A method, device, equipment and medium for recommending personalized content to students

Abstract

The present application provides a method, device, equipment and medium for recommending personalized content to students, and relates to the technical field of data processing. The method includes: obtaining a target data set and a preset classification dimension, wherein the target data set includes multiple student data; classifying the target data set according to the preset classification dimension to obtain multiple data groups; evaluating each data group according to a preset recommendation model to determine the performance index value of each data group; screening a first data group among multiple data groups, wherein the performance index value of the first data group meets a preset condition; analyzing the classification error rate of the first data group; if it is determined that the classification error rate is greater than or equal to a preset threshold, adjusting the target data set to obtain an adjusted data set, so as to train the preset recommendation model using the adjusted data set. The present application can identify and solve the recommendation bias problem of the recommendation model.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular to a method, device, equipment and medium for recommending personalized content to students. Background Art

[0002] With the rise of online education, students can easily learn various courses through the Internet, whether it is programming, language learning, artistic creation, or in-depth knowledge in professional fields, all can be obtained online. Online education breaks the time and space limitations of traditional education, allowing students to flexibly arrange learning time and content according to their interests and needs. At the same time, the application of personalized recommendation systems also makes learning resources richer and more accurate, helping students efficiently find the most suitable courses for themselves, and achieve personalized learning experience and comprehensive knowledge improvement.

[0003] The recommendation system for online courses may have recommendation bias when making recommendations. That is, the amount of data for some student groups is much smaller than that for other groups, which causes the recommendation algorithm to be more inclined to the groups with larger data during training, while ignoring the needs and preferences of the groups with smaller data. This data imbalance problem will cause the recommendation system to be unable to fairly provide personalized learning resources to all students, especially those students from minority preference groups, whose learning interests and needs may not be fully addressed and met. Therefore, a method is needed to identify and solve the recommendation bias problem of the recommendation model. Summary of the invention

[0004] The present application provides a method, apparatus, device and medium for recommending personalized content to students, which can identify and solve the recommendation bias problem of the recommendation model.

[0005] In a first aspect of the present application, a method for recommending personalized content to students is provided, the method comprising:

[0006] Obtaining a target data set and a preset classification dimension, wherein the target data set includes a plurality of student data;

[0007] Classifying the target data set according to the preset classification dimension to obtain multiple data groups;

[0008] Determine the performance index value of each data group according to the evaluation of each data group by a preset recommendation model;

[0009] Screening a first data group among the plurality of data groups, wherein a performance indicator value of the first data group meets a preset condition;

[0010] analyzing a classification error rate of the first data set;

[0011] If it is determined that the classification error rate is greater than or equal to a preset threshold, the target data set is adjusted to obtain an adjusted data set, so as to use the adjusted data set to train the preset recommendation model.

[0012] By adopting the above technical solution, the target data set is divided into multiple data groups through preset classification dimensions, and performance evaluation and classification error rate analysis are performed on each data group to identify the group that performs poorly in the recommendation process (i.e., the first data group). Then, these data groups are adjusted to balance the data distribution to retrain the recommendation model. This method ensures balanced representation of all student groups in model training, thereby reducing recommendation bias and improving the fairness and accuracy of the recommendation system.

[0013] Optionally, the evaluating each of the data groups according to a preset recommendation model to determine the performance indicator value of each of the data groups specifically includes:

[0014] Respectively count the first sample number, the second sample number, the third sample number and the fourth sample number output by the preset recommendation model for the data group, wherein the first sample number is the number of samples whose recommendation results are the same as the selection results, the second sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third sample number is the number of samples whose recommendation results are the same as the non-selected results, and the fourth sample number is the number of samples whose non-recommended results are the same as the selection results, the recommendation results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selection results are courses selected by students corresponding to the data group, and the non-selected results are courses not selected by students corresponding to the data group;

[0015] The performance indicator value is calculated according to the first sample quantity, the second sample quantity and the third sample quantity. The specific calculation formula is as follows:

[0016] ;

[0017] Among them, F is the performance indicator value, N1 is the first sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0018] By adopting the above technical solutions, the performance of the recommendation model on different data groups is statistically analyzed in detail, and the comprehensive performance index of each data group is calculated, so as to accurately evaluate the effect of the recommendation model on each student group. This index can help to identify which data groups (student groups) have performance problems in the recommendation process, especially those groups that are neglected due to small data volume.

[0019] Optionally, the evaluating each of the data groups according to a preset recommendation model to determine the performance indicator value of each of the data groups specifically includes:

[0020] The performance indicator value is calculated according to the first sample quantity, the second sample quantity, the third sample quantity and the fourth sample quantity. The specific calculation formula is as follows:

[0021] ;

[0022] Among them, F is the performance indicator value, N1 is the first sample number, N2 is the second sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0023] By using the above technical solution and calculating the comprehensive performance index value of each data group, the performance of the recommendation model in different student groups can be comprehensively evaluated. The specific formula takes into account the number of samples that are recommended correctly, not recommended correctly, recommended incorrectly, and not recommended incorrectly, thereby quantifying the overall accuracy of the model.

[0024] Optionally, analyzing the classification error rate of the first data group specifically includes:

[0025] Determine a first target sample number, a second target sample number, a third target sample number, and a fourth target sample number output by the preset recommendation model, wherein the first target sample number is the number of samples whose recommended results are identical to the selected results, the second target sample number is the number of samples whose non-recommended results are identical to the non-selected results, the third target sample number is the number of samples whose recommended results are identical to the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are identical to the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group;

[0026] The classification error rate is calculated according to the first target sample number, the second target sample number, the third target sample number, and the fourth target sample number. The specific calculation formula is as follows:

[0027] ;

[0028] Where D is the classification error rate, N1 ' is the first target sample size, N2 ' is the second target sample size, N3 ' is the third target sample size, N4 'is the fourth target sample size.

[0029] By adopting the above technical solution, the above technical solution effectively evaluates the misclassification of the recommendation model in each student group by detailed statistics and calculation of the classification error rate (D value) of each data group. The specific formula takes into account the number of samples with correct recommendations, incorrect recommendations, and incorrect non-recommendations, and quantifies the degree of misclassification of the model. This method can identify data groups with high classification error rates, reveal the biases and deficiencies of the model in these groups, and thus provide a basis for targeted adjustments and optimizations.

[0030] Optionally, if it is determined that the classification error rate is greater than or equal to a preset threshold, the target data set is adjusted to obtain an adjusted data set, so as to use the adjusted data set to train the preset recommendation model, specifically including:

[0031] Determine a first data amount of the first data group and a second data amount of a second data group, where the second data group is any one of the plurality of data groups except the first data group;

[0032] Generate target student data according to two adjacent student data in the first data group to obtain a first adjusted data group;

[0033] Delete multiple student data from the second data group to obtain a second adjusted data group, wherein the first adjusted data amount of the first adjusted data group is the same as the second adjusted data amount of the second adjusted data group, and the sum of the first adjusted data amount and the second adjusted data amount is the same as the sum of the first data amount and the second data amount.

[0034] By adopting the above technical solutions, the data volume differences in the data set are balanced, especially by generating new target student data and removing redundant data to adjust the data set to ensure balanced representation of all data groups. This adjustment helps to eliminate recommendation bias during model training and enhance the sensitivity and accuracy of the recommendation model for small data groups, thereby improving the fairness and overall performance of the recommendation system and ensuring that each student group can obtain more accurate and personalized learning resource recommendations.

[0035] Optionally, the screening of the first data group among the plurality of data groups specifically includes:

[0036] Determining whether the performance indicator value of the first data group is less than or equal to a preset threshold;

[0037] If it is determined that the performance indicator value is less than or equal to the preset threshold, the data group is determined to be the first data group.

[0038] By adopting the above technical solution, we can filter out data groups with performance index values ​​lower than or equal to the preset threshold, and effectively identify the student groups that perform poorly in the recommendation process. This method ensures that we can conduct in-depth analysis and adjustments for these problematic data groups, thereby discovering and correcting the biases and deficiencies of the recommendation system, and ultimately improving the overall fairness and accuracy of the recommendation model, ensuring that all student groups can receive high-quality personalized learning resource recommendations.

[0039] Optionally, the generating target student data according to two adjacent student data in the first data group to obtain the first adjusted data group is specifically calculated by the following formula:

[0040] ;

[0041] Among them, X n is the target student data, X i is the i-th student data in the first data group, λ is the preset random multiplier, X i+1 It is the i+1th student data in the first data group.

[0042] By adopting the above technical solutions, the diversity and representativeness of the data group are effectively increased, and the data set is balanced by synthesizing new data, thereby enhancing the learning ability of the recommendation model for groups with small data volumes, improving the overall fairness and performance of the model, and ensuring that all student groups can receive accurate and personalized recommendations.

[0043] In a second aspect of the present application, a device for recommending personalized content to students is provided, the device comprising an acquisition module, a classification module, a processing module, an analysis module and an adjustment module, wherein:

[0044] The acquisition module is used to acquire a target data set and a preset classification dimension, wherein the target data set includes a plurality of student data;

[0045] The classification module is used to classify the target data set according to the preset classification dimension to obtain multiple data groups;

[0046] The classification module is used to evaluate each of the data groups according to a preset recommendation model to determine the performance indicator value of each of the data groups;

[0047] The classification module is used to screen a first data group among the plurality of data groups, wherein a performance indicator value of the first data group meets a preset condition;

[0048] The analysis module is used to analyze the classification error rate of the first data group;

[0049] The adjustment module is used to adjust the target data set to obtain an adjusted data set if it is determined that the classification error rate is greater than or equal to a preset threshold, so as to use the adjusted data set to train the preset recommendation model.

[0050] Optionally, the classification module is used to respectively count a first sample number, a second sample number, a third sample number and a fourth sample number output by the preset recommendation model for the data group, wherein the first sample number is the number of samples whose recommended results are the same as the selected results, the second sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the data group, and the non-selected results are courses not selected by students corresponding to the data group;

[0051] The processing module is used to calculate the performance indicator value according to the first sample number, the second sample number and the third sample number. The specific calculation formula is as follows:

[0052] ;

[0053] Among them, F is the performance indicator value, N1 is the first sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0054] Optionally, the processing module is used to calculate the performance indicator value according to the first sample number, the second sample number, the third sample number and the fourth sample number. The specific calculation formula is as follows:

[0055] ;

[0056] Among them, F is the performance indicator value, N1 is the first sample number, N2 is the second sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0057] Optionally, the classification module is used to determine a first target sample number, a second target sample number, a third target sample number and a fourth target sample number output by the preset recommendation model, wherein the first target sample number is the number of samples whose recommended results are the same as the selected results, the second target sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third target sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group;

[0058] The processing module is used to calculate the classification error rate according to the first target sample number, the second target sample number, the third target sample number and the fourth target sample number. The specific calculation formula is as follows:

[0059] ;

[0060] Where D is the classification error rate, N1 ' is the first target sample size, N2 ' is the second target sample size, N3 ' is the third target sample size, N4 ' is the fourth target sample size.

[0061] Optionally, the adjustment module is used to determine a first data volume of the first data group and a second data volume of a second data group, where the second data group is any one of the plurality of data groups except the first data group;

[0062] The adjustment module is used to generate target student data according to two adjacent student data in the first data group to obtain a first adjusted data group;

[0063] The adjustment module is used to delete multiple student data from the second data group to obtain a second adjusted data group, wherein the first adjusted data amount of the first adjusted data group is the same as the second adjusted data amount of the second adjusted data group, and the sum of the first adjusted data amount and the second adjusted data amount is the same as the sum of the first data amount and the second data amount.

[0064] Optionally, the analysis module is used to determine whether the performance indicator value of the first data group is less than or equal to a preset threshold;

[0065] The analysis module is configured to determine that the data group is the first data group if it is determined that the performance indicator value is less than or equal to a preset threshold.

[0066] Optionally, the processing module is used to generate target student data according to two adjacent student data in the first data group to obtain a first adjusted data group, which is specifically calculated by the following formula:

[0067] ;

[0068] Among them, X n is the target student data, X i is the i-th student data in the first data group, λ is the preset random multiplier, X i+1 It is the i+1th student data in the first data group.

[0069] In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes any one of the methods described above.

[0070] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions, and when the instructions are executed, any of the methods described above is executed.

[0071] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0072] The target dataset is divided into multiple data groups by presetting the classification dimension, and the performance evaluation and classification error rate analysis are performed on each data group to identify the group that performs poorly in the recommendation process (i.e., the first data group). Then, these data groups are adjusted to balance the data distribution and retrain the recommendation model. This method ensures that all student groups are evenly represented in model training, thereby reducing recommendation bias and improving the fairness and accuracy of the recommendation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a flowchart of a method for recommending personalized content to students disclosed in an embodiment of the present application;

[0074] Figure 2 It is a module diagram of a student personalized content recommendation device disclosed in an embodiment of the present application;

[0075] Figure 3It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present application.

[0076] Explanation of the accompanying drawings: 201, acquisition module; 202, classification module; 203, processing module; 204, analysis module; 205, adjustment module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0077] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0078] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0079] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0080] With the rise of online education, students can flexibly and conveniently learn various courses through the Internet and enjoy the precise learning resources brought by personalized recommendation systems. However, online course recommendation systems may have recommendation bias problems. The data volume of some student groups is small, which causes the algorithm to be more inclined to the groups with larger data volumes during training, ignoring the needs and preferences of the groups with smaller data volumes. This data imbalance problem makes it impossible for the recommendation system to serve all students fairly, especially the learning interests and needs of minority groups may not receive sufficient attention. Therefore, it is crucial to identify and solve the recommendation bias problem in the recommendation model.

[0081] This embodiment discloses a method for recommending personalized content to students. Figure 1 , including the following steps S110-S170:

[0082] S110, obtaining a target data set and a preset classification dimension.

[0083] The method for recommending personalized content to students disclosed in the embodiment of the present application is applied to a server, wherein the server includes but is not limited to electronic devices such as mobile phones, tablet computers, wearable devices, PCs (Personal Computers), etc., and may also be a background server running a method for recommending personalized content to students. The server may be implemented as an independent server or a server cluster consisting of multiple servers.

[0084] Based on the online course sales data, the target data set about student online courses is obtained. The target data set includes multiple student data. Each student data includes basic information filled in by platform students on the platform, such as age, gender, educational background, interests and hobbies, etc. Student data also includes online courses selected by students on the platform, as well as courses related to the course learning process, etc. The preset classification dimensions are used to classify student data. The preset classification dimensions include age, gender, region, grade, major, education, learning frequency, etc.

[0085] S120, classifying the target data set according to a preset classification dimension to obtain multiple data groups.

[0086] Determine the specific classification criteria of the preset classification dimension. For example, when the preset classification dimension is age group, it includes 8-25 years old, 26-35 years old, 36-45 years old, etc. According to the preset classification dimension, divide the student data in the target data set into multiple data groups. Group the data set according to one dimension, such as age. Group all students who meet a certain age group into one group. Perform multi-dimensional combination grouping on the data set, such as a combination of age and gender. Each combination represents a data group, such as 18-25 years old males, 18-25 years old females, 26-35 years old males, etc. Traverse each record in the target data set and assign it to the corresponding data group according to the criteria of the preset classification dimension. Add each record to the corresponding data group according to the classification criteria. For example, if a record belongs to the 18-25 years old male group, add it to this group.

[0087] S130, evaluating each data group according to a preset recommendation model, and determining a performance indicator value for each data group.

[0088] In a possible implementation, the performance index value of each data group is determined according to the evaluation of each data group by the preset recommendation model, specifically including: respectively counting the first sample number, the second sample number, the third sample number and the fourth sample number output by the preset recommendation model for the data group, wherein the first sample number is the number of samples whose recommended results are the same as the selected results, the second sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by the students corresponding to the data group, and the non-selected results are courses not selected by the students corresponding to the data group; the performance index value is calculated according to the first sample number, the second sample number and the third sample number, and the specific calculation formula is as follows:

[0089] ;

[0090] Wherein, F is the performance index value, N1 is the first sample size, N3 is the third sample size, and N4 is the fourth sample size.

[0091] Specifically, recommendation algorithms often rely on a large amount of historical data for learning and prediction. During model training, if the data of different student groups are not balanced, the model will tend to favor student groups with better performance and more data, because the data of these student groups contribute more to the model. Suppose there is an online course platform whose user group consists of 80% male users and 20% female users. Without data balancing, the recommendation system is more likely to learn and optimize the preferences of male users during training, while the preferences of female users are not learned enough, resulting in lower quality recommendations for female users.

[0092] Cluster testing is an effective way to detect and evaluate the performance differences of recommendation systems on different student groups. By grouping student data into different groups and evaluating the performance of the model on each group separately, potential bias and unfairness can be discovered. The performance index value refers to the proportion of recommendation results of the model for samples of different categories, which is used to evaluate the advantages and disadvantages of the model.

[0093] For each data group, the recommendation model recommends online courses to different individual students based on the training results. After the recommendation, the courses actually selected by the individual students are counted. Depending on whether the model recommends certain courses to students and whether the students choose these courses, it will be divided into four situations, and four sample sizes can be obtained according to the four situations. The number of samples with the same recommendation result and the same selection result, that is, the courses recommended by the model are also the courses actually selected by the students. The number of samples obtained in this case is the first sample size. The number of samples with the same non-recommended result and the same non-selected result, that is, the courses not recommended by the model and the students did not choose the courses. The number of samples obtained in this case is the second sample size. The number of samples with the same recommendation result and the same non-selected result, that is, the courses recommended by the model but not selected by the students. The number of samples obtained in this case is the third sample size. The number of samples with the same non-recommended result and the same selection result, that is, the courses not recommended by the model but actually selected by the students. The number of samples obtained in this case is the fourth sample size.

[0094] First, we calculate the first ratio of the number of samples correctly predicted as positive by the model to the number of samples predicted as positive. The first ratio measures how much of the recommended courses are relevant. The first ratio is an important indicator when the recommendation system needs to avoid recommending irrelevant courses to students. For example, on a course recommendation platform, we hope that the recommended courses are as interesting as possible to improve the user experience.

[0095] The first ratio of the number of samples correctly predicted as positive by the model to the number of samples predicted as positive is calculated as follows:

[0096] ;

[0097] Wherein, P is the first ratio, N1 is the first sample size, and N3 is the third sample size.

[0098] Then calculate the second ratio of the number of recommended courses that students are actually interested in and click on to all courses that they are actually interested in. The second ratio reflects the extent to which the system covers the courses that students are interested in. The second ratio is a key indicator when the recommendation system needs to recommend as many courses as possible to students that they may be interested in. For example, on a learning platform, it is hoped that students can discover as many courses that they are interested in as possible so that they can better achieve their learning goals.

[0099] The calculation formula for the second ratio is as follows:

[0100] ;

[0101] Wherein, R is the second ratio, N1 is the first sample size, and N4 is the fourth sample size.

[0102] Finally, the performance index value is calculated based on the first ratio and the second ratio, which is also the harmonic mean of the first ratio and the second ratio, comprehensively considering the accuracy and coverage of the recommendation. The performance index value strikes a balance between the first ratio and the second ratio, and is suitable for scenarios where both recommendation accuracy and coverage need to be considered. In course recommendation, the performance index value is a comprehensive indicator used to evaluate the overall performance of the system in recommending relevant courses and covering courses of interest. For example, if a recommendation system needs to accurately recommend relevant courses and cover all courses of interest to students as much as possible, the performance index value is a good evaluation indicator.

[0103] The performance index value is calculated as follows:

[0104] ;

[0105] Wherein, F is the performance index value, P is the first ratio, and R is the second ratio.

[0106] In a possible implementation, the performance index value of each data group is determined according to the evaluation of each data group by the preset recommendation model, specifically including: calculating the performance index value according to the first sample number, the second sample number, the third sample number and the fourth sample number, and the specific calculation formula is as follows:

[0107] ;

[0108] Among them, F is the performance index value, N1 is the first sample number, N2 is the second sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0109] Specifically, another performance indicator value provided by the embodiment of the present application is the ratio of all correctly predicted course recommendations to the total number of recommendations, which is used to reflect the overall recommendation accuracy of the system, but it may not be sufficient to fully evaluate the system performance in course recommendations, especially when the number of recommended courses and the number of non-recommended courses are unbalanced. For example, if most courses are not recommended, then the accuracy will still be high even if the system recommends almost no courses.

[0110] S140, screening a first data group among the multiple data groups.

[0111] In a possible implementation, screening a first data group among multiple data groups specifically includes: determining whether a performance indicator value of the first data group is less than or equal to a preset threshold; if it is determined that the performance indicator value is less than or equal to the preset threshold, determining that the data group is the first data group.

[0112] Specifically, a preset threshold of a performance indicator value is defined in advance, and the threshold is used to determine whether each data group meets the performance requirements. If the performance indicator value of the data group can meet the performance requirements, there will be no bias problem when the model makes subsequent recommendations. On the contrary, if the performance indicator value of the data group cannot meet the performance requirements, it is necessary to further verify the first data group to determine whether the model recommendation may have a recommendation bias problem.

[0113] S150, analyzing the classification error rate of the first data group.

[0114] In a possible implementation, analyzing the classification error rate of the first data group specifically includes: determining a first target sample number, a second target sample number, a third target sample number, and a fourth target sample number output by a preset recommendation model, wherein the first target sample number is the number of samples whose recommended results are the same as the selected results, the second target sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third target sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group; calculating the classification error rate based on the first target sample number, the second target sample number, the third target sample number, and the fourth target sample number, the specific calculation formula is as follows:

[0115] ;

[0116] Among them, D is the classification error rate, N1 ' is the first target sample size, N2 ' is the second target sample size, N3 ' is the third target sample size, N4 ' is the fourth target sample size.

[0117] Specifically, for the first data group whose screened performance indicator values ​​cannot meet the performance requirements, four target sample sizes are determined: the sample size where the recommended result is the same as the selected result, that is, the course recommended by the model is also the course actually selected by the student. The sample size obtained in this case is the first target sample size. The sample size where the non-recommended result is the same as the non-selected result, that is, the course not recommended by the model and the student did not select the course. The sample size obtained in this case is the second target sample size. The sample size where the recommended result is the same as the non-selected result, that is, the course recommended by the model but not selected by the student. The sample size obtained in this case is the third target sample size. The sample size where the non-recommended result is the same as the selected result, that is, the course not recommended by the model but actually selected by the student. The sample size obtained in this case is the fourth target sample size.

[0118] The classification error rate directly measures the errors in the model's recommendation results. By calculating the ratio of N3 (false positives) to N4 (false negatives), we can understand the model's specific performance in terms of wrong recommendations (recommending courses that should not be recommended) and missed recommendations (not recommending courses that should be recommended). It helps evaluate the performance of the recommendation model, especially the degree of its misclassification. By analyzing the classification error rate, we can understand the performance of the model on different data groups and make corresponding adjustments and optimizations.

[0119] S160: If it is determined that the classification error rate is greater than or equal to the preset threshold, the target data set is adjusted to obtain an adjusted data set.

[0120] In a possible implementation, if it is determined that the classification error rate is greater than or equal to a preset threshold, the target data set is adjusted to obtain an adjusted data set, so as to use the adjusted data set to train the preset recommendation model, specifically including: determining a first data amount of the first data group, and a second data amount of the second data group, the second data group being any one of the multiple data groups except the first data group; generating target student data based on two adjacent student data in the first data group to obtain a first adjusted data group; deleting multiple student data from the second data group to obtain a second adjusted data group, wherein the first adjusted data amount of the first adjusted data group is the same as the second adjusted data amount of the second adjusted data group, and the sum of the first adjusted data amount and the second adjusted data amount is the same as the sum of the first data amount and the second data amount.

[0121] Specifically, first, the data volume of the first data group and the second data group needs to be determined. The first data group is a data group with a relatively low data volume and a recommendation bias problem. The second data group is a data group other than the first data group among multiple data groups, and the second data group is a data group with a relatively high data volume. In order to solve the problem of recommendation bias of the model due to different data volumes of different data groups, the data volume of different data groups needs to be adjusted to achieve a balance between the different data groups.

[0122] The number of student data in the first data group is counted to obtain the first data volume, and the number of student data in the second data group is counted to obtain the second data volume. According to two adjacent student data in the first data group, the target student data is generated to obtain the first adjusted data group; at the same time, multiple student data are deleted from the second data group to obtain the second adjusted data group. The method of generating target student data usually uses data enhancement technology, such as SMOTE (synthetic minority oversampling technology), which first selects two adjacent student data from the first data group, and generates new student data through interpolation methods or other synthesis technologies.

[0123] For any minority class sample X in the first data set i , select a sample from the nearest neighbors, preferably the next student data sample, to generate new target student data as follows:

[0124] ;

[0125] Among them, X n is the target student data, X i is the i-th student data in the first data group, λ is the preset random multiplier, X i+1 It is the i+1th student data in the first data group.

[0126] In SMOTE (Synthetic Minority Oversampling Technology), the preset random multiplier is a random number with a value range of [0, 1]. It plays a key role in generating new synthetic samples. Specifically, the preset random multiplier controls the position of the new sample between the original sample and its neighboring samples. When λ=0, X n =X i , the new sample is equal to the original sample. When λ=1, X n =X i+1 , the new sample is equivalent to the neighboring sample X i+1 When λ is between 0 and 1, X n It is in X i and X i+1 Somewhere in between.

[0127] In this way, SMOTE is able to generate new samples that are evenly distributed between the original samples and their neighboring samples, thereby increasing the diversity of the minority class and reducing the risk of overfitting.

[0128] Delete redundant student data from the second data group to balance the adjusted data volume, calculate the number of student data that needs to be deleted, and make the data volume of the first adjusted data group and the second adjusted data group equal. In short, the first data group is adjusted to obtain the first adjusted data group, and the second data group is adjusted to obtain the second adjusted data group. The data volume of the first adjusted data group is the first adjusted data volume, and the data volume of the second adjusted data group is the second adjusted data volume. Use the adjusted data set to retrain the preset recommendation model to improve the performance of the model on different data groups and reduce the classification error rate. The first adjusted data volume and the second adjusted data volume meet the following requirements: the first adjusted data volume of the first adjusted data group is the same as the second adjusted data volume of the second adjusted data group, and the sum of the first adjusted data volume and the second adjusted data volume is the same as the sum of the first data volume and the second data volume.

[0129] For example, the first data set contains 100 student data, and the second data set contains 200 student data. 50 new student data are generated from the first data set to obtain a first adjusted data set. 50 student data are deleted from the second data set to obtain a second adjusted data set. Thus, the first adjusted data amount of the first adjusted data set is 150, and the second adjusted data amount of the second adjusted data set is 150.

[0130] By adopting the technical solution of the present application, the target data set is divided into multiple data groups by preset classification dimensions, and performance evaluation and classification error rate analysis are performed on each data group, so as to identify the group with poor performance in the recommendation process (i.e., the first data group). Then, by adjusting these data groups and balancing the data distribution, the recommendation model is retrained. This method ensures balanced representation of all student groups in model training, thereby reducing recommendation bias and improving the fairness and accuracy of the recommendation system.

[0131] This embodiment also discloses a device for recommending personalized content to students, which includes an acquisition module, a classification module, a processing module, an analysis module, and an adjustment module, wherein:

[0132] The acquisition module is used to obtain the target data set and preset classification dimensions, and the target data set includes multiple student data.

[0133] The classification module is used to classify the target data set according to the preset classification dimension to obtain multiple data groups.

[0134] The classification module is used to evaluate each data group according to the preset recommendation model and determine the performance indicator value of each data group.

[0135] The classification module is used to screen a first data group among multiple data groups, wherein a performance indicator value of the first data group meets a preset condition.

[0136] The analysis module is used to analyze the classification error rate of the first data group.

[0137] The adjustment module is used to adjust the target data set if it is determined that the classification error rate is greater than or equal to a preset threshold, so as to obtain an adjusted data set, so as to train a preset recommendation model using the adjusted data set.

[0138] In one possible implementation, a classification module is used to respectively count a first sample number, a second sample number, a third sample number and a fourth sample number output by a preset recommendation model for a data group, wherein the first sample number is the number of samples whose recommended results are the same as the selected results, the second sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the data group, and the non-selected results are courses not selected by students corresponding to the data group.

[0139] The processing module is used to calculate the performance indicator value according to the first sample quantity, the second sample quantity and the third sample quantity. The specific calculation formula is as follows:

[0140] ;

[0141] Wherein, F is the performance index value, N1 is the first sample size, N3 is the third sample size, and N4 is the fourth sample size.

[0142] In a possible implementation manner, the processing module is used to calculate the performance indicator value according to the first sample quantity, the second sample quantity, the third sample quantity and the fourth sample quantity. The specific calculation formula is as follows:

[0143] ;

[0144] Among them, F is the performance index value, N1 is the first sample number, N2 is the second sample number, N3 is the third sample number, and N4 is the fourth sample number.

[0145] In one possible implementation, a classification module is used to determine a first target sample number, a second target sample number, a third target sample number and a fourth target sample number output by a preset recommendation model, wherein the first target sample number is the number of samples whose recommended results are the same as the selected results, the second target sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third target sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group.

[0146] The processing module is used to calculate the classification error rate according to the first target sample number, the second target sample number, the third target sample number and the fourth target sample number. The specific calculation formula is as follows:

[0147] ;

[0148] Among them, D is the classification error rate, N1 ' is the first target sample size, N2 ' is the second target sample size, N3 ' is the third target sample size, N4 ' is the fourth target sample size.

[0149] In a possible implementation, the adjustment module is used to determine a first data amount of the first data group and a second data amount of the second data group, where the second data group is any one of the multiple data groups except the first data group.

[0150] The adjustment module is used to generate target student data according to two adjacent student data in the first data group to obtain a first adjusted data group.

[0151] An adjustment module is used to delete multiple student data from the second data group to obtain a second adjusted data group, wherein the first adjusted data amount of the first adjusted data group is the same as the second adjusted data amount of the second adjusted data group, and the sum of the first adjusted data amount and the second adjusted data amount is the same as the sum of the first data amount and the second data amount.

[0152] In a possible implementation, the analysis module is used to determine whether the performance indicator value of the first data group is less than or equal to a preset threshold.

[0153] The analysis module is used to determine that the data group is the first data group if it is determined that the performance indicator value is less than or equal to a preset threshold.

[0154] In a possible implementation, the processing module is used to generate target student data according to two adjacent student data in the first data group to obtain the first adjusted data group, which is specifically calculated by the following formula:

[0155] ;

[0156] Among them, X n is the target student data, X i is the i-th student data in the first data group, λ is the preset random multiplier, X i+1 It is the i+1th student data in the first data group.

[0157] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0158] This embodiment also discloses an electronic device, referring to Figure 3 The electronic device may include: at least one processor 301 , at least one communication bus 302 , a user interface 303 , a network interface 304 , and at least one memory 305 .

[0159] The communication bus 302 is used to realize the connection and communication between these components.

[0160] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0161] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0162] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0163] The memory 305 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may also be at least one storage device located away from the aforementioned processor 301. The memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface 303 module, and an application program for a method for recommending personalized content to students.

[0164] exist Figure 3In the electronic device shown, the user interface 303 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 301 can be used to call an application program storing a method for recommending personalized content to students in the memory 305. When executed by one or more processors 301, the electronic device executes one or more methods in the above-mentioned embodiments.

[0165] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0166] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0167] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0168] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0169] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0170] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory 305 and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory 305 includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk.

[0171] The present application also discloses a computer-readable storage medium, wherein the computer-readable storage medium stores instructions. When executed by one or more processors 301, the electronic device executes one or more of the methods described in the above embodiments.

[0172] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, it will be easy for those skilled in the art to think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for recommending personalized content to students, characterized in that: The method comprises: Obtaining a target data set and a preset classification dimension, wherein the target data set includes a plurality of student data; Classifying the target data set according to the preset classification dimension to obtain multiple data groups; Determine the performance index value of each data group according to the evaluation of each data group by a preset recommendation model; Screening a first data group among the plurality of data groups, wherein a performance indicator value of the first data group meets a preset condition; analyzing a classification error rate of the first data set; If it is determined that the classification error rate is greater than or equal to a preset threshold, the target data set is adjusted to obtain an adjusted data set, so as to use the adjusted data set to train the preset recommendation model; The evaluation of each of the data groups according to the preset recommendation model to determine the performance indicator value of each of the data groups specifically includes: respectively counting the first sample number, the second sample number, the third sample number and the fourth sample number output by the preset recommendation model for the data group, and calculating the performance indicator value according to the first sample number, the second sample number and the third sample number; wherein the first sample number is the number of samples whose recommended results are the same as the selected results, the second sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are the courses output by the preset recommendation model, the non-recommended results are the courses not output by the preset recommendation model, the selected results are the courses selected by the students corresponding to the data group, and the non-selected results are the courses not selected by the students corresponding to the data group; The analyzing the classification error rate of the first data group specifically includes: Determine the first target sample number, the second target sample number, the third target sample number and the fourth target sample number output by the preset recommendation model, and calculate the classification error rate according to the first target sample number, the second target sample number, the third target sample number and the fourth target sample number; wherein the first target sample number is the number of samples whose recommended results are the same as the selected results, the second target sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third target sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group.

2. A method for recommending personalized content to students according to claim 1, characterized in that: The performance indicator value is calculated according to the first sample quantity, the second sample quantity and the third sample quantity. The specific calculation formula is as follows: Among them, F is the performance indicator value, N1 is the first sample number, N3 is the third sample number, and N4 is the fourth sample number.

3. A method for recommending personalized content to students according to claim 2, characterized in that: The step of evaluating each of the data groups according to the preset recommendation model to determine the performance indicator value of each of the data groups specifically includes: The performance indicator value is calculated according to the first sample quantity, the second sample quantity, the third sample quantity and the fourth sample quantity. The specific calculation formula is as follows: Among them, F is the performance indicator value, N1 is the first sample number, N2 is the second sample number, N3 is the third sample number, and N4 is the fourth sample number.

4. A method for recommending personalized content to students according to claim 1, characterized in that: The classification error rate is calculated according to the first target sample number, the second target sample number, the third target sample number, and the fourth target sample number. The specific calculation formula is as follows: Wherein, D is the classification error rate, N1' is the first target sample number, N2' is the second target sample number, N3' is the third target sample number, and N4' is the fourth target sample number.

5. A method for recommending personalized content to students according to claim 2, characterized in that: If it is determined that the classification error rate is greater than or equal to a preset threshold, the target data set is adjusted to obtain an adjusted data set, and the preset recommendation model is trained using the adjusted data set, specifically including: Determine a first data amount of the first data group and a second data amount of a second data group, where the second data group is any one of the plurality of data groups except the first data group; Generate target student data according to two adjacent student data in the first data group to obtain a first adjusted data group; Delete multiple student data from the second data group to obtain a second adjusted data group, wherein the first adjusted data amount of the first adjusted data group is the same as the second adjusted data amount of the second adjusted data group, and the sum of the first adjusted data amount and the second adjusted data amount is the same as the sum of the first data amount and the second data amount.

6. A method for recommending personalized content to students according to claim 1, characterized in that: The screening of the first data group among the plurality of data groups specifically includes: Determining whether the performance indicator value of the first data group is less than or equal to a preset threshold; If it is determined that the performance indicator value is less than or equal to the preset threshold, the data group is determined to be the first data group.

7. A method for recommending personalized content to students according to claim 5, characterized in that: The target student data is generated according to two adjacent student data in the first data group to obtain the first adjustment data group, which is specifically calculated by the following formula: X n =X i +λ×(X i+1 -X i ) Among them, X n is the target student data, X i is the i-th student data in the first data group, λ is the preset random multiplier, X i+1 It is the i+1th student data in the first data group.

8. A personalized content recommendation device for students, characterized in that: The device comprises an acquisition module, a classification module, a processing module, an analysis module and an adjustment module, wherein: The acquisition module is used to acquire a target data set and a preset classification dimension, wherein the target data set includes a plurality of student data; The classification module is used to classify the target data set according to the preset classification dimension to obtain multiple data groups; The classification module is used to determine the performance indicator value of each data group according to the evaluation of each data group by the preset recommendation model; specifically comprising: respectively counting the first sample number, the second sample number, the third sample number and the fourth sample number output by the preset recommendation model for the data group, and calculating the performance indicator value according to the first sample number, the second sample number and the third sample number; wherein the first sample number is the number of samples with the same recommended result and the same selected result, the second sample number is the number of samples with the same non-recommended result and the same non-selected result, the third sample number is the number of samples with the same recommended result and the same non-selected result, and the fourth sample number is the number of samples with the same non-recommended result and the same selected result, the recommended result is the course output by the preset recommendation model, the non-recommended result is the course not output by the preset recommendation model, the selected result is the course selected by the students corresponding to the data group, and the non-selected result is the course not selected by the students corresponding to the data group; The classification module is used to screen a first data group among the plurality of data groups, wherein a performance indicator value of the first data group meets a preset condition; The analysis module is used to analyze the classification error rate of the first data group, specifically including: Determine a first target sample number, a second target sample number, a third target sample number and a fourth target sample number output by the preset recommendation model, and calculate the classification error rate according to the first target sample number, the second target sample number, the third target sample number and the fourth target sample number; wherein the first target sample number is the number of samples whose recommended results are the same as the selected results, the second target sample number is the number of samples whose non-recommended results are the same as the non-selected results, the third target sample number is the number of samples whose recommended results are the same as the non-selected results, and the fourth target sample number is the number of samples whose non-recommended results are the same as the selected results, the recommended results are courses output by the preset recommendation model, the non-recommended results are courses not output by the preset recommendation model, the selected results are courses selected by students corresponding to the first data group, and the non-selected results are courses not selected by students corresponding to the first data group; The adjustment module is used to adjust the target data set to obtain an adjusted data set if it is determined that the classification error rate is greater than or equal to a preset threshold, so as to use the adjusted data set to train the preset recommendation model.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface, a network interface and a memory, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images