A scenario optimization method based on teacher-student ARG game social platform

Abstract

The present invention relates to the field of big data analysis technology, and in particular to a method for optimizing a scenario based on a teacher-student ARG game social platform, the method comprising: obtaining learning dimension data of each target student in each preset time period under each preset dimension, randomly combining all preset dimensions, and clustering the learning dimension data under each data dimension combination; determining a learning state characterization index for each target student for each data dimension combination; determining an overall state characterization index for all target students for each data dimension combination; screening out a target dimension combination according to the overall state characterization index and a target clustering cluster; optimizing the teacher-student ARG game social platform scenario according to the target dimension combination. The present invention realizes adaptive dimensionality reduction of multiple preset dimensions by analyzing a large amount of learning dimension data, thereby improving the rationality of optimizing the teacher-student ARG game social platform.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a scene optimization method based on a teacher-student ARG game social platform. Background Art

[0002] In campus life, teachers can use the teacher-student ARG (Alternate Reality Gaming) game social platform to design classroom interactive activities, such as setting up ARG game tasks or puzzles related to the course content, allowing students to learn in the game and cooperate with classmates to solve problems, stimulating students' interest in learning. At the same time, they can also develop ARG game tasks related to knowledge of various subjects on the teacher-student ARG game social platform, allowing students to learn textbook knowledge through games. Since the learning status of different students is often different, the teacher-student ARG game social platform can be used to adaptively recommend suitable ARG games for students with different learning statuses to optimize the teacher-student ARG game social platform. Specifically, according to the data of different students using the teacher-student ARG game social platform under different preset dimensions, all students can be clustered using a clustering algorithm to obtain student groups with different learning situations, thereby recommending suitable ARG games to student groups with different learning situations to optimize the teacher-student ARG game social platform. Among them, the preset dimension can be a dimension related to the ARG game recommendation. The clustering algorithm can realize data mining.

[0003] However, clustering algorithms may have poor performance in the process of clustering analysis of high-dimensional data samples. For example, common clustering methods, such as the DBSCAN algorithm, have poor performance in the process of clustering analysis of high-dimensional data samples. The learning data of students on the teacher-student ARG game social platform is usually composed of multiple preset dimensions, which leads to poor mining ability of the DBSCAN algorithm, resulting in poor rationality in optimizing the teacher-student ARG game social platform. Therefore, it is crucial to reduce the dimensionality of data under multiple preset dimensions.

[0004] At present, the commonly used dimensionality reduction method is to use principal component analysis to reduce data dimensionality. However, when principal component analysis is used to reduce the data dimensionality of students' learning data on the teacher-student ARG game social platform, since principal component analysis often assumes that the data is linearly related, and there may be nonlinear relationships between the data of multiple preset dimensions in the learning data of students on the teacher-student ARG game social platform, principal component analysis may not be able to capture these relationships, resulting in the loss of important information in the learning data of students on the teacher-student ARG game social platform after dimensionality reduction, which in turn leads to poor rationality in optimizing the teacher-student ARG game social platform. Summary of the invention

[0005] In order to solve the technical problem of poor rationality in optimizing the teacher-student ARG game social platform, the present invention proposes a scene optimization method based on the teacher-student ARG game social platform.

[0006] The present invention provides a scene optimization method based on a teacher-student ARG game social platform, the method comprising:

[0007] Obtain the learning dimension data of each target student in each preset time period under each preset dimension, randomly combine all preset dimensions to obtain different data dimension combinations, and cluster the learning dimension data under each data dimension combination to obtain each target clustering cluster;

[0008] According to all learning dimension data of each target student in each data dimension combination and the information richness of the learning dimension data in each data dimension combination, determine the learning status representation index of each target student for each data dimension combination;

[0009] According to the learning status representation index of each data dimension combination for all target students and the learning influence degree under each data dimension combination, determine the overall status representation index of each data dimension combination for all target students, wherein the learning influence degree represents the overall influence of the preset dimensions in the data dimension combination on the main learning conditions of all target students;

[0010] According to the overall state representation index of the data dimension combination and all target clusters under the data dimension combination, the target dimension combination is screened out from all data dimension combinations;

[0011] According to the target dimension combination, the teacher-student ARG game social platform scenario is optimized.

[0012] Optionally, clustering the learning dimension data under each data dimension combination to obtain each target cluster cluster includes:

[0013] Determine any data dimension combination as a tag dimension combination;

[0014] The learning dimension data of each target student in the same preset time period under all preset dimensions in the marked dimension combination are used to form the target data points of each target student under the marked dimension combination in the preset time period, thereby obtaining a target data point set under the marked dimension combination;

[0015] Density clustering is performed on all target data points in the target data point set, and each obtained cluster is determined as a target cluster, thereby obtaining each target cluster under the label dimension combination.

[0016] Optionally, determining the learning status characterization index of each target student for each data dimension combination according to all learning dimension data of each target student in each data dimension combination and the information richness of the learning dimension data in each data dimension combination includes:

[0017] According to the data distribution information entropy under all preset dimensions in each data dimension combination, the information richness under each data dimension combination is determined, wherein the data distribution information entropy is positively correlated with the information richness, and the information richness under the data dimension combination represents the information richness of the learning dimension data under the data dimension combination;

[0018] Based on all learning dimension data of each target student in each data dimension combination and the information richness of each data dimension combination, the learning status characterization indicators of each data dimension combination for each target student are determined.

[0019] Optionally, determining the learning status characterization index of each target student for each data dimension combination according to all learning dimension data of each target student in each data dimension combination and the information richness of each data dimension combination includes:

[0020] Determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination;

[0021] Determine the correlation coefficient between each two preset dimensions of the marked student according to all learning dimension data of each two preset dimensions in the marked dimension combination, wherein the correlation coefficient between the two preset dimensions represents the information association between the learning dimension data under the two preset dimensions;

[0022] According to the correlation coefficient between all preset dimensions of the marked student in the marked dimension combination, determine the initial representation factor of the marked dimension combination for the marked student, wherein the correlation coefficient is negatively correlated with the initial representation factor, and the initial representation factor of the marked dimension combination for the marked student represents the information difference between the learning dimension data of the marked student under different preset dimensions in the marked dimension combination;

[0023] Perform data change trend analysis and processing based on all learning dimension data of the marked students in each preset dimension to obtain the data change trend of the marked students in each preset dimension;

[0024] Filter out label dimensions that represent student performance from all preset dimensions;

[0025] Determine a reference representation factor of the marked dimension combination for the marked student according to the difference between the data change trend of the marked student under the label dimension and the data change trend under all preset dimensions in the marked dimension combination;

[0026] According to the information richness of the marking dimension combination, the initial representation factor and reference representation factor of the marking dimension combination for the marked students, the representation index of the learning status of the marked students by the marking dimension combination is determined.

[0027] Optionally, the formula corresponding to the data dimension combination for the target student's learning status representation indicator is:

[0028] w ij =norm(F j × ij1 × ij2 );

[0029]

[0030] Among them, w ij is the learning status representation index of the j-th data dimension combination for the i-th target student; j is the serial number of the data dimension combination; i is the serial number of the target student; norm() is the normalization function; F j is the information richness of the jth data dimension combination; w ij1 and w ij2 are the initial representation factor and reference representation factor of the j-th data dimension combination for the i-th target student, respectively; n j is the number of preset dimensions in the jth data dimension combination; a and c are the serial numbers of different preset dimensions in the jth data dimension combination; γ ijac is the correlation coefficient between the a-th preset dimension and the c-th preset dimension for the i-th target student in the j-th data dimension combination; f(γ ijac ) is the independent variable with value γ ijac The function value when , and γ ijac With f(γ ijac ) is negatively correlated; f(|k i -k ija |) is the independent variable with the value of |k i -k ija |, and |k i -k ija |with f(|k i -k ija |) is negatively correlated; || is the absolute value function; k i is the data change trend of the i-th target student in the label dimension; k ija It is the data change trend of the i-th target student under the a-th preset dimension in the j-th data dimension combination.

[0031] Optionally, determining the overall status representation index of each data dimension combination for all target students according to the learning status representation index of each data dimension combination for all target students and the learning influence degree of each data dimension combination includes:

[0032] The cumulative value of the learning status representation index of each data dimension combination for all target students is determined as the initial overall factor of each data dimension combination for all target students;

[0033] Perform learning status impact analysis based on all learning dimension data of all target students in each data dimension combination to obtain the degree of learning impact in each data dimension combination;

[0034] According to the degree of learning impact under each data dimension combination and the initial overall factor of each data dimension combination for all target students, the overall state representation index of each data dimension combination for all target students is determined, among which the degree of learning impact and the initial overall factor are positively correlated with the overall state representation index.

[0035] Optionally, performing a learning status impact analysis process on all learning dimension data of all target students in each data dimension combination to obtain a learning impact degree in each data dimension combination includes:

[0036] Determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination;

[0037] Determine the initial influence degree of the marked student under each preset dimension in the marked dimension combination according to all the learning dimension data of the marked student under the marked dimension combination, wherein the initial influence degree represents the importance of the learning dimension data of the marked student under the preset dimension in the learning dimension data of the marked student in the marked dimension combination;

[0038] The learning impact degree under each data dimension combination is determined based on all initial impact degrees of all target students under each data dimension combination, where the initial impact degree and the learning impact degree are positively correlated.

[0039] Optionally, determining the initial influence degree of the marked student in each preset dimension in the marked dimension combination according to all learning dimension data of the marked student in the marked dimension combination includes:

[0040] The contribution degree of the principal component of each preset dimension of the marked student in the marked dimension combination is obtained as the initial influence degree of the marked student in each preset dimension in the marked dimension combination.

[0041] Optionally, the method of screening out a target dimension combination from all data dimension combinations according to the overall state characterization index of the data dimension combination and all target clusters under the data dimension combination includes:

[0042] Filter out label dimensions that represent student performance from all preset dimensions;

[0043] The variance of the label dimension data corresponding to all target data points in each target cluster is determined as the label variance corresponding to each target cluster, wherein the label dimension data is the learning dimension data under the label dimension;

[0044] According to the overall state representation index of the label dimension combination for all target students and the label variance corresponding to all target clusters under the label dimension combination, determine the information mining index under the label dimension combination, wherein the overall state representation index is positively correlated with the information mining index, and the label variance is negatively correlated with the information mining index;

[0045] The data dimension combination with the largest information mining index is selected from all data dimension combinations as the target dimension combination.

[0046] Optionally, optimizing the teacher-student ARG game social platform scenario according to the target dimension combination includes:

[0047] The pre-acquired learning dimension data of each target student in the current time period under all preset dimensions in the target dimension combination are used to form a current data point corresponding to each target student;

[0048] According to the current data points corresponding to all target students, all target students are clustered to obtain the current clusters;

[0049] According to the current data points corresponding to all target students in each current cluster, ARG game content is recommended to all target students in each current cluster through the teacher-student ARG game social platform, thereby optimizing the teacher-student ARG game social platform scenario.

[0050] The present invention has the following beneficial effects:

[0051] The present invention is based on a scenario optimization method of the teacher-student ARG game social platform. By analyzing a large amount of learning dimension data, the adaptive dimensionality reduction of multiple preset dimensions is realized, and the technical problem of poor rationality of optimizing the teacher-student ARG game social platform is solved, thereby improving the rationality of optimizing the teacher-student ARG game social platform. When recommending ARG games to students through the teacher-student ARG game social platform, the pre-designed ARG games are often uniformly recommended to the students. However, the learning situations of different students are often different, so directly recommending the pre-designed ARG games uniformly to the students may result in poor rationality of ARG game recommendations. Therefore, it is often necessary to optimize the teacher-student ARG game social platform to achieve adaptive ARG game recommendations for students. Secondly, the present invention screens out a target dimension combination that can accurately characterize the learning situation of students, and realizes adaptive dimensionality reduction of multiple preset dimensions, so that the clustering effect in the optimization process of the teacher-student ARG game social platform can be improved to a certain extent, and then the rationality of optimizing the teacher-student ARG game social platform can be improved. Moreover, compared with the use of principal component analysis for data dimensionality reduction, the present invention comprehensively considers multiple indicators related to students' learning situation when screening target dimension combinations for data dimensionality reduction, such as learning impact degree, information richness, learning state characterization indicators and overall state characterization indicators, thereby improving the accuracy of determining the target dimension combination, improving the effect of data dimensionality reduction to a certain extent, and further improving the rationality of optimizing the teacher-student ARG game social platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 It is a flow chart of a scene optimization method based on a teacher-student ARG game social platform of the present invention;

[0054] Figure 2 A flowchart of a method for determining a learning state characterization indicator of the present invention;

[0055] Figure 3 The flowchart of the method for determining the degree of learning influence of the present invention. DETAILED DESCRIPTION

[0056] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0057] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0058] refer to Figure 1 , shows the process of some embodiments of a method for optimizing a scene based on a teacher-student ARG game social platform according to the present invention. The method for optimizing a scene based on a teacher-student ARG game social platform includes the following steps:

[0059] Step S1, obtain the learning dimension data of each target student in each preset time period under each preset dimension, randomly combine all preset dimensions to obtain different data dimension combinations, and cluster the learning dimension data under each data dimension combination to obtain each target cluster cluster.

[0060] Among them, the target students may be students who use the teacher-student ARG game social platform for learning. The preset dimension may be a pre-set dimension related to the learning type to be recommended. The learning type to be recommended may be a learning type recommended by the ARG game content related to the learning to be carried out. For example, the learning type to be recommended may be a learning subject type, and the learning subject type may be, but is not limited to, mathematics, Chinese and English. The preset dimension may be, but is not limited to, a task completion percentage dimension, an interaction frequency dimension, an average test score dimension and a grade level dimension. The preset time period may be a preset time period. For example, the duration corresponding to the preset time period may be one week. The number of preset time periods may be a preset number, for example, the number of preset time periods may be 8. The end time of the previous preset time period in the adjacent preset time period may be the start time of the next preset time period. The end time of the latest preset time period may be the current time or the historical time. The historical time is before the current time. If the end time of the latest preset time period is the current time, the latest preset time period may be recorded as the current time period. At this time, all preset time periods may constitute the time period to which the recent stage belongs, for example, the time period constituted by all preset time periods may be the last two months. The learning dimension data may be a preset dimension value. A data dimension combination can have at least two preset dimensions.

[0061] The teacher-student ARG game social platform is a platform that combines augmented reality technology and social networking functions, aiming to provide users with immersive gaming experience and social interaction. In campus life, teachers can use the ARG game social platform to design classroom interactive activities. For example, they can set up ARG game tasks or puzzles related to the course content, allowing students to learn in the game and cooperate with classmates to solve problems, stimulating students' interest in learning. At the same time, they can also develop ARG game tasks related to knowledge of various subjects on the ARG game social platform, allowing students to learn textbook knowledge through games, etc.

[0062] It should be noted that obtaining the learning dimension data of each target student in each preset time period under each preset dimension can facilitate the subsequent analysis of the learning situation of each target student, and can facilitate the subsequent adaptive recommendation of ARG games for each target student. And all preset dimensions are randomly combined to obtain multiple different data dimension combinations, which can facilitate the subsequent screening of data dimension combinations suitable for dimensionality reduction from multiple different data dimension combinations. Secondly, determining multiple target clustering clusters under the data dimension combination can facilitate the subsequent verification of whether different learning situations can be divided when the dimension is reduced to the data dimension combination, that is, whether different target clustering clusters can represent different learning situations, so as to facilitate the subsequent adaptive ARG game recommendation for target students in different learning situations in the current time period.

[0063] As an example, this step may include the following steps:

[0064] The first step is to obtain the learning dimension data of each target student in each preset time period under each preset dimension.

[0065] Since the method for obtaining the learning dimension data of each target student in each preset time period under each preset dimension is the same, here, obtaining the learning dimension data of a target student in a preset time period under the task completion percentage dimension is taken as an example, and if the recommended learning type is mathematics, then the method for obtaining the learning dimension data of the target student in the preset time period under the task completion percentage dimension can be specifically: obtaining the mathematics-related workload that the target student needs to complete in the preset time period as the first workload; obtaining the mathematics-related workload completed by the target student in the preset time period as the second workload; and determining the ratio of the second workload to the first workload as the learning dimension data of the target student in the preset time period under the task completion percentage dimension.

[0066] The second step is to randomly combine all preset dimensions to obtain different data dimension combinations.

[0067] The data dimension combination may be composed of different preset dimensions. For example, the preset dimensions in the data dimension combination may be obtained by non-repetitive sampling. The number of preset dimensions in the data dimension combination may be greater than or equal to 2 and less than or equal to a preset number. The preset number may be a preset number, for example, the preset number may be equal to half of the total number of all preset dimensions.

[0068] For example, if there are three preset dimensions, namely the task completion percentage dimension, the interaction frequency dimension and the average test score dimension, and if the number of preset dimensions in the data dimension combination is 2, then three data dimension combinations can be obtained, namely (task completion percentage dimension, interaction frequency dimension), (task completion percentage dimension, average test score dimension) and (interaction frequency dimension, average test score dimension).

[0069] It should be noted that the purpose of obtaining the learning dimension data of each target student in each preset time period under each preset dimension is mainly to facilitate the subsequent screening of the most suitable data dimension combination for dimensionality reduction from different data dimension combinations. Therefore, the learning dimension data collected at this time, whether in the past time period or in the current stage, does not affect the subsequent screening process. Therefore, the preset time period can be a historical time period or a current time period. Secondly, the target student can take an exam related to the recommended learning type at least once in each preset time period, and can obtain at least one test score. Furthermore, when there are more preset dimensions, the screening effect of screening out the most suitable data dimension combination for dimensionality reduction from different data dimension combinations is relatively better. Since dimensionality reduction is performed, the number of preset dimensions in the data dimension combination can be less than the total number of all preset dimensions.

[0070] The third step is to determine any data dimension combination as a label dimension combination.

[0071] The fourth step is to combine the learning dimension data of each target student in the same preset time period under all preset dimensions in the above-mentioned marked dimension combination to form the target data points of each target student under the marked dimension combination within the preset time period, and obtain the target data point set under the above-mentioned marked dimension combination.

[0072] Among them, the target data points under the marking dimension combination of the target student within a preset time period may include: the numerical values ​​corresponding to all preset dimensions in the marking dimension combination of the target student within the preset time period. For example, if the marking dimension combination includes: the task completion percentage dimension, the interaction frequency dimension and the average test score dimension, then the method for obtaining the target data points under the marking dimension combination of a target student within a preset time period may be: the three-dimensional coordinate point composed of the task completion percentage, interaction frequency and average test score of the target student within the preset time period is recorded as the target data point under the marking dimension combination of the target student within the preset time period. The set of target data points under the marking dimension combination may include: the target data points of all target students under the marking dimension combination in all preset time periods.

[0073] In the fifth step, a density clustering algorithm can be used to perform density clustering on all target data points in the above target data point set, and each obtained cluster is determined as a target cluster, thereby obtaining each target cluster under the above label dimension combination.

[0074] Among them, all target clusters under the label dimension combination may include: all clusters obtained by density clustering all target data points in the target data point set under the label dimension combination. The density clustering algorithm may be, but is not limited to, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, the OPTICS (Ordering Points To Identify the Clustering Structure) algorithm, and the DENCLUE (DENsity-based Clustering of Applications with Noise) algorithm.

[0075] It should be noted that, since the learning situation of the same target student in different time periods may be different, the target data points formed by the same target student in different preset time periods may be divided into different target clusters.

[0076] Step S2, determining the learning status characterization index of each target student for each data dimension combination based on all learning dimension data of each target student under each data dimension combination and the information richness of the learning dimension data under each data dimension combination.

[0077] It should be noted that the greater the index representing the learning status of the target student for the quantitative data dimension combination, the better the data dimension combination can represent the learning situation of the target student, and the better the effect of dimensionality reduction to this data dimension combination.

[0078] As an example, this step may include the following steps:

[0079] The first step is to determine the information richness of each data dimension combination based on the data distribution information entropy under all preset dimensions in each data dimension combination.

[0080] Among them, the information richness under the data dimension combination can represent the information richness of the learning dimension data under the data dimension combination. The data distribution information entropy can be positively correlated with the information richness.

[0081] It should be noted that the greater the information richness of a quantitative data dimension combination, the better the data dimension combination can represent the student's learning situation, and the better the effect of dimensionality reduction to this data dimension combination.

[0082] For example, determining the information richness of each combination of data dimensions may include the following sub-steps:

[0083] In the first sub-step, based on all the learning dimension data of all the target students in each preset dimension, the corresponding occurrence probability of each different learning dimension data under each preset dimension is determined.

[0084] Among them, all learning dimension data of all target students under any preset dimension may include: all values ​​under the preset dimension collected in all preset time periods. For example, taking the task completion percentage dimension as an example, all learning dimension data of all target students under the task completion percentage dimension may include: all task completion percentages collected in all preset time periods. The corresponding probability of occurrence of learning dimension data may be: the proportion of this kind of learning dimension data in all learning dimension data under the preset dimension to which it belongs. For example, if there is 1 task completion percentage equal to 80%, and there are 8 task completion percentages collected in all preset time periods, then the corresponding probability of occurrence of the task completion percentage equal to 80% may be 1 / 8.

[0085] In the second sub-step, the data distribution information entropy under each preset dimension is determined according to the corresponding occurrence probabilities of all different learning dimension data under each preset dimension.

[0086] In the third sub-step, according to the information entropy of data distribution, the formula corresponding to the information richness under the combination of data dimensions can be determined as follows:

[0087] Among them, F j is the information richness of the jth data dimension combination. j is the sequence number of the data dimension combination. j is the number of preset dimensions in the jth data dimension combination. a is the sequence number of the preset dimension in the jth data dimension combination. Na is the number of all different learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination. μ is the sequence number of all different learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination. a A is the number of all learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination. aμ It is the number of μ-th different learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination. It is based on b The logarithm of . b is a preset value greater than 1, for example, b can be 2. It is the data distribution information entropy under the a-th preset dimension in the j-th data dimension combination. It is the probability of occurrence of the μth different learning dimension data under the ath preset dimension in the jth data dimension combination.

[0088] It should be noted that when When F is larger, it often means that the distribution of learning dimension data under the a-th preset dimension is more chaotic, the uncertainty of learning dimension data under the a-th preset dimension is higher, the amount of information contained in the a-th preset dimension is richer, and the a-th preset dimension is more likely to obtain more information in the process of data mining. Therefore, when F j When it is larger, it often means that most of the preset dimensions in the j-th data dimension combination are more likely to obtain more information in the process of data mining, and it often means that the j-th data dimension combination can better represent the students' learning situation.

[0089] Optionally, the method for determining the information richness under a data dimension combination may also include: determining the variance of all learning dimension data under each preset dimension as the target variance under each preset dimension; and determining the mean of the target variances under all preset dimensions in each data dimension combination as the information richness under each data dimension combination.

[0090] The second step is to determine the learning status characterization indicators of each target student for each data dimension combination based on all learning dimension data of each target student under each data dimension combination and the information richness of each data dimension combination.

[0091] For example, the flowchart for determining the data dimension combination for each target student's learning status representation indicator can be as follows: Figure 2 As shown, the following steps may be specifically included:

[0092] Step 201, determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination.

[0093] Step 202, based on all learning dimension data of the marked students under every two preset dimensions in the marked dimension combination, determine the correlation coefficient between every two preset dimensions of the marked students.

[0094] Among them, the correlation coefficient between two preset dimensions can characterize the information association between the learning dimension data under the two preset dimensions. The method for obtaining the correlation coefficient between each two preset dimensions of the marked students can be: the two preset dimensions are respectively recorded as the first preset dimension and the second preset dimension, and all the learning dimension data of the marked students under the first preset dimension are formed into a first sequence; and all the learning dimension data of the marked students under the second preset dimension are formed into a second sequence; the normalized value of the Pearson correlation coefficient between the first sequence and the second sequence is determined as the correlation coefficient between the two preset dimensions of the marked students.

[0095] Step 203, determining an initial characterization factor of the marked dimension combination for the marked student according to the correlation coefficients among all preset dimensions in the marked dimension combination of the marked student.

[0096] The correlation coefficient is negatively correlated with the initial representation factor. The initial representation factor of the above-mentioned marking dimension combination for the above-mentioned marked student represents the information difference between the learning dimension data of the above-mentioned marked student under different preset dimensions in the above-mentioned marking dimension combination.

[0097] Step 204 , performing data change trend analysis and processing based on all learning dimension data of the marked students in each preset dimension, to obtain the data change trend of the marked students in each preset dimension.

[0098] Among them, the method for obtaining the data change trend of the marked students under the preset dimension can be: perform straight line fitting on all learning dimension data of the marked students under the preset dimension, and use the slope of the fitting straight line obtained at this time as the data change trend of the marked students under the preset dimension.

[0099] Step 205, filter out label dimensions that characterize student performance from all preset dimensions.

[0100] Among them, the grade level or average test score dimension under the grade level dimension can represent the student's performance, so the grade level dimension or average test score dimension can be used as a label dimension. Among them, the higher the average test score, the higher the grade level. For example, when the average test score is less than 60 points, the grade level can be recorded as 0; when the average test score is higher than or equal to 60 points and less than 80 points, the grade level can be recorded as 1; when the average test score is higher than or equal to 80 points, the grade level can be recorded as 2.

[0101] It should be noted that student performance is an important factor in characterizing the student's learning status. Therefore, all preset dimensions obtained often include dimensions that characterize the student's performance.

[0102] Step 206 , determining a reference characterization factor of the marked dimension combination for the marked student according to the difference between the data change trend of the marked student under the label dimension and the data change trend under all preset dimensions in the marked dimension combination.

[0103] Step 207, determining a characterization index of the learning status of the marked student by the marked dimension combination according to the information richness of the marked dimension combination, the initial characterization factor of the marked student by the marked dimension combination, and the reference characterization factor.

[0104] For example, the formula for determining the data dimension combination corresponding to the target student's learning status representation indicator can be:

[0105] w ij =norm(f j × ij1 × ij2 );

[0106]

[0107] Among them, w ij is the learning status characterization index of the jth data dimension combination for the ith target student. j is the serial number of the data dimension combination. i is the serial number of the target student. norm() is the normalization function. F j is the information richness of the jth data dimension combination. ij1 and w ij2 are the initial characterization factor and reference characterization factor of the j-th data dimension combination for the i-th target student. j is the number of preset dimensions in the jth data dimension combination. a and c are the serial numbers of different preset dimensions in the jth data dimension combination. γ ijac is the correlation coefficient between the a-th preset dimension and the c-th preset dimension for the i-th target student in the j-th data dimension combination. γ ijac The value range of f(γ ijac ) is the independent variable with value γ ijac The function value when , and γ ijac With f(γ ijac ) is negatively correlated. f(|k i -k ija |) is the independent variable with the value |k i -k ija |, and |k i -kija |with f(|k i -k ija |) is negatively correlated. For example, β1 and β2 are pre-set factors greater than 0, mainly used to prevent the denominator from being 0. For example, β1 and β2 can both be 0.001. || is the absolute value function. k i is the data change trend of the i-th target student in the label dimension. k ija It is the data change trend of the i-th target student under the a-th preset dimension in the j-th data dimension combination.

[0108] It should be noted that when F j When γ is larger, it often means that the j-th data dimension combination can better represent the student’s learning situation. ijac The smaller it is, the smaller the correlation between the a-th preset dimension and the c-th preset dimension of the i-th target student is, which often means that the learning dimension data under the a-th preset dimension and the c-th preset dimension are less similar, which often means that the learning dimension data under the a-th preset dimension and the c-th preset dimension are more likely to complement each other in student learning, and the a-th preset dimension and the c-th preset dimension are more important. Therefore, when w ij1 When |k is larger, it often means that the learning dimension data of the i-th target student under different preset dimensions in the j-th data dimension combination are more likely to complement each other's learning situation, and it often means that the effect of dimensionality reduction to the j-th data dimension combination is better. Since the label dimension can represent the student's academic performance, and the student's academic performance is often an important factor in representing the student's learning situation, the change of the learning dimension data under the label dimension can often represent the main change of the student's learning situation. i -k ija The smaller | is, the closer the data change trend of the i-th target student in the a-th preset dimension in the j-th data dimension combination is to the data change trend in the label dimension. Therefore, when w ij2 When w is larger, it often means that the data changes of the i-th target student under different preset dimensions in the j-th data dimension combination are closer to the main changes in the student's learning situation, and it often means that the j-th data dimension combination can better represent the student's learning situation. ij The larger it is, the more likely it is that the j-th data dimension combination can better represent the learning situation of the i-th target student.

[0109] Step S3, determining the overall status representation index of each data dimension combination for all target students based on the learning status representation index of each data dimension combination for all target students and the degree of learning influence under each data dimension combination.

[0110] Among them, the degree of learning impact can represent the overall impact of the preset dimensions in the data dimension combination on the main learning situations of all target students.

[0111] It should be noted that the larger the overall status representation index of all target students for the quantitative data dimension combination is, the better the data dimension combination can represent the students' learning situation, and the better the effect of dimensionality reduction to this data dimension combination is.

[0112] As an example, this step may include the following steps:

[0113] In the first step, the cumulative value of the learning status characterization indicators of each data dimension combination for all target students is determined as the initial overall factor for each data dimension combination for all target students.

[0114] The second step is to conduct a learning status impact analysis based on all learning dimension data of all target students in each data dimension combination to obtain the degree of learning impact in each data dimension combination.

[0115] It should be noted that the greater the degree of learning impact under the quantitative data dimension combination, the more this data dimension combination can represent the students' learning situation, and the better the effect of dimensionality reduction to this data dimension combination.

[0116] For example, a flowchart for determining the degree of learning impact under a combination of data dimensions can be as follows: Figure 3 As shown, the following steps may be specifically included:

[0117] Step 301, determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination.

[0118] Step 302, based on all learning dimension data of the marked student under the marked dimension combination, determine the initial influence degree of the marked student under each preset dimension in the marked dimension combination.

[0119] The initial impact degree may represent the importance of the learning dimension data of the marked student under the preset dimension in the learning dimension data of the marked student in the marked dimension combination. All learning dimension data of the marked student under the marked dimension combination may include: learning dimension data of the marked student under all preset dimensions in all preset time periods in the marked dimension combination.

[0120] For example, the contribution degree of the principal component of each preset dimension of the marked student under the marked dimension combination can be obtained as the initial influence degree of the marked student under each preset dimension in the marked dimension combination, which can specifically include: according to all the learning dimension data of the marked student under the marked dimension combination, through principal component analysis, the contribution degree of the learning dimension data of the marked student under each preset dimension in the marked dimension combination to the principal component is obtained as the initial influence degree of the marked student under each preset dimension in the marked dimension combination. The principal component at this time is the principal component obtained by principal component analysis of all the learning dimension data of the marked student under the marked dimension combination, and the principal component can often characterize the main learning situation of the student. Among them, the contribution degree is also called the influence degree or contribution.

[0121] Step 303, determining the learning influence degree under each data dimension combination according to all initial influence degrees of all target students under each data dimension combination.

[0122] Among them, the initial impact degree and the learning impact degree are positively correlated.

[0123] For example, the formula for determining the degree of learning impact under a combination of data dimensions can be:

[0124]

[0125] Among them, Q j is the learning influence degree under the jth data dimension combination. j is the sequence number of the data dimension combination. j is the number of preset dimensions in the jth data dimension combination. a is the sequence number of the preset dimension in the jth data dimension combination. Q ja is the dimensional influence degree of the a-th preset dimension in the j-th data dimension combination. n is the number of target students. i is the sequence number of the target student. ija It is the initial influence degree of the i-th target student under the a-th preset dimension in the j-th data dimension combination.

[0126] It should be noted that the principal component can often represent the main learning situation of students. ija When Q is larger, it often means that the a-th preset dimension of the i-th target student contributes more to the principal component of the j-th data dimension combination, which often means that the a-th preset dimension has a greater impact on the learning situation of the i-th target student. ja The larger the value, the greater the impact of the a-th preset dimension on the learning situation of most target students. jWhen it is larger, it often means that the preset dimension in the j-th data dimension combination has a greater impact on the learning situation of most target students, and it often means that the j-th data dimension combination can better represent the students' learning situation.

[0127] Optionally, the data dimension combinations can be preliminarily screened by the degree of learning influence, specifically: the total number of data dimension combinations is recorded as the first total number, and half of the first total number is determined as the reference number, and the reference number of data dimension combinations with the largest learning influence are screened out to participate in the subsequent steps, and the reference number of data dimension combinations with the smallest learning influence can be removed. That is, it is considered that the reference number of data dimension combinations with the smallest learning influence are often not the target dimension combinations that need to be screened, and can be excluded from the subsequent steps.

[0128] The third step is to determine the overall status representation indicators of each data dimension combination for all target students based on the degree of learning impact under each data dimension combination and the initial overall factor of each data dimension combination for all target students.

[0129] Among them, the degree of learning influence and the initial overall factor can be positively correlated with the overall state representation index.

[0130] For example, the formula for determining the data dimension combination corresponding to the overall status representation index of all target students can be:

[0131] Among them, L j is the overall status characterization index of the jth data dimension combination for all target students. j is the sequence number of the data dimension combination. norm() is the normalization function. Q j is the learning impact degree under the jth data dimension combination. n is the number of target students. i is the sequence number of the target student. w ij It is the indicator representing the learning status of the i-th target student by the j-th data dimension combination. is the initial overall factor of the j-th data dimension combination for all target students.

[0132] It should be noted that when Q j When w is larger, it often means that the j-th data dimension combination can better represent the student’s learning situation. ij The larger the value, the better the j-th data dimension combination can represent the learning situation of the i-th target student. j The larger it is, the better the j-th data dimension combination can represent the student's learning situation, and the better the effect of dimensionality reduction to the j-th data dimension combination.

[0133] Step S4, based on the overall state characterization index of the data dimension combination and all target clusters under the data dimension combination, a target dimension combination is screened out from all data dimension combinations.

[0134] It should be noted that the larger the information mining index under the quantitative data dimension combination, the better the data dimension combination can represent the student's learning situation, and the better the effect of dimensionality reduction to the data dimension combination. In addition, the learning dimension data under the selected target dimension combination can often more accurately represent the student's learning situation, realizing adaptive data dimensionality reduction.

[0135] As an example, this step may include the following steps:

[0136] The first step is to filter out the label dimensions that represent the students' performance from all preset dimensions.

[0137] In the second step, the variance of the label dimension data corresponding to all target data points in each target cluster is determined as the label variance corresponding to each target cluster.

[0138] The label dimension data may be learning dimension data under the label dimension. The label dimension data corresponding to the target data point may be obtained by: the label dimension collected by the target student to which the target data point belongs within the preset time period to which the target data point belongs.

[0139] For example, if the first target data point is the target data point of a target student under a labeled dimension combination within a preset time period, and the label dimension is a grade level dimension, then the label dimension data corresponding to the first target data point can be the grade level of the target student within the preset time period.

[0140] The third step is to determine the information mining index under the above-mentioned label dimension combination based on the overall status characterization index of all target students and the label variance corresponding to all target clusters under the above-mentioned label dimension combination.

[0141] Among them, the overall state representation index can be positively correlated with the information mining index, and the label variance can be negatively correlated with the information mining index.

[0142] For example, the formula for determining the information mining index under the data dimension combination can be:

[0143] Among them, H j is the information mining index under the jth data dimension combination. j is the sequence number of the data dimension combination. j G is the overall status representation indicator of the j-th data dimension combination for all target students. jis the number of all target clusters under the jth data dimension combination. g is the ordinal number of all target clusters under the jth data dimension combination. β3 is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0. For example, β3 can be 0.001. σ jg It is the label variance corresponding to the g-th target cluster under the j-th data dimension combination.

[0144] It should be noted that when L j When σ is larger, it often means that the j-th data dimension combination can better represent the student's learning situation, and it often means that the effect of dimension reduction to the j-th data dimension combination is better. jg The smaller it is, the less chaotic the distribution of label dimension data corresponding to the target data point in the g-th target cluster under the j-th data dimension combination is, the better the clustering effect of the g-th target cluster under the j-th data dimension combination is, and the stronger its data mining ability is. Therefore, when When H is larger, it often means that the clustering effect of different target clusters under the jth data dimension combination is better, it often means that the jth data dimension combination is more likely to classify the same learning situation into the same target cluster, it often means that the jth data dimension combination has a stronger ability to describe the student's learning situation, it often means that the jth data dimension combination can better represent the student's learning situation, and it often means that the effect of dimensionality reduction to the jth data dimension combination is better. Therefore, when H j The larger it is, the better the j-th data dimension combination can represent the student's learning situation, and the better the effect of dimensionality reduction to the j-th data dimension combination.

[0145] The fourth step is to select the data dimension combination with the largest information mining index from all data dimension combinations as the target dimension combination.

[0146] Step S5, optimizing the teacher-student ARG game social platform scenario according to the target dimension combination.

[0147] As an example, this step may include the following steps:

[0148] In the first step, the pre-acquired learning dimension data of each target student in the current time period under all preset dimensions in the above target dimension combination are used to form the current data point corresponding to each target student.

[0149] For example, if the target dimension combination includes: task completion percentage dimension, interaction frequency dimension and average test score dimension, the method for obtaining the current data point corresponding to the target student can be: record the three-dimensional coordinate point composed of the target student's task completion percentage, interaction frequency and average test score in the current time period as the current data point corresponding to the target student.

[0150] The second step is to cluster all target students according to the current data points corresponding to all target students to obtain the current clusters.

[0151] Among them, different current clusters can represent student groups with different learning status.

[0152] For example, the DBSCAN algorithm can be used to perform density clustering on all current data points, and each cluster obtained at this time is used as a reference cluster, and the target students corresponding to all current data points in each reference cluster constitute a current cluster.

[0153] The third step is to recommend ARG game content to all target students in each current cluster through the teacher-student ARG game social platform according to the current data points corresponding to all target students in each current cluster, and to optimize the teacher-student ARG game social platform scenario by: determining the mean of the label dimension data corresponding to the current data points corresponding to all target students in each current cluster as the label representative value corresponding to each current cluster; sorting all current clusters in descending order of the label representative values ​​to obtain the current cluster sequence, and dividing the current cluster sequence into a preset number of subsequences; dividing the ARG game content related to the learning type to be recommended into a preset number of categories according to the difficulty; and recommending the ARG game content of the preset number of categories to the target students in the preset number of subsequences in order from difficult to easy, so that the teacher-student ARG game social platform can adaptively recommend appropriate ARG game content to each target student, thereby optimizing the teacher-student ARG game social platform scenario. The preset number may be a pre-set number, for example, the preset number may be 2. The same content may be recommended to target students in the same subsequence. When the label representative value corresponding to the current cluster in the subsequence is larger, more difficult ARG game content can be recommended to the target students in the subsequence.

[0154] In summary, the present invention screens out a target dimension combination that can accurately characterize the student learning situation, and realizes adaptive dimensionality reduction of multiple preset dimensions, thereby improving the clustering effect in the optimization process of the teacher-student ARG game social platform to a certain extent, and further improving the rationality of optimizing the teacher-student ARG game social platform. Moreover, compared with the use of principal component analysis for data dimensionality reduction, the present invention comprehensively considers multiple indicators related to the student learning situation when screening the target dimension combination for data dimensionality reduction, such as the degree of learning impact, information richness, learning state characterization indicators and overall state characterization indicators, thereby improving the accuracy of the target dimension combination determination, improving the effect of data dimensionality reduction to a certain extent, and further improving the rationality of optimizing the teacher-student ARG game social platform.

[0155] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A scenario optimization method based on a teacher-student ARG game social platform, characterized in that: The following steps are involved: Obtain the learning dimension data of each target student in each preset time period under each preset dimension, randomly combine all preset dimensions to obtain different data dimension combinations, and cluster the learning dimension data under each data dimension combination to obtain each target clustering cluster; According to all learning dimension data of each target student in each data dimension combination and the information richness of the learning dimension data in each data dimension combination, determine the learning status representation index of each target student for each data dimension combination; According to the learning status representation index of each data dimension combination for all target students and the learning influence degree under each data dimension combination, determine the overall status representation index of each data dimension combination for all target students, wherein the learning influence degree represents the overall influence of the preset dimensions in the data dimension combination on the main learning conditions of all target students; According to the overall state representation index of the data dimension combination and all target clusters under the data dimension combination, the target dimension combination is screened out from all data dimension combinations; According to the target dimension combination, the teacher-student ARG game social platform scenario is optimized; The step of determining the learning status characterization index of each target student for each data dimension combination according to all learning dimension data of each target student in each data dimension combination and the information richness of the learning dimension data in each data dimension combination includes: According to the data distribution information entropy under all preset dimensions in each data dimension combination, the information richness under each data dimension combination is determined, wherein the data distribution information entropy is positively correlated with the information richness, and the information richness under the data dimension combination represents the information richness of the learning dimension data under the data dimension combination; According to all learning dimension data of each target student in each data dimension combination and the information richness of each data dimension combination, determine the learning status representation index of each target student for each data dimension combination; The step of determining the learning status characterization index of each target student for each data dimension combination according to all learning dimension data of each target student in each data dimension combination and the information richness of each data dimension combination includes: Determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination; Determine the correlation coefficient between each two preset dimensions of the marked student according to all learning dimension data of each two preset dimensions in the marked dimension combination, wherein the correlation coefficient between the two preset dimensions represents the information association between the learning dimension data under the two preset dimensions; According to the correlation coefficient between all preset dimensions of the marked student in the marked dimension combination, determine the initial representation factor of the marked dimension combination for the marked student, wherein the correlation coefficient is negatively correlated with the initial representation factor, and the initial representation factor of the marked dimension combination for the marked student represents the information difference between the learning dimension data of the marked student under different preset dimensions in the marked dimension combination; Perform data change trend analysis and processing based on all learning dimension data of the marked students in each preset dimension to obtain the data change trend of the marked students in each preset dimension; Filter out label dimensions that represent student performance from all preset dimensions; Determine a reference representation factor of the marked dimension combination for the marked student according to the difference between the data change trend of the marked student under the label dimension and the data change trend under all preset dimensions in the marked dimension combination; According to the information richness of the marking dimension combination, the initial representation factor of the marking dimension combination for the marked student and the reference representation factor, the representation index of the learning state of the marked student by the marking dimension combination is determined; The formula corresponding to the data dimension combination for the target student's learning status representation indicator is: ； ； ;in, is the learning status representation index of the j-th data dimension combination for the i-th target student; j is the serial number of the data dimension combination; i is the serial number of the target student; is the normalization function; is the information richness of the jth data dimension combination; and are the initial representation factor and reference representation factor of the j-th data dimension combination for the i-th target student respectively; is the number of preset dimensions in the jth data dimension combination; a and c are the serial numbers of different preset dimensions in the jth data dimension combination; is the correlation coefficient between the a-th preset dimension and the c-th preset dimension for the i-th target student in the j-th data dimension combination; The independent variable takes the value The function value when , and and Negatively correlated; The independent variable takes the value The function value when , and and Negatively correlated; It is the absolute value function; is the data change trend of the i-th target student in the label dimension; is the data change trend of the i-th target student under the a-th preset dimension in the j-th data dimension combination; ; ; and It is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0; ;in, is the number of all different learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination; It is the serial number of all different learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination; is the number of all learning dimension data of all target students under the a-th preset dimension in the j-th data dimension combination; It is the number of all target students in the jth data dimension combination under the ath preset dimension. The number of different learning dimension data; It is based on b The logarithm of ; b is a preset value greater than 1; It is the data distribution information entropy under the a-th preset dimension in the j-th data dimension combination; It is the jth data dimension combination under the ath preset dimension. The occurrence probability corresponding to different learning dimension data.

2. A scenario optimization method based on a teacher-student ARG game social platform according to claim 1, characterized in that: The clustering of the learning dimension data under each data dimension combination to obtain each target clustering cluster includes: Determine any data dimension combination as a tag dimension combination; The learning dimension data of each target student in the same preset time period under all preset dimensions in the marked dimension combination are used to form the target data points of each target student under the marked dimension combination in the preset time period, thereby obtaining a target data point set under the marked dimension combination; Density clustering is performed on all target data points in the target data point set, and each obtained cluster is determined as a target cluster, thereby obtaining each target cluster under the label dimension combination.

3. A scenario optimization method based on a teacher-student ARG game social platform according to claim 1, characterized in that: Determining the overall status representation index of each data dimension combination for all target students according to the learning status representation index of each data dimension combination for all target students and the learning influence degree of each data dimension combination includes: The cumulative value of the learning status representation index of each data dimension combination for all target students is determined as the initial overall factor of each data dimension combination for all target students; Perform learning status impact analysis based on all learning dimension data of all target students in each data dimension combination to obtain the degree of learning impact in each data dimension combination; According to the degree of learning impact under each data dimension combination and the initial overall factor of each data dimension combination for all target students, the overall state representation index of each data dimension combination for all target students is determined, among which the degree of learning impact and the initial overall factor are positively correlated with the overall state representation index.

4. A scenario optimization method based on a teacher-student ARG game social platform according to claim 3, characterized in that: The learning state impact analysis is performed based on all learning dimension data of all target students in each data dimension combination to obtain the learning impact degree in each data dimension combination, including: Determine any target student as a marked student, and determine any data dimension combination as a marked dimension combination; Determine the initial influence degree of the marked student under each preset dimension in the marked dimension combination according to all the learning dimension data of the marked student under the marked dimension combination, wherein the initial influence degree represents the importance of the learning dimension data of the marked student under the preset dimension in the learning dimension data of the marked student in the marked dimension combination; The learning impact degree under each data dimension combination is determined based on all initial impact degrees of all target students under each data dimension combination, where the initial impact degree and the learning impact degree are positively correlated.

5. A scenario optimization method based on a teacher-student ARG game social platform according to claim 4, characterized in that: The step of determining the initial influence degree of the marked student in each preset dimension in the marked dimension combination according to all learning dimension data of the marked student in the marked dimension combination includes: The contribution degree of the principal component of each preset dimension of the marked student in the marked dimension combination is obtained as the initial influence degree of the marked student in each preset dimension in the marked dimension combination.

6. A scenario optimization method based on a teacher-student ARG game social platform according to claim 2, characterized in that: The target dimension combination is screened out from all data dimension combinations based on the overall state characterization index of the data dimension combination and all target clusters under the data dimension combination, including: Filter out label dimensions that represent student performance from all preset dimensions; The variance of the label dimension data corresponding to all target data points in each target cluster is determined as the label variance corresponding to each target cluster, wherein the label dimension data is the learning dimension data under the label dimension; According to the overall state representation index of the label dimension combination for all target students and the label variance corresponding to all target clusters under the label dimension combination, determine the information mining index under the label dimension combination, wherein the overall state representation index is positively correlated with the information mining index, and the label variance is negatively correlated with the information mining index; The data dimension combination with the largest information mining index is selected from all data dimension combinations as the target dimension combination.

7. The scene optimization method based on the teacher-student ARG game social platform according to claim 1 is characterized in that: The optimizing of the teacher-student ARG game social platform scenario according to the target dimension combination includes: The pre-acquired learning dimension data of each target student in the current time period under all preset dimensions in the target dimension combination are used to form a current data point corresponding to each target student; According to the current data points corresponding to all target students, all target students are clustered to obtain the current clusters; According to the current data points corresponding to all target students in each current cluster, ARG game content is recommended to all target students in each current cluster through the teacher-student ARG game social platform, thereby optimizing the teacher-student ARG game social platform scenario.

Images