recommended aspect of the target course; The number of trend lines representing the number of users of the target course; This represents the average distance between the trend line for the number of users of the i-th target course and the time axis; , This represents the slope of the trend line indicating the number of users for the (i+1)th and ithth users of the target course; Indicates the number of words in the intersection of Category 1 and Category 2 vocabulary; The frequency of words representing the r-th intersection of words from category one and category two; Represents the absolute value function; This represents the normalization function.

5. The teaching course recommendation method based on big data according to claim 4, characterized in that, The trend line of the number of users acquiring the target course includes: Using the course release time of the target course as the starting point of the timeline, the number of users on the corresponding day is obtained in units of days, and a time-user number scatter plot is constructed. A time window is preset, and the time window is shifted from the starting point of the timeline in steps of days to obtain several time-user number scatter plots within the time window. Linear fitting is performed on each of these time windows to obtain several user number trend lines for the target course.

6. The teaching course recommendation method based on big data according to claim 1, characterized in that, The recommendation coefficients of each user satisfy the expression: ； In the formula, This represents the recommendation coefficient for the u-th user; This represents the registration duration of the u-th user; This represents the number of courses included in the learning records of the u-th user; This represents the learning duration of the u-th user for the v-th course included in their learning record; This represents the total course duration for the u-th user in the v-th course included in their learning record; Represents the natural exponential function; This represents the normalization function.

7. The teaching course recommendation method based on big data according to claim 1, characterized in that, The aforementioned global and personalized recommendations for users include: Multiply the recommendation coefficient of user a by the number of first-category recommended courses for user a, and round up to obtain the number of final recommended courses of category a for user a; subtract the recommendation coefficient of user a from 1, multiply by the number of second-category recommended courses for user a, and round up to obtain the number of final recommended courses of category a for user a; obtain the courses with the highest recommendability corresponding to the number of final recommended courses of category a and the number of final recommended courses of category a, and use the final recommended courses for recommending courses to user a.

8. The teaching course recommendation method based on big data according to claim 7, characterized in that, The acquisition of the first recommended course for the a-th user includes: Based on the distance between users, the similarity of their learning records, and the difference in their recommendation coefficients, obtain the similarity between any two users; then select the user with the highest similarity to the a-th user. Let user a be the reference user of user a. The courses included in the learning records of user a's reference user, but not included in user a's learning records, are designated as user a's first courses to be recommended.

9. The teaching course recommendation method based on big data according to claim 7, characterized in that, The acquisition of the second recommended course for the a-th user includes: Multiply the first recommendation and second recommendation of the target course to obtain the overall recommendation of the target course. This overall recommendation is divided into global overall recommendation and regional overall recommendation. Obtain the average distance between the target user of all courses and the a-th user, and select the course with the closest average distance. The course with the highest overall regional recommendation among all courses. This course is the one with the highest overall recommendation value. These courses are collectively referred to as the second recommended course for user a.

10. A teaching course recommendation system based on big data, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a teaching course recommendation method based on big data according to any one of claims 1-9.

Images