Cooperative learning grouping method and system based on multi-objective optimization algorithm

Abstract

The invention provides a collaborative learning grouping method and system based on a multi-objective optimization algorithm, and the method comprises the steps: 1, collecting student data which comprises the gender, the number of owned skills, the skill mastering degree, the learning style and the score of a pre-repair course; 2, randomly grouping class members to obtain an initial grouping scheme under the condition of ensuring that the number of students with low score of the first course in each group is controllable and the group leader is fixed; 3, calculating a diversity value Dg of a single group, a diversity value Dc of the whole class, a pressure value # imgabs0 # among members in the group and a pressure value Pc of the whole class; 4, gradually improving the grouping quality through a multi-target genetic algorithm NSGA-II; and 5, taking the grouping scheme with the minimum congestion distance in the Pareto leading edge as a final output scheme. According to the method, grouping can be carried out by considering the student attributes in multiple aspects, so that grouping is more scientific and accurate, the advantages of each student are fully played, and good conditions are created for effective collaborative learning.

Description

Technical Field

[0001] The present application relates to the technical field of multi-objective optimization algorithms, and in particular to a collaborative learning grouping method and system based on a multi-objective optimization algorithm. Background Art

[0002] In the field of education and teaching, group cooperative learning is an extremely important teaching method. At present, there are many intelligent grouping software to realize learning grouping, but the main grouping methods are mostly teacher-defined grouping and random grouping. Teacher-defined grouping relies on the teacher's subjective judgment and teaching experience, and usually groups students based on a single dimension such as academic performance and classroom performance. It has obvious limitations and is difficult to fully take into account students' characteristics and needs. Although random grouping is easy to operate, it is too arbitrary and divorced from teaching reality. It cannot meet the grouping requirements of modern education for efficient and high-quality group cooperative learning.

[0003] The few intelligent grouping software based on optimization algorithms mostly only consider a single target value during optimization, such as only performing intelligent grouping based on predecessor performance levels. This does not fully consider the multi-dimensional differences among students and the scientificity and rationality of the grouping. It is also difficult to meet the actual needs of modern education in terms of group cooperative learning grouping.

[0004] For example, the method disclosed in the paper titled "An Evolutionary Algorithm for Forming Mixed Groups in a Network Collaborative Learning Environment" aims to build learning groups of equal size in a way that maximizes the quality of the entire solution and each group. This method has been implemented on the MOODLE platform. The details are as follows:

[0005] like Figure 1 As shown in the figure, the technical solution first obtains student personal data from the MOODLE platform (Moodle is an open source online learning management system with a powerful plug-in system and rich APIs. Teachers, developers, etc. can customize functions as needed to flexibly adapt to various teaching scenarios and personalized teaching needs.) and uses questionnaires (such as MBTI and Felder Silverman questionnaires) and students' performance in course assignments to collect information on students' learning styles, course grades, etc.

[0006] Subsequently, the teacher classifies the data collected about the students into different attributes and rates the attributes according to his / her own judgment, determining their importance and priority in the grouping process. Next, the technical solution designs the grouping solution quality factor as the fitness function based on the defined attributes, and then executes the genetic algorithm (GA) to optimize the fitness value. The fitness function is a value used to measure the quality of each grouping solution (individual), which can guide the algorithm to search for the optimal solution. Genetic algorithm is an optimization algorithm that simulates the natural evolution process. It screens individuals in the population through selection, crossover and mutation operations to find the optimal solution to the problem. It is often used for complex function optimization, combinatorial optimization and other problems. In the GA operation, the fitness threshold and the maximum number of generations are set as the stopping criteria. If the maximum fitness of the solution found reaches the fitness threshold, or the number of generations exceeds the maximum number of generations, the GA stops running.

[0007] After the algorithm is executed, the results will be presented to the teacher in the form of group graphs and relevant statistical data. The teacher will decide whether to accept the grouping scheme based on the presented results to determine whether to continue the iteration.

[0008] After the collaborative task, students will evaluate each other, evaluate their peers based on participation and satisfaction, and weight the task, and finally evaluate their peers and the group comprehensively. At the same time, based on the results of students' mutual evaluation and students' actual participation (reflected by the number and quality of posts, and teachers determine the quality of posts), the final score is calculated by comprehensively considering the mutual evaluation score, the degree of proximity to the median, and the actual participation, and the result can be modified according to professional opinions.

[0009] However, the above techniques only aim to maximize the quality of the entire solution and each group, without considering the simultaneous optimization of multiple objectives. In actual teaching scenarios, different teaching tasks and student groups may require the simultaneous optimization of multiple attributes. A single optimization goal is difficult to meet diverse teaching needs, resulting in insufficient flexibility of the grouping scheme when facing complex teaching situations. For example, some courses may require a higher group diversity to promote creativity, while ensuring that the stress level within the group is within a low range to ensure group work efficiency.

[0010] In addition, the above technology is based on traditional genetic algorithms, which have shortcomings when dealing with complex data relationships and large-scale data. It is prone to problems such as slow convergence speed, especially when faced with many student attributes and complex grouping requirements, it is difficult to quickly and effectively find a high-quality grouping solution. This makes the grouping process time-consuming and the results may not reach the ideal state, and it cannot efficiently adapt to the large-scale and diverse grouping tasks in modern education. Summary of the invention

[0011] In view of this, the present application provides a collaborative learning grouping method and system based on a multi-objective optimization algorithm to overcome the technical defect that a single optimization objective is difficult to meet diverse teaching needs.

[0012] To achieve the above objectives, the technical solutions adopted in this application are as follows:

[0013] The collaborative learning grouping method based on multi-objective optimization algorithm includes:

[0014] Step 1: Collect student data, including gender, number of skills, skill mastery, learning style, and prerequisite course grades;

[0015] Step 2: Under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, class members are randomly grouped to obtain an initial grouping plan, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than the set threshold;

[0016] Step 3: Calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ;

[0017] Step 4: Gradually improve the grouping quality through the multi-objective genetic algorithm NSGA-II;

[0018] Step 5: Take the grouping scheme with the smallest crowding distance in the Pareto front as the final output scheme.

[0019] Furthermore, in step 3, the diversity value D of a single group is calculated. g and the overall class diversity value D c The specific formula is:

[0020]

[0021] Where G represents the number of group members; t and s represent the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with one or two skills; the weight parameter w t 、w s and w l They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class; σ(D g ) represents the standard deviation of all group diversity values; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

[0022] Furthermore, in step 3, the pressure value between group members is calculated. And the pressure value P of the whole class c The specific formula is:

[0023]

[0024] Among them, the weight w k 、w s and w g Corresponding to historical collaboration records, relationship status and gender differences respectively, if one of the conditions in formula (3) is met, the pressure value between two members is Add the corresponding weights in g represents the total pressure value within a single group, σ(p g ) represents the standard deviation of the pressure values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

[0025] Furthermore, in step 4, the grouping quality is gradually improved by using an improved multi-objective genetic algorithm NSGA-II, which specifically includes:

[0026] Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different levels according to the Pareto dominance principle to obtain the Pareto frontier;

[0027] Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so as to subsequently select a solution with a large crowding distance to maintain diversity. The target space refers to all target values ​​in the dominance level of the current grouping scheme.

[0028] The specific formula for calculating crowding distance is as follows;

[0029]

[0030] in, and Respectively represent target values ​​of two grouping schemes adjacent to the current grouping scheme, wherein the target value is a diversity value or a pressure value; and They represent the maximum and minimum target values ​​at the dominance level of the current grouping scheme, respectively, and i represents the current grouping scheme;

[0031] Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group.

[0032] Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from a parent grouping scheme, the high-quota group is the first several groups in the parent grouping scheme after sorting from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any member of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring;

[0033] Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

[0034] A collaborative learning grouping system based on a multi-objective optimization algorithm, including:

[0035] A data collection module, used to collect student data, wherein the student data includes gender, number of skills possessed, degree of skill mastery, learning style, and grades of prerequisite courses;

[0036] An initialization module is used to randomly group class members to obtain an initial grouping plan under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than a set threshold;

[0037] Calculation module, used to calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ;

[0038] The optimization module is used to gradually improve the grouping quality through the multi-objective genetic algorithm NSGA-II;

[0039] The output module is used to take the grouping scheme with the smallest crowding distance in the Pareto frontier as the final output scheme.

[0040] Furthermore, the calculation module calculates the diversity value D of a single group according to the following formula: g and the overall class diversity value D c :

[0041]

[0042] Where G represents the number of group members; t and s represent the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with one or two skills; the weight parameter w t 、w s and w l They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class, σ(D g ) represents the standard deviation of all group diversity values; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

[0043] Furthermore, the calculation module calculates the pressure value between group members according to the following formula: And the pressure value P of the whole class c :

[0044]

[0045]

[0046] Among them, the weight w k 、w s and w g Corresponding to historical collaboration records, relationship status and gender differences respectively, if one of the conditions in formula (3) is met, the pressure value between two members is Add the corresponding weights in g represents the total pressure value within a single group, m represents the number of groups in a single class, σ(p g ) represents the standard deviation of the pressure values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

[0047] Furthermore, the optimization module is specifically used to perform the following steps:

[0048] Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different levels according to the Pareto dominance principle to obtain the Pareto frontier;

[0049] Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so as to subsequently select a solution with a large crowding distance to maintain diversity. The target space refers to all target values ​​in the dominance level of the current grouping scheme.

[0050] The specific formula for calculating crowding distance is as follows;

[0051]

[0052] in, and Respectively represent target values ​​of two grouping schemes adjacent to the current grouping scheme, wherein the target value is a diversity value or a pressure value; and They represent the maximum and minimum target values ​​at the dominance level of the current grouping scheme, respectively, and i represents the current grouping scheme;

[0053] Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group.

[0054] Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from a parent grouping scheme, the high-quota group is the first several groups in the parent grouping scheme after sorting from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any member of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring;

[0055] Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

[0056] Compared with the prior art, the beneficial effects of this application are:

[0057] 1. This application can consider various student attributes to group students, making the grouping more scientific and accurate, giving full play to the advantages of each student, and creating good conditions for effective collaborative learning;

[0058] 2. When optimizing grouping, this application can ensure that the intra-group stress value is as small as possible while ensuring that the intra-group heterogeneity is high. Higher heterogeneity can promote complementary advantages and knowledge sharing among members, which is conducive to stimulating innovative thinking and improving collaborative learning effects. Lower stress values ​​can enable students to conduct collaborative learning in a more relaxed and harmonious atmosphere, improving learning efficiency and participation. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 Flowchart of evolutionary algorithm for forming mixed groups based on network collaborative learning environment in the prior art;

[0061] Figure 2 This is a flow chart of the collaborative learning grouping method based on the multi-objective optimization algorithm of this application;

[0062] Figure 3 This is a structural block diagram of the collaborative learning grouping system based on the multi-objective optimization algorithm of this application. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0064] When constructing learning groups, existing technologies often only perform grouping operations based on a small number of conditions and standards. This simplistic approach cannot fully adapt to the complex and changing needs of actual teaching. For example, in actual classrooms, students' individual differences are reflected in many aspects, including learning ability, learning style, personality characteristics, and knowledge background, but existing technologies are difficult to fully consider these factors, resulting in the grouping results may not be able to give full play to the advantages of each student, affecting the effectiveness of group collaboration.

[0065] When optimizing grouping, existing technologies mostly focus on a single goal (such as overall solution and group quality), and are unable to simultaneously optimize multiple key goals such as diversity value and stress value. This results in poor flexibility in grouping solutions and is unable to meet the diverse requirements of different courses and learning tasks for group diversity and stress tolerance.

[0066] This application proposes a set of innovative technical solutions to address the defects of the above-mentioned existing technologies. By introducing advanced algorithm models and optimization strategies, it is possible to analyze student attribute data more comprehensively and deeply, fully consider the simultaneous optimization of multiple key goals, and achieve more scientific and accurate heterogeneous grouping.

[0067] like Figure 2 As shown, the present application provides a collaborative learning grouping method based on a multi-objective optimization algorithm, including:

[0068] Step 1: Collect student data, including gender, number of skills, skill mastery, learning style, and prerequisite course grades;

[0069] We collected rich data from students through questionnaires, Kolb Learning Style Inventory, and academic records. We extracted information such as gender, whether they were recommended group leaders, the number of skills they possessed, the degree of their skills mastered, their learning styles, and their grades in prerequisite courses from these data, providing a comprehensive data basis for subsequent group analysis.

[0070] Step 2: Under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, class members are randomly grouped to obtain an initial grouping plan, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than the set threshold;

[0071] For example, the group leader is selected through the collected data, and the group leader is fixed and the number of students with lower scores in the prerequisite courses in each group is less than or equal to one. Then, other members are randomly assigned to generate the initial group to obtain the initial grouping scheme. The set threshold can be 60 points (out of 100 points) or 72 points (out of 120 points); the set threshold can also be a grade value, such as C grade (A grade is the highest grade).

[0072] Step 3: Calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ;

[0073] The heterogeneity within the group is measured by designing the diversity value of member attributes. The calculation takes into account the number of different skills, the number of learning styles, and the differences in skill mastery within the group. The weight parameter is used to evaluate the impact of each factor on the overall diversity, and the global diversity value of the entire class is further calculated. The standard deviation and weight coefficient are introduced as penalty factors in the calculation to prevent excessive concentration or excessive divergence of heterogeneity. The specific calculation formula is as follows:

[0074]

[0075]

[0076] Where G represents the number of group members; t and s represent the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with one or two skills; the weight parameter w t 、w s and w l They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class, σ(D g ) represents the standard deviation of all group diversity values; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

[0077] At the same time, this application designs a stress value to measure the stress within the group, taking into account student relationships, cooperation history and gender factors, and introduces constraints, namely, each group contains at most one member with a low ability score to prevent excessive concentration of stress. Specifically, the low ability score member can be a member with a D grade in the prerequisite course. After calculating the stress value of the group through the weight parameter, the overall group stress value of the class is calculated, and the standard deviation and weight coefficient are also introduced as penalty items to ensure that the stress level of members is reasonable. The specific calculation formula is as follows:

[0078]

[0079] Among them, formula (3) first calculates the pressure value between group members s i and j Respectively represent two different members in the group. Weight w k 、w s and w g They correspond to historical collaboration records, relationship status, and gender differences, respectively. If one of the conditions in formula (3) is met, then the pressure value between the two members is On this basis, the pressure values ​​between each group are added and normalized to obtain the total pressure value p within the group.g Finally, the overall pressure value pc of the entire class is calculated, see formula (4). Where m represents the number of groups in a single class, σ(p g ) represents the standard deviation of the pressure values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

[0080] Step 4: Gradually improve the grouping quality through the multi-objective genetic algorithm NSGA-II;

[0081] NSGA-II (Non-dominated Sorting Genetic Algorithm II) is an efficient multi-objective optimization genetic algorithm that can optimize multiple objectives at the same time. The algorithm uses non-dominated sorting to classify individuals, seek balance among multiple objectives, avoid premature convergence, and improve flexibility and comprehensiveness.

[0082] The step 4 specifically includes:

[0083] Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different dominance levels according to the Pareto dominance principle to obtain the Pareto frontier;

[0084] Pareto dominance principle: If a grouping scheme is better than another grouping scheme in at least one objective and not inferior to another grouping scheme in other objectives, it is said to dominate (be better than) the other grouping scheme.

[0085] Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so that solutions with large crowding distance can be selected later to maintain diversity;

[0086] The crowding distance calculation is based on the comparison of the difference between the target values ​​of adjacent grouping schemes (i.e., the diversity value or pressure value of the entire class) and the maximum and minimum values ​​of the target space (i.e., all target values ​​in the dominance level of the current grouping scheme). The specific calculation formula is as follows;

[0087]

[0088] in, and They represent the target values ​​of the two grouping schemes adjacent to the current grouping scheme. and They represent the maximum and minimum target values ​​at this level respectively, and i represents the current grouping scheme.

[0089] Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group.

[0090] Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from one parent grouping scheme, the high-quota group is the first several groups after sorting the parent grouping scheme from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any members of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring;

[0091] PMX (Partially Mapped Crossover) is an important crossover operation method in genetic algorithms. This method can enhance the diversity and adaptability of offspring while retaining the excellent characteristics of the parent generation.

[0092] Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

[0093] By designing a scientific and reasonable member attribute diversity index and stress index to quantify the heterogeneity and stress within the group, considering a variety of relevant factors and introducing appropriate penalty factors and constraints, we can ensure that the stress level is effectively controlled while achieving diversity among group members, thereby promoting the improvement of collaborative learning effects.

[0094] Step 5: The grouping scheme that performs best in maximizing diversity and minimizing pressure in the Pareto frontier is taken as the final output scheme.

[0095] When performing optimization iterations, two iteration stop conditions can be set to judge convergence, including the stability of each generation of the optimal grouping scheme (i.e., the number of solution changes in the Pareto front solution set) and the similarity of the mean of the objective function value (i.e., the difference between the mean of the two objective values ​​of the Pareto front solution set in consecutive iterations). After the iteration stops, the grouping scheme with the smallest crowding distance (i.e., the solution with the most average performance on the two objective values) is selected from the Pareto front as the final output scheme.

[0096] This application uses an advanced multi-objective optimization algorithm (NSGA-II algorithm) to optimize the grouping scheme, which involves key operations such as non-dominated sorting, crowding distance calculation, tournament selection, improved crossover and mutation. Through these operations, a balance is sought between multiple optimization objectives (such as maximizing diversity and minimizing pressure), and high-quality grouping schemes are found quickly and effectively, thereby improving the adaptability and efficiency of the algorithm in complex data environments.

[0097] like Figure 3 As shown, the present application also provides a collaborative learning grouping system based on a multi-objective optimization algorithm, including:

[0098] A data collection module 310 is used to collect student data, wherein the student data includes gender, number of skills possessed, degree of skill mastery, learning style, and grades of prerequisite courses;

[0099] Initialization module 320, randomly grouping class members to obtain an initial grouping scheme under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than a set threshold;

[0100] Calculation module 330, used to calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ;

[0101] An optimization module 340 is used to gradually improve the grouping quality through a multi-objective genetic algorithm NSGA-II;

[0102] The output module 350 is used to take the grouping scheme with the smallest crowding distance in the Pareto front as the final output scheme.

[0103] Furthermore, the calculation module calculates the diversity value D of a single group according to the following formula: g and the overall class diversity value D c :

[0104]

[0105] Where G is the number of group members; t and s are the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with zero or one skill; the weight parameter w t 、w s and w i They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class, σ(Dg) represents the standard deviation of the diversity values ​​of all groups; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

[0106] Furthermore, the calculation module calculates the pressure value between group members according to the following formula: And the pressure value P of the whole class c :

[0107]

[0108] Among them, the weight w k 、w s and w g Corresponding to historical collaboration records, relationship status and gender differences respectively, if one of the conditions in formula (3) is met, the pressure value between two members is Add the corresponding weights in g represents the total stress value within a single group; m represents the number of groups in a single class, σ(Pg) represents the standard deviation of the stress values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

[0109] Furthermore, the optimization module 340 is specifically configured to perform the following steps:

[0110] Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different levels according to the Pareto dominance principle to obtain the Pareto frontier;

[0111] Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so as to subsequently select a solution with a large crowding distance to maintain diversity. The target space refers to all target values ​​in the dominance level of the current grouping scheme.

[0112] The specific formula for calculating crowding distance is as follows;

[0113]

[0114] in, and Respectively represent target values ​​of two grouping schemes adjacent to the current grouping scheme, wherein the target value is a diversity value or a pressure value; and They represent the maximum and minimum target values ​​at the dominance level of the current grouping scheme, respectively, and i represents the current grouping scheme;

[0115] Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group.

[0116] Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from a parent grouping scheme, the high-quota group is the first several groups in the parent grouping scheme after sorting from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any member of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring;

[0117] Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

[0118] As for the device embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units, such as distributed on a server and a client. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0119] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A collaborative learning grouping method based on a multi-objective optimization algorithm, characterized in that: include: Step 1: Collect student data, including gender, number of skills, skill mastery, learning style, and prerequisite course grades; Step 2: Under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, class members are randomly grouped to obtain an initial grouping plan, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than the set threshold; Step 3: Calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ; Step 4: Gradually improve the grouping quality through the multi-objective genetic algorithm NSGA-II; Step 5: The grouping scheme with the smallest crowding distance in the Pareto frontier is taken as the final output scheme.

2. The collaborative learning grouping method based on a multi-objective optimization algorithm as claimed in claim 1, characterized in that: In step 3, the diversity value D of a single group is calculated. g and the overall class diversity value D c The specific formula is: Where G represents the number of group members; t and s represent the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with one or two skills; the weight parameter w t 、w s and w l They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class; σ(D g ) represents the standard deviation of all group diversity values; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

3. The collaborative learning grouping method based on a multi-objective optimization algorithm as claimed in claim 2, characterized in that: In step 3, the pressure value between group members is calculated And the pressure value P of the whole class c The specific formula is: Among them, the weight w k 、w s and w g Corresponding to historical collaboration records, relationship status and gender differences respectively, if one of the conditions in formula (3) is met, the pressure value between the two members is Add the corresponding weights in; m represents the number of groups in a single class; p g represents the total pressure value within a single group, σ(p g ) represents the standard deviation of the pressure values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

4. The collaborative learning grouping method based on a multi-objective optimization algorithm as claimed in claim 3, characterized in that: In step 4, the multi-objective genetic algorithm NSGA-II is used to gradually improve the grouping quality, which specifically includes: Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different dominance levels according to the Pareto dominance principle to obtain the Pareto frontier; Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so as to subsequently select a solution with a large crowding distance to maintain diversity. The target space refers to all target values ​​in the dominance level of the current grouping scheme. The specific formula for calculating crowding distance is as follows; in, and Respectively represent target values ​​of two grouping schemes adjacent to the current grouping scheme, where the target value is the diversity value of the entire class or the pressure value of the entire class; and They represent the maximum and minimum target values ​​at the dominance level of the current grouping scheme, respectively, and i represents the current grouping scheme; Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group. Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from a parent grouping scheme, the high-quota group is the first several groups in the parent grouping scheme after sorting from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any member of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring; Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

5. A collaborative learning grouping system based on a multi-objective optimization algorithm, characterized in that: include: A data collection module, used to collect student data, wherein the student data includes gender, number of skills possessed, degree of skill mastery, learning style, and grades of prerequisite courses; An initialization module is used to randomly group class members to obtain an initial grouping plan under the condition that the number of students with low scores in prerequisite courses in each group is controllable and the group leader is fixed, wherein the low scores in prerequisite courses refer to the scores in prerequisite courses being lower than a set threshold; Calculation module, used to calculate the diversity value D of a single group g , the diversity value of the entire class D c , Pressure value between group members And the pressure value P of the whole class c ; The optimization module is used to gradually improve the grouping quality through the multi-objective genetic algorithm NSGA-II; The output module is used to take the grouping scheme with the smallest crowding distance in the Pareto frontier as the final output scheme.

6. The collaborative learning grouping system based on a multi-objective optimization algorithm as claimed in claim 5, characterized in that The calculation can be performed from two aspects. The calculation module calculates the diversity value D of a single group according to the following formula: g and the overall class diversity value D c : Where G represents the number of group members; t and s represent the number of different skills and learning styles in the group, respectively; m1 corresponds to the number of members with three or more skills, and m2 corresponds to the number of members with one or two skills; the weight parameter w t 、w s and w l They are used to evaluate the impact of skill diversity, learning style diversity and skill mastery on the overall diversity value; m represents the number of groups in a single class; σ(D g ) represents the standard deviation of all group diversity values; weight coefficient It is a penalty factor to prevent heterogeneity from being too concentrated or too dispersed in certain groups.

7. The collaborative learning grouping system based on a multi-objective optimization algorithm as claimed in claim 6, characterized in that: The calculation module calculates the pressure value between group members according to the following formula And the pressure value P of the whole class c : Among them, the weight w k 、w s and w g Corresponding to historical collaboration records, relationship status and gender differences respectively, if one of the conditions in formula (3) is met, the pressure value between two members is Add the corresponding weights in g represents the total pressure value within a single group, σ(p g ) represents the standard deviation of the pressure values ​​of all groups, It is the weight coefficient of the pressure penalty term, which prevents the pressure from being too concentrated or too dispersed in some groups.

8. The collaborative learning grouping system based on a multi-objective optimization algorithm as claimed in claim 7, characterized in that: The optimization module is specifically used to perform the following steps: Step 4.1: Non-dominated sorting: divide all initial grouping schemes into different levels according to the Pareto dominance principle to obtain the Pareto frontier; Step 4.2: Calculate the crowding distance to evaluate the distribution density of the grouping scheme in the target space so as to subsequently select a solution with a large crowding distance to maintain diversity. The target space refers to all target values ​​in the dominance level of the current grouping scheme. The specific formula for calculating crowding distance is as follows; in, and Respectively represent target values ​​of two grouping schemes adjacent to the current grouping scheme, wherein the target value is a diversity value or a pressure value; and They represent the maximum and minimum target values ​​at the dominance level of the current grouping scheme, respectively, and i represents the current grouping scheme; Step 4.3: Championship selection: select several grouping schemes from the initial grouping scheme set to compete, compare the properties of the selected grouping schemes, including dominance level and crowding distance, give priority to comparing dominance level, and select the grouping scheme with the lowest dominance level. If the dominance levels are the same, the grouping scheme with the largest crowding distance is given priority to be selected as the parent grouping scheme, and so on. Then, select another parent grouping scheme from the remaining grouping schemes in the same way. All the selected parent grouping schemes constitute a new parent grouping scheme group. Step 4.4: Crossover, based on partial mapping crossover improvement, first select a high-quota group from one parent grouping scheme, the high-quota group is the first several groups after sorting the parent grouping scheme from high to low according to the difference between the diversity value of a single group and the total pressure value within the group, then extract an additional reserved group from another parent grouping scheme, the additional reserved group is a randomly selected group that does not contain any members of the high-quota group, then retain the high-quota group and the additional reserved group while maintaining the group leader unchanged, so as to retain the elite and combine the characteristics of the parent generation; finally, under the condition that the number of students with a D grade in the prerequisite courses in each group is less than or equal to 1, fill the remaining students other than the high-quota group and the additional reserved group members into each group to ensure that the group is complete and balanced, retain the excellent characteristics of the parent generation and enhance the diversity and adaptability of the offspring; Step 4.5: Mutation: Randomly swap two students with the same scores in the prerequisite courses in the two groups to add more possibilities to the algorithm while ensuring the balance of the groups.

Images