Online course learning recommendation method based on knowledge graph

Abstract

The invention relates to the technical field of online course learning recommendation, in particular to an online course learning recommendation method based on a knowledge graph. According to the method, the obtained user information and the online course information are preprocessed, the use value of the obtained data can be improved, meanwhile, the preprocessed user information and the preprocessed online course information are input into the knowledge graph, the knowledge graph can better label the data, and the user experience is improved. According to the method and the system, on-line course recommendation can be more accurate by performing course package recommendation on the user according to the adaptation degree, and meanwhile, whether a recommendation result meets a standard or not is determined according to the page view of each recommended on-line course, and a reason for not meeting the standard is preliminarily determined, so that the recommendation efficiency is improved. Therefore, pushing can be more accurately carried out according to the requirements of the user, the efficiency of the user in the course learning process is improved, and then the online course recommendation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of online course learning recommendation, and particularly to an online course learning recommendation method based on a knowledge graph. Background Art

[0002] The online course learning recommendation method based on a knowledge graph is a new type of recommendation method that combines knowledge graph technology and recommendation algorithms. This method constructs a course knowledge graph by analyzing the behavior data of learners on the online course platform, the course attributes, and the association relationships between courses, and provides personalized course recommendations based on this graph. Therefore, this method can provide more accurate course recommendations for learners and can customize personalized course recommendation plans according to the personal circumstances and interest preferences of learners.

[0003] Chinese Patent Publication No.: CN116501970A discloses an online course recommendation method based on a knowledge graph and convolution, including: the model first uses a feature extraction module to extract and transform the user's historical interaction information and the feature information in the course field into embedding vectors, and fuse these vectors into the knowledge graph to generate the embedding expressions of the user and the course. Finally, these information are passed to adjacent nodes through a message propagation algorithm to obtain the final embedding vectors of the user and the course for recommendation.

[0004] It can be seen that the above solution has the following problems: it fails to adjust the recommended online courses in a timely manner according to the user's needs, resulting in low efficiency of the user in the process of course learning, and further reducing the efficiency of online course recommendation. Summary of the Invention

[0005] For this reason, the present invention provides an online course learning recommendation method based on a knowledge graph to overcome the problem in the prior art that the recommended online courses cannot be adjusted in a timely manner according to the user's needs, resulting in low efficiency of the user in the process of course learning and further reducing the efficiency of online course recommendation.

[0006] To achieve the above object, the present invention provides an online course learning recommendation method based on a knowledge graph, including:

[0007] Obtain user information and online course information;

[0008] Preprocess the user information and the online course information respectively;

[0009] Input the preprocessed user information and the online course information into the knowledge graph. The knowledge graph packs the online courses into several course packages according to the characteristics of the online courses, outputs several preset labels for each course package, and outputs several demand labels for the user according to the user information, where the preset label is the common feature of all the courses in the course package;

[0010] Calculate the fitness between the course package and the user requirements, where the fitness is the ratio of the number of identical tags between the preset tags of the course package and the requirement tags of the user to the total number of the user's requirement tags;

[0011] Recommend the course package to the user according to the fitness, and when the fitness is higher than the preset fitness, push the course package to the user;

[0012] Determine whether the recommendation result of the online course meets the standard based on the view count of each recommended course package, and determine the corresponding processing method based on the reasons for not meeting the standard;

[0013] Adjust the online course learning recommendation method based on the processing method.

[0014] Further, the process of determining whether the recommendation result of the online course meets the standard based on the view count of each course package includes: determining whether the recommendation result meets the standard based on the comparison result between the view count and the pre-stored preset view count; if the view count is greater than or equal to the first preset view count, determine that the recommendation result meets the standard; if the view count is less than the first preset view count and greater than the second preset view count, then determine whether the recommendation result meets the standard based on the cumulative learning duration of the user for the recommended course package, where the cumulative learning duration is the average duration for the users who receive the course package to learn the course; if the view count is less than or equal to the second preset view count, determine that the recommendation result does not meet the standard, and determine the reason why the recommendation result does not meet the standard based on the difference between the view count at the previous moment and the current view count.

[0015] Further, the process of determining whether the recommendation result meets the standard based on the cumulative learning duration of the user for the recommended course package includes: comparing the cumulative learning duration with the preset learning duration; if the cumulative learning duration is less than or equal to the preset learning duration, determine that the recommendation result does not meet the standard, and determine the reason why the recommendation result does not meet the standard based on the difference between the view count at the previous moment and the current view count; if the cumulative learning duration is greater than the preset learning duration, determine that the recommendation result meets the standard, and then adjust the preset view count based on the actual usage ratio, where the actual usage ratio is the ratio of the number of users whose cumulative learning duration is greater than the preset learning duration to the total number of users who receive the push of the course package.

[0016] Further, the process of adjusting the preset view count based on the actual usage ratio includes: calculating the difference between the actual usage ratio and the preset usage ratio; reducing the preset view count based on the obtained difference, and the difference is proportional to the reduction amplitude of the preset view count.

[0017] Further, the process of adjusting the preset view count further includes: calculating the difference between the average interaction ratio of the user and the online course and the preset interaction ratio; reducing the preset view count based on the obtained difference, and the difference is proportional to the reduction amplitude of the view count.

[0018] Further, the process of determining the reason that the recommendation result does not meet the standard when the view count is less than or equal to the second preset view count includes: calculating the absolute value of the difference between the view count at the previous moment and the current view count and the preset difference, and determining the reason that the recommendation result does not meet the standard based on the comparison result between the absolute value and the pre-stored preset absolute value, where: if the absolute value is less than or equal to the first preset absolute value, it is determined that there is a problem with the course package itself, and a processing notice is issued; if the absolute value is greater than the first preset absolute value and less than the second preset absolute value, the reason that the recommendation result does not meet the standard is determined based on the variance of the historical view count; if the absolute value is greater than or equal to the second preset absolute value, it is determined that there is a problem with the setting of the preset adaptation degree, and the preset adaptation degree is adjusted based on the difference between the corresponding preset view count and the actual view count.

[0019] Further, the process of determining the reason that the recommendation result does not meet the standard based on the variance of the historical view count includes: comparing the variance with the preset variance; if the variance is greater than the preset variance, drawing a historical moment-historical view count curve for the course package, and determining the reason that the recommendation result does not meet the standard based on the characteristics of the historical moment-historical view count curve; if the variance is less than or equal to the preset variance, it is determined that there is a problem with the course package itself, and a processing notice is issued.

[0020] Further, the process of determining the characteristics of the curve based on the autocorrelation function includes: determining the lag number; according to the lag number, calculating the autocorrelation value using the normalization method, and drawing a lag number-autocorrelation value curve; counting the number of peaks of the curve, and comparing the number of peaks with the lag number; if the number is greater than the lag number, it is determined that the curve has periodicity; if the number is less than or equal to the lag number, it is determined that the curve does not have periodicity.

[0021] Further, the process of determining the reason why the recommended result does not meet the standard based on the characteristics of the lag number-autocorrelation value curve includes: determining the reason why the recommended result does not meet the standard based on whether the curve is periodic; if the lag number-autocorrelation value curve is periodic, it is determined that the reason is that the current course package is in the off-season for browsing, and then the view volume of the corresponding course package is continuously monitored; if the lag number-autocorrelation value curve is not periodic, the reason is determined according to the comparison result of the view volume at the previous moment and the current view volume; if the view volume at the previous moment is greater than the current view volume, it is determined that the reason is that there is a problem with the course package itself, and a processing notice is issued; if the view volume at the previous moment is less than or equal to the current view volume, it is determined that the reason is that the current course package is in the off-season for browsing, and the view volume of the corresponding course package is continuously monitored.

[0022] Further, the process of adjusting the preset adaptation degree based on the difference between the corresponding preset view volume and the actual view volume includes: increasing the preset adaptation degree based on the difference, and the difference is proportional to the increase amplitude of the preset adaptation degree.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows. By preprocessing the obtained user information and online course information, the present invention can improve the utilization value of the obtained data. At the same time, by inputting the preprocessed user information and online course information into the knowledge graph, the knowledge graph can better label these data, calculate the adaptation degree between the user and the course package based on the labels, and recommend course packages to users according to the adaptation degree, which can make the online course recommendation more accurate. At the same time, by determining whether the recommended result meets the standard according to the view volume of each recommended online course and initially determining the reason for not meeting the standard, it is possible to push more accurately according to the needs of users, improve the efficiency of users in the process of course learning, and further improve the efficiency of online course recommendation.

[0024] Further, the present invention determines whether the recommended result meets the standard by comparing the actual view volume with the pre-stored preset view volume and initially determines the reason for the non-compliance of the recommended result, which can quickly determine whether the online course recommendation can achieve the effect, thereby further improving the efficiency of online course recommendation.

[0025] Further, by using the cumulative learning duration, the present invention determines again whether the recommended result meets the standard when the actual view volume is within the preset view volume range, which can make the determination of whether the recommended result meets the standard more accurate, so that users can learn more efficiently according to the recommended courses, and further improve the efficiency of online course recommendation.

[0026] Furthermore, the present invention adjusts the preset view count based on the difference between the actual usage ratio and the preset usage ratio, which can adjust the preset view count more accurately, making the determination result of whether the recommendation result meets the standard more accurate, and further improving the efficiency of online course recommendation.

[0027] Furthermore, the present invention adjusts the preset view count based on the difference between the average interaction ratio of online courses and the preset interaction ratio, further making the determination result of whether the recommendation result meets the standard more accurate, and further improving the efficiency of online course recommendation.

[0028] Furthermore, the present invention determines the reason why the recommendation result does not meet the standard based on the absolute value of the difference between the view count at the previous moment and the current view count and the preset difference, which can more accurately determine the reason why the recommendation result does not meet the standard, so that subsequent adjustments can be made more accurately according to the reason, and further improve the efficiency of online course recommendation.

[0029] Furthermore, when the absolute value of the difference between the view count at the previous moment and the current view count is within the interval of the preset absolute value, the present invention determines the reason why the recommendation result does not meet the standard based on the variance of historical view counts and the preset variance, which can more accurately determine the reason why the recommendation result does not meet the standard, so that adjustments can be made more accurately according to the reason, and further improve the efficiency of online course recommendation.

[0030] Furthermore, the present invention determines whether the historical moment-historical view count curve is periodic through the autocorrelation function, which can more scientifically and accurately determine whether the recommendation result of the online course recommendation has a pattern, so that the reason why the recommendation result does not meet the standard can be determined more precisely subsequently, and further improve the efficiency of online course recommendation.

[0031] Furthermore, the present invention determines the reason why the recommendation result does not meet the standard by judging whether the lag number-autocorrelation value curve is periodic, which can more precisely determine the reason why the recommendation result does not meet the standard, so that the online course recommendation method can be adjusted more accurately according to the reason, and further improve the efficiency of online course recommendation.

[0032] Furthermore, the present invention adjusts the preset fitness based on the difference between the preset view count and the actual view count, which can more accurately adjust the preset fitness, making the determination of the recommendation result more accurate, and further improving the efficiency of online course recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of the steps of the online course learning recommendation method based on a knowledge graph according to an embodiment of the present invention;

[0034] Figure 2The flowchart of the steps determined based on the comparison result between the view count of each course package and the preset view count in the embodiments of the present invention;

[0035] Figure 3 The flowchart of the steps determined based on the cumulative learning duration of the user for the recommended course package and the preset learning duration in the embodiments of the present invention;

[0036] Figure 4 The flowchart of the steps determined based on the absolute value of the difference between the view count at the previous moment and the current view count and the preset absolute value pre - stored in the embodiments of the present invention. Detailed implementation manners

[0037] In order to make the objectives and advantages of the present invention more clear, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0038] The preferred implementation manners of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these implementation manners are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0039] Please refer to Figure 1 as shown, the flowchart of the steps of the online course learning recommendation method based on the knowledge graph in the embodiments of the present invention.

[0040] The steps in the actual operation process of the embodiments of the present invention include:

[0041] S1. Obtain user information and online course information;

[0042] S2. Pre - process the user information and the online course information respectively;

[0043] S3. Input the pre - processed user information and the online course information into the knowledge graph. The knowledge graph packages the online courses into several course packages according to the characteristics of the online courses, and outputs several preset tags for each course package and several demand tags for the user according to the user information. Among them, the preset tags are the common characteristics of all the courses in the course package;

[0044] S4. Calculate the fitness between the course package and the user's demands. Among them, the fitness is the ratio of the number of the same tags in the preset tags of the course package and the demand tags of the user to the total number of the user's demand tags;

[0045] S5. Recommend the course package to the user according to the fitness, and when the fitness is higher than the preset fitness, push the course package to the user;

[0046] S6. Determine whether the recommendation results of the online courses meet the standards based on the view counts of the recommended course packages, and determine the corresponding handling methods based on the reasons for non-compliance;

[0047] S7. Adjust the online course learning recommendation method based on the handling method.

[0048] Specifically, first obtain user information and online course information through a website analysis tool, and then preprocess the user information, including the following steps: First, use data cleaning to ensure that each user's information is unique and handle missing values, filling the missing values with the mean. Second, use data transformation to normalize numerical data (such as age) to eliminate the scale differences between different features, and encode categorical data (such as gender and occupation). Then, according to the course content and data characteristics, select features useful for user behavior analysis. Finally, integrate the cleaned and transformed user data together for storage. Preprocess the online course information, including the following steps: First, use data cleaning methods to remove duplicate course information and ensure that each course's information is unique, and for missing course information (such as teacher name and course description), adopt filling or deletion strategies for processing. Second, use data transformation to normalize numerical data (such as class hours and credits) and perform word segmentation and word vectorization on text data (such as course description) for text analysis.

[0049] Specifically, input the preprocessed user information and online course information into the constructed knowledge graph. The system will first deeply analyze the online course information. Each course is regarded as a node in the graph, and the relationships such as associations, dependencies, and similarities between courses are represented by edges. The system will use advanced technologies such as clustering algorithms or topic models according to the characteristics of the courses, such as subject fields, teaching methods, target audiences, etc., to intelligently package these courses into several course packages with internal logical connections. Among them, each course package will be given a series of preset labels, which accurately reflect the common characteristics of all the courses in the course package. For example, a course package may contain all the basic courses on "Data Analysis and Machine Learning", so its preset labels may include "Computer Science", "Data Analysis", "Introduction to Machine Learning", etc. At the same time, the user information will also be deeply mined, and several demand labels of the user will be generated according to multi-dimensional information such as the user's personal characteristics, historical behaviors, and interests. These demand labels accurately depict the user's current learning needs, interest preferences, and potential learning goals. For example, a user who is interested in the field of artificial intelligence may have demand labels such as "Frontier of AI Technology", "Python Programming", "Deep Learning", etc.

[0050] Please refer toFigure 2 As shown, it is a flowchart of the steps determined based on the comparison result between the view count of each course package and the preset view count in an embodiment of the present invention. The process of determining whether the recommended result of the online course meets the standard based on the view count of each course package in an embodiment of the present invention includes: determining whether the recommended result meets the standard based on the comparison result between the view count and the preset view count stored in advance; if the view count is greater than or equal to the first preset view count, determining that the recommended result meets the standard; if the view count is less than the first preset view count and greater than the second preset view count, then determining whether the recommended result meets the standard based on the cumulative learning duration of the user for the recommended course package, where the cumulative learning duration is the average duration of the users who receive the course package to study the course; if the view count is less than or equal to the second preset view count, determining that the recommended result does not meet the standard, and determining the reason why the recommended result does not meet the standard based on the difference between the view count at the previous moment and the current view count.

[0051] Specifically, in this embodiment, the view count L0 can be divided into the first preset view count L1 and the second preset view count L2. It is set that in the view count standard, the first preset view count L1 = 3000, and the second preset view count L2 = 0.7×L1. It should be noted that in other embodiments, the values of L1 and L2 can also be determined according to the relative online course requirements, and the numerical values of the view count are all rounded up; the specific process of comparing the view count L with L1 and L2 is as follows:

[0052] If the view count L is greater than or equal to the first preset view count L1, then determine that the recommended result meets the standard based on the view count;

[0053] If the view count L is less than the first preset view count L1 and greater than the second preset view count L2, it indicates that it is necessary to make a secondary determination of whether the recommended result meets the standard based on the cumulative learning duration P of the user for the recommended course package, where the cumulative learning duration is the average duration of the users who receive the course package to study the course;

[0054] If the view count L is less than or equal to the second preset view count L2, then determine that the recommended result does not meet the standard based on the view count, and determine the reason why the recommended result does not meet the standard based on the difference Q between the view count at the previous moment and the current view count.

[0055] Please refer to Figure 3As shown, it is a flowchart of the steps for the embodiment of the present invention to determine based on the cumulative learning duration of the user for the recommended course package and the preset learning duration. The process of the embodiment of the present invention to determine whether the recommended result meets the standard based on the cumulative learning duration of the user for the recommended course package includes: comparing the cumulative learning duration with the preset learning duration; if the cumulative learning duration is less than or equal to the preset learning duration, it is determined that the recommended result does not meet the standard, and the reason why the recommended result does not meet the standard is determined based on the difference between the previous moment's view count and the current view count; if the cumulative learning duration is greater than the preset learning duration, it is determined that the recommended result meets the standard, and the preset view count is adjusted based on the actual usage ratio, where the actual usage ratio is the ratio of the number of users whose cumulative learning duration is greater than the preset learning duration to the total number of users who received the push of this course package.

[0056] Specifically, in the case where the view count L is less than the first preset view count L1 and greater than the second preset view count L2, it is determined whether the recommended result meets the standard based on the cumulative learning duration and the preset learning duration; in this embodiment, the preset learning duration P0 = 5h is set, and the specific process of comparing the cumulative learning duration and the preset learning duration is as follows:

[0057] If the cumulative learning duration P is less than or equal to the preset learning duration P0, it indicates that the user's learning duration does not meet the standard, and it is determined that the recommended result does not meet the standard, and the reason why the recommended result does not meet the standard is determined based on the difference R between the previous moment's view count and the current view count;

[0058] If the cumulative learning duration P is greater than the preset learning duration P0, it indicates that the user's learning duration meets the standard, and it is determined that the recommended result meets the standard, and the preset view count is adjusted based on the actual usage ratio.

[0059] Specifically, the process of the embodiment of the present invention to adjust the preset view count based on the actual usage ratio includes: calculating the difference between the actual usage ratio and the preset usage ratio; reducing the preset view count based on the obtained difference, and the difference is proportional to the reduction amplitude of the preset view count.

[0060] Specifically, in this embodiment, the preset difference T0 = 0.2, and the specific process of comparing the difference between the actual usage ratio and the preset usage ratio with the preset difference is as follows:

[0061] If the difference T is less than or equal to the preset difference T0, the preset view count is adjusted to 0.8 times the original preset view count;

[0062] If the difference T is greater than the preset difference T0, the preset view count is adjusted to 0.5 times the original preset view count.

[0063] Specifically, the process of adjusting the preset view count in the embodiments of the present invention further includes: calculating the difference between the average interaction ratio of the user and the online course and the preset interaction ratio; reducing the preset view count based on the obtained difference, and the difference is proportional to the reduction amplitude of the view count.

[0064] Specifically, in this embodiment, the interaction ratio is the completion degree of the interactive content in the course, that is, the ratio of the number of completed interactive contents to the total number of interactions in the course. Among them, the interactive content is the questions and answers during the course teaching and the in-class tests. The preset difference U0 between the average interaction ratio and the preset interaction ratio is set to 0.2. The specific process of comparing the difference between the average interaction ratio and the preset interaction ratio with the preset difference is as follows:

[0065] If the difference U is less than or equal to the preset difference U0, the preset view count is adjusted to 0.76 times the original preset view count;

[0066] If the difference U is greater than the preset difference U0, the preset view count is adjusted to 0.57 times the original preset view count.

[0067] Please refer to Figure 4 As shown, it is a flowchart of the steps for the embodiments of the present invention to determine based on the absolute value of the difference between the view count at the previous moment and the current view count and the preset difference and the pre-stored preset absolute value. The process for the embodiments of the present invention to determine the reason why the recommended result does not meet the standard when the view count is less than or equal to the second preset view count includes: calculating the absolute value of the difference between the view count at the previous moment and the current view count and the preset difference, and determining the reason why the recommended result does not meet the standard based on the comparison result of the absolute value and the pre-stored preset absolute value. Among them: if the absolute value is less than or equal to the first preset absolute value, it is determined that there is a problem with the course package itself, and a processing notice is issued; if the absolute value is greater than the first preset absolute value and less than the second preset absolute value, the reason why the recommended result does not meet the standard is determined based on the variance of the historical view count; if the absolute value is greater than or equal to the second preset absolute value, it is determined that there is a problem with the setting of the preset adaptation degree, and the preset adaptation degree is adjusted based on the difference between the corresponding preset view count and the actual view count.

[0068] Specifically, in this embodiment, the absolute value W0 can be divided into the first preset absolute value W1 and the second preset absolute value W2. It is set that the first preset absolute value W1 = 100 in the absolute value standard, and the second preset absolute value W2 = 1.5 × W1. It should be noted that in other embodiments, the values of W1 and W2 can also be determined according to the needs of the online course. The specific process of comparing the absolute value with the pre-stored preset absolute value is as follows:

[0069] If the absolute value of W is less than or equal to the first preset absolute value W1, it indicates that the view volume at the previous moment and the current moment is low and the change is stable. Then, it is determined that there is a problem with the course package itself, and a processing notice is issued.

[0070] If the absolute value of W is greater than the first preset absolute value W1 and less than the second preset absolute value W2, it means that the reason for the recommended result not meeting the standard cannot be determined in this case. It is necessary to re-determine the reason for the recommended result not meeting the standard based on the variance G of the historical view volume.

[0071] If the absolute value of W is greater than or equal to the second preset absolute value W2, it is determined that there is a problem with the setting of the preset adaptation degree. Then, the preset adaptation degree is adjusted based on the difference H between the corresponding preset view volume and the actual view volume.

[0072] Specifically, the process of determining the reason for the recommended result not meeting the standard based on the variance of the historical view volume in the embodiments of the present invention includes: comparing the variance with the preset variance; if the variance is greater than the preset variance, a historical moment - historical view volume curve for the data packet is plotted, and the reason for the recommended result not meeting the standard is determined based on the characteristics of the historical moment - historical view volume curve; if the variance is less than or equal to the preset variance, it is determined that there is a problem with the course package itself, and a processing notice is issued.

[0073] Specifically, in this embodiment, the preset variance G0 = 0.94. The specific process of comparing the variance with the preset variance is as follows:

[0074] If the variance G is greater than the preset variance G0, it indicates that the numerical discreteness of the historical view volume is relatively high. Then, a historical moment - historical view volume curve is plotted, and the reason for the recommended result not meeting the standard is determined based on the characteristics of the historical moment - historical view volume curve.

[0075] If the variance G is less than or equal to the preset variance G0, it indicates that the numerical discreteness of the historical view volume is relatively low. It is determined that there is a problem with the course package itself, which may be a problem with the online courses in the course package, and a processing notice for reorganizing the courses in the course package is issued.

[0076] Specifically, the process of determining the characteristics of the curve based on the autocorrelation function in the embodiments of the present invention includes: determining the lag number; according to the lag number, calculating the autocorrelation value using the normalization method and plotting a lag number - autocorrelation value curve; counting the number of peaks of the curve and comparing the number of peaks with the lag number; if the number is greater than the lag number, it is determined that the curve has periodicity; if the number is less than or equal to the lag number, it is determined that the curve does not have periodicity.

[0077] Specifically, the autocorrelation function is the cross-correlation of a signal (or time series) with itself at different time points. It measures the similarity of the signal at two different time points as a function of the time difference between these two time points. Statistically, autocorrelation is defined as the Pearson correlation between the values at different times in two random processes.

[0078] Specifically, the process of determining the reason why the recommended result does not meet the standard according to the characteristics of the lag number-autocorrelation value curve in the embodiment of the present invention includes: determining the reason why the recommended result does not meet the standard based on whether the curve is periodic; if the lag number-autocorrelation value curve is periodic, it is determined that the reason is that the current course package is in the off-season of browsing, and then the browsing volume of the corresponding course package is continuously monitored; if the lag number-autocorrelation value curve is not periodic, the reason is determined according to the comparison result of the browsing volume at the previous moment and the current browsing volume; if the browsing volume at the previous moment is greater than the current browsing volume, it is determined that there is a problem with the course package itself, and a processing notice is issued; if the browsing volume at the previous moment is less than or equal to the current browsing volume, it is determined that the reason is that the current course package is in the off-season of browsing, and the browsing volume of the corresponding course package is continuously monitored.

[0079] Specifically, in this embodiment, the determination process based on whether the lag number-autocorrelation value curve is periodic is as follows:

[0080] If the lag number-autocorrelation value curve is periodic, it is determined that the current course package is in the off-season of user use, and then the browsing volume of the data packet is continuously monitored in the future;

[0081] If the lag number-autocorrelation value curve is not periodic, the browsing volume at the previous moment is compared with the current browsing volume, where:

[0082] If the browsing volume at the previous moment is greater than the current browsing volume, it means that the browsing volume is in a downward trend, and it is determined that there is a problem with the course package itself that causes this situation, and a processing notice for reorganizing the data in the data packet is issued;

[0083] If the browsing volume at the previous moment is less than or equal to the current browsing volume, it means that the browsing volume of the user for this course package is in a fluctuating situation, and the browsing volume of the data packet is continuously monitored in the future.

[0084] Specifically, the process of adjusting the preset adaptation degree based on the difference between the corresponding preset browsing volume and the actual browsing volume in the embodiment of the present invention includes: increasing the preset adaptation degree based on the difference, and the difference is proportional to the increase amplitude of the preset adaptation degree.

[0085] Specifically, in this embodiment, the preset difference H0 = 500. The specific process of comparing the difference H with the preset difference H0 is as follows:

[0086] If the difference H is less than or equal to the preset difference H0, then adjust the preset adaptability to 1.2 times the original preset adaptability;

[0087] If the difference H is greater than the preset difference H0, then adjust the preset adaptability to 1.7 times the original preset adaptability.

[0088] In order to make the purpose and advantages of the present invention more clear and understandable, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0089] Embodiment 1

[0090] This embodiment calculates the adaptability between the user and the course package. Among them, the demand tags for this user are as follows: deep learning, Python programming, AI technology frontier, object detection, object tracking; the preset tags for the course package are as follows: computer science, data analysis, object detection, Python programming, AI technology frontier. At this time, the demand tags of the user are compared with the preset tags of the course package. The same tags among them are the three tags of Python programming, AI technology frontier, and object detection. Therefore, the adaptability between the user and this course package is the number of the same tags between the user and the course package / the total number of the user's demand tags, that is, 3 / 5 = 0.6.

[0091] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0092] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An online course learning recommendation method based on a knowledge graph, characterized in that, including: obtaining user information and online course information; preprocessing the user information and the online course information respectively; inputting the preprocessed user information and the online course information into a knowledge graph, which packages the online courses into several course packages according to the characteristics of the online courses, outputs several preset labels for each course package and outputs several demand labels for the user according to the user information, wherein the preset label is the common feature of all the courses in the course package; calculating the fitness between the course package and the user's demands, wherein the fitness is the ratio of the number of the same labels in the preset labels of the course package and the demand labels of the user to the total number of the demand labels; recommending the course package to the user according to the fitness, and pushing the course package to the user when the fitness is higher than the preset fitness; judging whether the recommendation result of the online course meets the standard based on the view volume of each recommended course package, and determining the corresponding processing method based on the reason for not meeting the standard; adjusting the online course learning recommendation method based on the processing method.

2. The online course learning recommendation method based on a knowledge graph according to claim 1, wherein The process of judging whether the recommendation result of the online course meets the standard based on the view volume of each course package includes: judging whether the recommendation result meets the standard based on the comparison result between the view volume and the pre-stored preset view volume; if the view volume is greater than or equal to the first preset view volume, judging that the recommendation result meets the standard; if the view volume is less than the first preset view volume and greater than the second preset view volume, then judging whether the recommendation result meets the standard based on the cumulative learning duration of the user for the recommended course package, wherein the cumulative learning duration is the average duration for the users receiving the course package to learn the course; if the view volume is less than or equal to the second preset view volume, judging that the recommendation result does not meet the standard, and determining the reason for the recommendation result not meeting the standard based on the difference between the view volume at the previous moment and the current view volume.

3. The online course learning recommendation method based on a knowledge graph according to claim 2, wherein The process of judging whether the recommendation result meets the standard based on the cumulative learning duration of the user for the recommended course package includes: comparing the cumulative learning duration with the preset learning duration; if the cumulative learning duration is less than or equal to the preset learning duration, judging that the recommendation result does not meet the standard, and determining the reason for the recommendation result not meeting the standard based on the difference between the view volume at the previous moment and the current view volume; if the cumulative learning duration is greater than the preset learning duration, judging that the recommendation result meets the standard, and adjusting the preset view volume based on the actual usage ratio, wherein the actual usage ratio is the ratio of the number of users whose cumulative learning duration is greater than the preset learning duration to the total number of users receiving the push of the course package.

4. The online course learning recommendation method based on a knowledge graph according to claim 3, wherein The process of adjusting the preset view volume based on the actual usage ratio includes: calculating the difference between the actual usage ratio and the preset usage ratio; reducing the preset view volume based on the obtained difference, and the difference is proportional to the reduction amplitude of the preset view volume.

5. The online course learning recommendation method based on a knowledge graph according to claim 4, wherein The process of adjusting the preset view volume further includes: calculating the difference between the average interaction ratio of the user and the online course and the preset interaction ratio; reducing the preset view volume based on the obtained difference, and the difference is proportional to the reduction amplitude of the view volume.

6. The online course learning recommendation method based on a knowledge graph according to claim 2, wherein The process of determining the reason why the recommended result does not meet the standard when the view count is less than or equal to the second preset view count includes: Calculate the absolute value of the difference between the view count at the previous moment and the current view count and the preset difference, and determine the reason why the recommended result does not meet the standard based on the comparison result between the absolute value and the pre-stored preset absolute value, where: If the absolute value is less than or equal to the first preset absolute value, determine that there is a problem with the course package itself and issue a processing notice; If the absolute value is greater than the first preset absolute value and less than the second preset absolute value, determine the reason why the recommended result does not meet the standard based on the variance of the historical view count; If the absolute value is greater than or equal to the second preset absolute value, determine that there is a problem with the setting of the preset adaptation degree, and adjust the preset adaptation degree based on the difference between the corresponding preset view count and the actual view count.

7. The online course learning recommendation method based on a knowledge graph according to claim 6, wherein The process of determining the reason why the recommended result does not meet the standard based on the variance of the historical view count includes: Compare the variance with the preset variance; If the variance is greater than the preset variance, draw a historical moment - historical view count curve for the course package, and determine the reason why the recommended result does not meet the standard based on the characteristics of the historical moment - historical view count curve; If the variance is less than or equal to the preset variance, determine that there is a problem with the course package itself and issue a processing notice.

8. The online course learning recommendation method based on a knowledge graph according to claim 7, wherein The process of determining the characteristics of the curve based on the autocorrelation function includes: Determine the lag number; According to the lag number, calculate the autocorrelation value using the normalization method and draw a lag number - autocorrelation value curve; Count the number of peaks of the curve and compare the number of peaks with the lag number; If the number is greater than the lag number, determine that the curve is periodic; If the number is less than or equal to the lag number, determine that the curve is not periodic.

9. The online course learning recommendation method based on a knowledge graph according to claim 8, wherein The process of determining the reason why the recommended result does not meet the standard based on the characteristics of the lag number - autocorrelation value curve includes: Determine the reason why the recommended result does not meet the standard based on whether the curve is periodic; If the lag number - autocorrelation value curve is periodic, determine that the reason is that the course package is currently in the off-season for views, and continuously monitor the view count of the corresponding course package; If the lag number - autocorrelation value curve is not periodic, determine the reason based on the comparison result between the view count at the previous moment and the current view count; If the view count at the previous moment is greater than the current view count, determine that the reason is that there is a problem with the course package itself and issue a processing notice; If the view count at the previous moment is less than or equal to the current view count, determine that the reason is that the course package is currently in the off-season for views, and continuously monitor the view count of the corresponding course package.

10. The online course learning recommendation method based on a knowledge graph according to claim 6, characterized in that The process of adjusting the preset adaptation degree based on the difference between the corresponding preset view count and the actual view count includes: Increase the preset adaptation degree based on the difference, and the difference is proportional to the increase amplitude of the preset adaptation degree.

Images