Course assessment and evaluation feedback system and method

Abstract

The invention discloses a course examination evaluation feedback system and method, and the method comprises the steps: collecting first modal evaluation data and second modal evaluation data of offline classroom data and online classroom data of students, carrying out the analysis and evaluation of item factors, calculating an initial total learning scoring function, and obtaining the scores of the students; meanwhile, target fragile score item factors in the initial total learning score function are searched for reflecting unstable factors of the initial total learning score function; and an alignment adjustment coefficient is set for optimization solution, and the initial total learning scoring function is updated in dual ways with target fragile scoring item factors, so that the score influence on student scores is reduced to the maximum extent, and a final target total learning scoring function is obtained.

Description

Technical Field

[0001] The present invention relates to the field of database processing, and in particular to a course assessment and evaluation feedback method and system. Background Art

[0002] With the rapid development of information technology, the degree of informatization in the education sector has gradually increased. This is particularly true for student information management and performance assessment, with more and more schools beginning to use intelligent student management systems. Traditional performance assessment methods typically rely solely on final exam scores, but this single approach no longer fully reflects a student's comprehensive learning ability. With the prevalence of online education and multimedia instruction, student performance and engagement in both traditional and virtual classrooms are becoming increasingly important. Consequently, the need to use comprehensive data to assess student performance is growing, with both online and offline learning data becoming increasingly important evaluation indicators.

[0003] Students' performance includes offline classroom activities and online learning activities. The data on offline classroom activities are collected through video data sources (i.e., collected during multimedia teaching), while online learning activities are collected through logs or other text data sources. Offline classroom activities mainly record students' participation in traditional classrooms, such as the number of times they raise their hands, the number of times they answer questions, the frequency of participating in discussions, the number of times they ask questions, etc. in video data sources. Online learning activities collect students' activity data in virtual classrooms through online learning platforms, such as online learning time, video viewing, online homework submission, discussion participation, etc. These data can more comprehensively reflect students' learning situations in different learning environments. In the teaching process, the study found that only by combining the two with students' basic information data for comprehensive evaluation can the students' total learning score function (or total learning score) be more accurately derived.

[0004] However, further research found that offline classroom activities and online learning activities are essentially two different modal data sources. Due to the problem of inconsistent dimensional directions and inconsistent data distribution between the offline classroom activity evaluation and the online learning activity evaluation (for example: inconsistent dimensional directions lead to conflicts in the evaluation functions after fusion; for example: especially in the total learning score function formed by the fusion of two cross-modal evaluation functions, there is a contradiction (i.e., negative correlation) between the two scoring item factors, resulting in inconsistent dimensional directions), the above problems can easily lead to inaccurate comprehensive evaluation scores output by the fused total learning score function, resulting in evaluation bias. Summary of the Invention

[0005] The purpose of the present invention is to provide a course assessment and evaluation feedback system and method, which solves the above-mentioned technical problems pointed out in the prior art.

[0006] The present invention provides a course assessment and evaluation feedback system, comprising an encoding module, a screening module, a feedback update processing module and an operation output processing module;

[0007] The encoding module is used to collect and obtain basic information data of students and use the basic information data to generate an initial QR code identification;

[0008] The screening module is configured to collect first modality evaluation data and second modality evaluation data for the current student; and determine an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data;

[0009] The feedback update processing module is used to set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; introduce the alignment adjustment coefficient to correct the target vulnerability scoring factor in the initial total learning score function, and update the target total learning score function after correction;

[0010] The operation output processing module is used to evaluate the student's score based on the target total learning score function by performing the first modality evaluation data, the second modality evaluation data and the basic information data, output the current student's total evaluation score, and generate a target QR code identifier in combination with the QR code identifier.

[0011] Accordingly, the present invention provides a course assessment and evaluation feedback method, comprising the following steps:

[0012] Collect and obtain basic information data of students, and use the basic information data to generate an initial QR code identification;

[0013] Collecting first modality evaluation data and second modality evaluation data for the current student; determining an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data;

[0014] The initial total learning score function = S = f(x1, x2, x3, x4, x5, x6, ...);

[0015] Among them, xi is the i-th scoring factor, f is the scoring function;

[0016] Calculate the conflict degree of the pairwise scoring item factors in the initial total learning scoring function;

[0017] First, calculate the average conflict degree of each rating item factor with other rating item factors:

[0018]

[0019] Where Ci is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function, and N is the total number of scoring factors;

[0020] Set a threshold tau. If Ci>tau, the scoring factor xi is considered to be a scoring factor of target vulnerability.

[0021] Set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; introduce the alignment adjustment coefficient to correct the target vulnerability score factor in the initial total learning score function, and update the target total learning score function after correction;

[0022] Based on the target total learning score function, the student's score is evaluated by the first modality evaluation data, the second modality evaluation data and the basic information data, the current student's total evaluation score is output, and the target QR code identifier is generated in combination with the QR code identifier.

[0023] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0024] Analysis of the above-mentioned course assessment and evaluation feedback system and method provided by the present invention shows that in specific applications, the alignment adjustment coefficient is set to optimize and solve to update the initial total learning score function, and the initial particle group is generated according to the alignment adjustment coefficient. The average fitness value of the particles is iteratively calculated as the actual fitness value of the current iteration number to avoid falling into the local optimal solution, and the adaptive shrinkage factor is used to adjust the number of current iterations. The exponential optimization of the speed and position of the alignment adjustment coefficient C is close to the optimal solution, which improves the accuracy of finding the optimal alignment adjustment coefficient C. * possibility, thereby obtaining the optimal alignment adjustment coefficient C * ;

[0025] Further through the optimal alignment adjustment coefficient C * The corresponding alignment adjustment coefficient of the target fragile scoring factor is dynamically adjusted to correct the unstable influence of the target fragile scoring factor on the initial total learning scoring function, thereby obtaining the corrected alignment adjustment coefficient of the target fragile scoring factor.

[0026] Furthermore, by substituting the particles in the initial particle swarm into the initial total learning score function S, we can observe whether these adjustment coefficients have effectively corrected the problems in the initial total learning score function. The modified initial total learning score function is modified twice to obtain the final target total learning score function, which minimizes the impact on student scores. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic diagram of the main architecture principle of a course assessment and evaluation feedback system in Example 1;

[0028] Figure 2 This is an overall flow chart of a course assessment and evaluation feedback method in Example 2;

[0029] Figure 3 This is a flow chart of obtaining an initial total learning score function for a course assessment and evaluation feedback method in Example 2;

[0030] Figure 4 This is a schematic diagram of scoring item factors for a course assessment and evaluation feedback method in Example 2;

[0031] Figure 5 This is a flow chart of a function for generating a target total learning score for a course assessment and evaluation feedback method in Example 2;

[0032] Figure 6 This is a flow chart of an optimized alignment adjustment coefficient for a course assessment and evaluation feedback method in Example 2;

[0033] Label: encoding module 10; screening module 20; feedback update processing module 30; operation output processing module 40; first modal processing module 21; second modal processing module 22; calculation initial module 23; particle swarm module 31; constraint module 32; correction module 33; target module 34; distribution random factor module 321; shrinkage random factor module 322; particle velocity calculation module 323; randomness adjustment factor module 324; time series module 325; optimization module 326. DETAILED DESCRIPTION

[0034] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] The present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0036] Example 1

[0037] like Figure 1 As shown, the present invention provides a course assessment and evaluation feedback system, comprising: an encoding module 10, a screening module 20, a feedback update processing module 30 and an operation output processing module 40;

[0038] The encoding module 10 is used to collect and obtain basic information data of students and use the basic information data to generate an initial QR code identification;

[0039] The screening module 20 is configured to collect first modality evaluation data and second modality evaluation data for the current student; and determine an initial overall learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data;

[0040] The initial total learning score function = S = f(x1, x2, x3, x4, x5, x6, ...);

[0041] Among them, xi is the i-th scoring factor, f is the scoring function;

[0042] Calculate the conflict degree of the pairwise scoring item factors in the initial total learning scoring function;

[0043] First, calculate the average conflict degree of each rating item factor with other rating item factors:

[0044]

[0045] Where Ci is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function, and N is the total number of scoring factors;

[0046] Set a threshold tau. If Ci>tau, the scoring factor xi is considered to be a scoring factor of target vulnerability.

[0047] The feedback update processing module 30 is used to set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; introduce the alignment adjustment coefficient to correct the target vulnerability scoring factor in the initial total learning score function, and update the target total learning score function after correction;

[0048] The operation output processing module 40 is used to evaluate the student's score based on the target total learning score function by performing the first modality evaluation data, the second modality evaluation data and the basic information data, output the current student's total evaluation score, and generate a target QR code identifier in combination with the QR code identifier.

[0049] Specifically, the screening module 20 includes a first modal processing module 21, a second modal processing module 22 and an initial calculation module 23;

[0050] The first modality processing module 21 is used to video record students' offline classroom behavior and electronically monitor the video data source; monitor offline classroom behavior through the electronic monitoring, and collect motion capture and sound capture of students' offline classroom behavior; analyze the effectiveness of students' offline classroom question answers based on the motion capture and sound capture of offline classroom behavior, and obtain first modality evaluation data of the current student's offline classroom behavior based on the question answer effectiveness;

[0051] The second modal processing module 22 is used to monitor the student's online classroom behavior access to the online learning platform, analyze the student's online learning time, online homework completion, and online review performance based on the current student's access monitoring, and analyze the access records of the access monitoring; obtain online learning data of the current student's online classroom behavior through the access records as the second modal evaluation data;

[0052] The initial module 23 is used to analyze the offline classroom behavior of the first modal evaluation data, score the students' motion capture and sound capture, and obtain a motion normative score and a discussion interaction score; score the effectiveness of the question answers to obtain a question answer effectiveness score; analyze the online classroom behavior of the second modal evaluation data, score the online learning time, online homework completion, and online review score, and obtain an online learning time score, an online homework completion score, and an online review score; and use the scores of the first modal evaluation data and the first modal evaluation data as scoring item factors;

[0053] Calculate the initial total learning score function S=f(x1, x2, x3, x4, x5, x6, ...) by using the first modal evaluation data and the scoring item factors of the first modal evaluation data;

[0054] Specifically, the feedback update processing module 30 includes a particle swarm module 31, a constraint module 32, a correction module 33 and a target module 34;

[0055] The particle swarm module 31 is used to set the alignment adjustment coefficient C=(C1, C2, C3, C4, .., C n ), and using a hybrid mapping method to generate a particle swarm with a number of particles corresponding to the alignment adjustment coefficient, wherein each particle represents a candidate solution for the alignment adjustment coefficient; generating an initial position for the particle swarm to obtain an initial particle swarm;

[0056] The constraint module 32 is used to constrain the position of each particle in the initial particle swarm, iterate each constrained particle, calculate the fitness value of the particle in each iteration, calculate the average of the fitness values ​​calculated for all iterations, obtain the average fitness value as the actual fitness value of the current iteration, and determine the current number of iterations based on the actual fitness value;

[0057] The adaptive shrinkage factor is used to adjust the particle speed and the latest moving position of the particles in the initial particle group at the current iteration number; the alignment adjustment coefficient C is optimized by the particle speed and the latest moving position to obtain the optimal alignment adjustment coefficient C. * ;

[0058] The correction module 33 is used to adjust the optimal alignment coefficient C * Dynamically adjust the alignment adjustment coefficient corresponding to the target vulnerable scoring factor to obtain the modified alignment adjustment coefficient of the target vulnerable scoring factor

[0059] Where C i is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function; β is the correction factor;

[0060] The target module 34 is used to substitute each particle in the initial particle swarm into the initial total learning score function S, update the initial total learning score function S, and obtain a second total learning score function S'=f(C1x1, C2x2, ..., C n x N );

[0061] Where C n Expressed as the scoring factor x N Alignment adjustment factor;

[0062] Alignment adjustment coefficient of the score item factor after correction of target vulnerability Updating the second total learning score function S' to obtain a target total learning score function;

[0063] The expression is:

[0064] Where, It expresses the alignment adjustment coefficient after the optimization of the scoring factor in the target total learning scoring function;

[0065] Specifically, the constraint module 32 includes a distribution random factor module 321, a contraction random factor module 322, a particle velocity calculation module 323, a randomness adjustment factor module 324, a time series module 325, and an optimization module 326;

[0066] The distribution random factor module 321 is used to calculate the distribution variance of the particles in the initial particle group;

[0067] Preset a particle distribution variance threshold o, and determine whether the particles in the initial particle group are greater than or equal to the particle distribution variance threshold o;

[0068] If not, it is determined that the initial positions of the particles in the initial particle group are evenly distributed in the search space;

[0069] If yes, it is determined that the particles in the initial particle swarm have randomness in the search space as a distribution random factor;

[0070] The shrinkage random factor module 322 is used to set a shrinkage factor range threshold j according to the current number of iterations, and the shrinkage factor range threshold j includes a shrinkage factor minimum value j1 and a shrinkage factor maximum value j2;

[0071] Using the maximum value of the contraction factor j2 to perform preliminary iterations of extensive search for particles, the contraction factor is reduced according to the iterative changes of the extensive search for particles to obtain a reduced contraction factor;

[0072] When the reduced shrinkage factor is close to the minimum shrinkage factor j1, the number of iterations t and the maximum number of iterations T of the current reduced shrinkage factor are recorded;

[0073] Calculating a nonlinear adaptive shrinkage factor using the iteration number t of the current reduced shrinkage factor and the maximum iteration number T, and using the adaptive shrinkage factor as a shrinkage random factor;

[0074] The particle velocity calculation module 323 is used to adjust the particle velocity in the search space for each particle iteration using the adaptive shrinkage factor;

[0075] The particle velocity is calculated by the inertia term, individual cognition term, and social cognition term in the search space. The calculation formula is:

[0076] v(t+1)=χ(t)[w×v(t)+c1r1(pbest-x(t))+c2r2(gbest-x(t))];

[0077] Where w×v(t) represents the inertia term, w represents the inertia factor (that is, the degree to which the inertia weight w controls the particle speed and affects the particle's search range), and v(t) represents the current speed of the particle;

[0078] χ(t) represents the current iteration position of the particle (or the position of the particle at the current iteration);

[0079] c1r1(pbest-x(t)) is the individual cognitive term, and pbest is the position found by the particle in the previous iteration;

[0080] c2r2(gbest-x(t)) is represented as the social cognition term, gbest is represented as the position found by the particle in the previous iteration;

[0081] c1 and c2 are represented as learning factors;

[0082] r1 and r2 are interval random numbers;

[0083] The random adjustment factor module 324 is used to calculate the random adjustment factor appearing in the particle velocity using the distribution random factor, the contraction random factor, the inertia factor, and the interval random numbers of r1 and r2;

[0084] The time series module 325 is used to update the particle velocity by using the random adjustment factor to obtain a new particle velocity;

[0085] The position of the particle in the current iteration is updated according to the change amplitude of the new particle velocity to obtain the latest moving position of the particle in the search space;

[0086] Record the time point of the particle's latest moving position in each iterative search in the search space to form a time series;

[0087] The optimization module 326 is used to optimize the alignment adjustment coefficient C of the time series by using the random adjustment factor. * :

[0088]

[0089] Where n represents the number of time intervals; n-1 represents adjacent time points;

[0090] ti represents the latest moving position of the particle at the i-th time point;

[0091] Expressed as the logarithm of the change in time interval;

[0092] λ represents the random adjustment weight;

[0093] Rtotal is expressed as the randomness adjustment factor;

[0094] In the above formula, when When it is equal to 1, the adjacent time intervals in the time series change evenly, and the random adjustment factor has no fluctuation; when When it is equal to 0, the adjacent time intervals in the time series vary greatly, and the random adjustment factor fluctuates, and the optimized alignment adjustment coefficient C is obtained. * ;

[0095] In summary, the course assessment and evaluation feedback system proposed in the present invention collects video recordings of students in offline classes and electronic monitoring of video data sources, analyzes motion capture and sound capture of the electronic monitoring to determine whether the students' discussion interactions or answers are correct, thereby obtaining first modal evaluation data of the current students; monitors students' online classroom behavior through online learning platforms to understand their online learning time, online homework completion, and online review scores, thereby obtaining second modal evaluation data; analyzes scoring item factors of the first modal evaluation data and the first modal evaluation data, and calculates the initial total learning score function of the first modal evaluation data and the first modal evaluation data based on the scoring item factors; performs pairwise conflict analysis on the scoring item factors to find conflicting scoring item factors as target fragile scoring item factors. By identifying the target fragile scoring item factors, it is possible to understand the less stable scoring item factors in the initial total learning score function, thereby improving the stability of the entire total score function; and updates the initial total learning score function by setting an alignment adjustment coefficient for optimization and solution, thereby obtaining a final target total learning score function, reducing the unstable impact on the student's total score, and enabling the student to obtain a final accurate score.

[0096] Example 2

[0097] like Figure 2 As shown, the present invention also provides a course assessment and evaluation feedback method, comprising the following steps:

[0098] S1: Collect and obtain basic information data of students, and use the basic information data to generate an initial QR code identification;

[0099] It should be noted that the basic information data of the school's students is collected in many aspects, including student ID, name, major, class, learning track (i.e., learning track is the number of hours of learning for the major), etc.

[0100] And based on these basic information data, use coding software to edit them into QR codes, and use QR codes to record this information;

[0101] S2: collecting first modality evaluation data and second modality evaluation data for the current student; determining an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data;

[0102] The initial total learning score function = S = f(x1, x2, x3, x4, x5, x6, ...);

[0103] Among them, xi is the i-th scoring factor, f is the scoring function;

[0104] Explanation: Each scoring factor can be derived from either primary or secondary evaluation data. For example, primary data can include the number of questions answered, discussion frequency, and number of questions asked in the video data source; secondary data can include online learning time, online homework completion, and online review score values.

[0105] Calculate the conflict degree of the pairwise scoring item factors in the initial total learning scoring function;

[0106] First, calculate the average conflict degree of each rating item factor with other rating item factors:

[0107]

[0108] Where Ci is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function, and N is the total number of scoring factors;

[0109] Set a threshold tau (e.g., tau = 0.7). If Ci>tau, the scoring factor xi is considered to be a scoring factor of target vulnerability.

[0110] It should be noted that the conflict degree is an indicator that measures the similarity or difference between two scoring factors. Generally, a high conflict degree means that the relationship between the two factors is more complex or may interfere with each other. The conflict degree of each scoring factor xi with all other scoring factors is set, and their average conflict degree is calculated. A threshold tau is introduced to determine the fragile scoring factors. The fragile scoring factors can point out those scoring factors with a high conflict degree. Because their relationship with other scoring factors is more complex or their mutual influence is greater, the final scoring may be unstable.

[0111] S3: Setting the alignment adjustment coefficient to the initial total learning score function, and using the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; introducing the alignment adjustment coefficient to correct the target vulnerability score factor in the initial total learning score function, and updating the target total learning score function after correction (i.e., thereby optimizing the reliability of the total score function);

[0112] S4: Based on the target total learning score function, the student's score is evaluated by the first modality evaluation data, the second modality evaluation data and the basic information data, the total evaluation score of the current student is output, and the target QR code identifier is generated in combination with the QR code identifier.

[0113] It should be noted that the first modality evaluation data refers to: monitoring and recording students' behavior and performance in traditional offline classes (such as the number of questions answered, the frequency of participating in discussions, and the number of questions asked in the video data source); the second modality evaluation data refers to: recording students' activities in virtual classes through online learning platforms (such as online learning time, online homework completion, and online review score values, etc.); analyzing the collected offline and online learning data, comprehensively evaluating students' academic performance based on these data, collecting first modality evaluation data and second modality evaluation data for current students; determining an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data; calculating the conflict degree of each pair of scoring item factors in the initial total learning scoring function to screen out scoring item factors with vulnerable targets;

[0114] Then, the alignment adjustment coefficient is set to the initial total learning score function, and the particle swarm optimization algorithm is used to determine the alignment adjustment coefficient in the initial total learning score function; the alignment adjustment coefficient is introduced to correct the target vulnerable scoring factor in the initial total learning score function, and the target total learning score function is updated after correction (i.e., the reliability of the total score function is optimized). Based on the target total learning score function, the student's score is evaluated by the first modal evaluation data, the second modal evaluation data, and the basic information data, and the total evaluation score of the current student is output;

[0115] In subsequent use, the system will obtain the total evaluation score of the current student and combine it with the basic information data to regenerate a target QR code identifier containing the student's basic information (i.e. student number, name, major, class, learning track); teachers and students can view their own rating information by scanning the above target QR code identifier;

[0116] Specifically, if Figure 3 As shown, in step S2, first modality evaluation data and second modality evaluation data are collected for the current student; based on the first modality evaluation data and the second modality evaluation data, an initial total learning scoring function composed of multiple scoring item factors is determined, including:

[0117] S21: Video recording of students’ offline classroom behavior and electronic monitoring of video data sources;

[0118] Monitor offline classroom behavior through electronic monitoring, and capture motion and sound of students' offline classroom behavior;

[0119] Analyzing the effectiveness of the student's offline classroom question answers based on the motion capture and sound capture of the offline classroom behavior, and obtaining first modality evaluation data of the current student's offline classroom behavior based on the question answer effectiveness;

[0120] It should be noted that electronic monitoring of students and video data sources is carried out through electronic cameras in offline classrooms, and each student's behavior in offline classes is identified through electronic monitoring. That is, the video recording captures whether the student's behavior is normal (i.e., whether the student is turning around and whispering in the name of discussing a problem), and whether the sound is actually discussing the problem (i.e., whether the student is talking about other things while answering questions or discussing problems);

[0121] Then, through motion capture analysis of the student's interaction with other students (i.e., the discussion questions of students in front, behind, and on both sides), combined with the audio of the discussion questions recognized by the video data source, we determine whether the student's discussion interaction or the questions answered are correct, thereby proving the validity of the student's answer (i.e., proving the correctness of the student's answer, thereby reflecting the student's knowledge mastery), and obtaining the first modal evaluation data of the current student's offline classroom behavior;

[0122] S22: monitoring the student's online classroom behavior on the online learning platform, analyzing the student's online learning time, online homework completion, and online review performance based on the current student's access monitoring, and analyzing the access records of the access monitoring;

[0123] The online learning data of the current students’ online classroom behaviors are obtained through access records as the second modality evaluation data;

[0124] It should be noted that the online learning platform monitors students' online learning time, online homework completion, and online review scores, and uses students' access records to understand their mastery of online classroom knowledge, thereby obtaining students' online learning data as the second modality evaluation data;

[0125] S23: Analyze the offline classroom behavior of the first modality evaluation data, score the students' motion capture and sound capture to obtain a motion standardization score and a discussion interaction score; score the effectiveness of the question answers to obtain a question answer effectiveness score;

[0126] Analyzing the online classroom behavior of the second modality evaluation data, scoring the online learning time, online homework completion, and online review score to obtain an online learning time score, an online homework completion score, and an online review score;

[0127] The scores of the first modal evaluation data and the first modal evaluation data (i.e., the scores of the first modal evaluation data and the first modal evaluation data include but are not limited to the above-mentioned action standardization score and online review performance score, etc.) are used as scoring item factors;

[0128] Calculate the initial total learning score function S=f(x1, x2, x3, x4, x5, x6, ...) by using the first modal evaluation data and the scoring item factors of the first modal evaluation data;

[0129] It should be noted that the first modality evaluation data is used to analyze the motion capture of offline classroom behavior to obtain a motion normative score, which calibrates whether the students' offline classroom behavior is normative (i.e., whispering); the video capture of students' discussions is then used to score whether they are the issues discussed in the activity, and then the motion capture and sound capture are combined to analyze whether the questions answered by students are correct, thereby scoring the effectiveness of the answers to the questions (i.e., the scoring of the first modality evaluation data includes but is not limited to the motion normative score, the discussion interaction score, and the question answering effectiveness score, as well as the number, depth, and effectiveness of the discussion questions, and the score of whether the students are attentive to the lecture, etc.);

[0130] The online classroom behavior of the second modality assessment data is analyzed to score whether the students' online learning time has met the time requirements for the learning objectives, score the homework completion rate and accuracy rate, and score the review score according to the online review test results or review completion rate (that is, the scoring of the second modality assessment data includes but is not limited to the online learning time score, the online homework completion score and the online review score score, and also includes the score of the marking of learning knowledge points during online learning, such as Figure 4 shown);

[0131] Specifically, if Figure 5 As shown, in step S3, the alignment adjustment coefficient is set to the initial total learning score function, and the alignment adjustment coefficient in the initial total learning score function is determined by using the particle swarm optimization algorithm; the alignment adjustment coefficient is introduced to correct the target vulnerability score item factor in the initial total learning score function, and the target total learning score function is updated after correction, including:

[0132] S31: Set the alignment adjustment coefficient C=(C1, C2, C3, C4, .., C n ), and using a hybrid mapping method to generate a particle swarm with a number of particles corresponding to the alignment adjustment coefficient, each particle representing a candidate solution for the alignment adjustment coefficient;

[0133] Generate an initial position for the particle swarm to obtain an initial particle swarm;

[0134] Explanation: The execution goal of the above step S31 is to set the alignment adjustment coefficient C and then optimize it to obtain the optimal alignment adjustment coefficient C, thereby updating the total learning score function. First, the particles are initialized and defined: that is, each particle represents a candidate solution for the alignment adjustment coefficient C (that is, the alignment adjustment coefficient is set according to the initial total learning score function, and the initial total learning score function is S = f (x1, x2, x3, x4, x5, x6, ...); therefore, the number of N scoring item factors corresponding to the alignment adjustment coefficient, that is, Cn alignment adjustment coefficients, represents multiple alignment adjustment coefficients C (for example, C = (C1, C2, C3, C4, .., C n )) generates n particles as a particle swarm (that is, the number of particles corresponding to the alignment adjustment coefficient), denoted as C, which can be a multidimensional vector. Using the hybrid mapping method to generate the initial position of the particle swarm can be understood as initializing the initial position (i.e., the initial value of the alignment adjustment coefficient C) and velocity of each particle, usually randomly generated within the set search space.

[0135] S32: constraining the position of each particle in the initial particle swarm, iterating each constrained particle, calculating a fitness value for each particle in each iteration, calculating an average of the fitness values ​​calculated for all iterations, obtaining an average fitness value, and using the average fitness value as the actual fitness value of the current iteration. The current number of iterations is determined based on the actual fitness value (i.e., by calculating the average fitness value of all particles in each iteration, it is possible to understand whether the position of the entire particle swarm in the current iteration is in a relatively optimal solution).

[0136] The adaptive shrinkage factor is used to adjust the particle speed and the latest moving position of the particles in the initial particle group at the current iteration number; the alignment adjustment coefficient C is optimized by the particle speed and the latest moving position to obtain the optimal alignment adjustment coefficient C. * ;

[0137] It should be noted that in the particle swarm optimization algorithm, particles represent candidate solutions to the problem (in this case, the initial value of the alignment adjustment coefficient C); the diversity of the initial particle swarm is crucial to the search process; through the hybrid mapping method, a diverse particle swarm can be generated, that is, the initial particles have a wide distribution, which helps to avoid the algorithm from falling into the local optimal solution; through the wide distribution of initial particles, the particle swarm can start the iterative search from different starting points and calculate the fitness values ​​of all particles in the particle swarm for each iteration, and calculate the average fitness value based on the fitness value of each iteration as the actual fitness value of the current iteration; thereby avoiding falling into the local optimal solution, the diverse initial particle swarm helps to find a suitable alignment adjustment coefficient C on a global scale;

[0138] The alignment adjustment coefficient C of each iteration of the particle is calculated based on the current number of iterations determined by the actual fitness value: the calculation formula is Where λ(t) is the alignment adjustment coefficient C of the tth iteration, λ0 is the initial alignment adjustment coefficient C, and λ max is the final alignment adjustment coefficient C * , T is the total number of iterations, t is the current number of iterations; in the particle swarm optimization process, the position and speed of the particle are updated according to certain factors (such as the personal best position, the global best position); the introduction of the adaptive shrinkage factor is to dynamically adjust the change amplitude of the particle speed and position; ensure that the particle search process can not only cover a wider solution space, but also make fine adjustments when approaching the optimal solution, and improve the accuracy of finding the optimal alignment adjustment coefficient C * possibility;

[0139] Use the adaptive shrinkage factor to adjust the speed and position of the particles, that is, update the speed and position of the particles, forcing the subsequent particles to develop in a better direction; accelerate the particles to find the optimal solution, and ensure that the particles in the group will gradually tend to the optimal solution, thereby optimizing the alignment adjustment coefficient C through the optimal solution * ;

[0140] S33: Adjust the optimal alignment coefficient C * Dynamically adjust the alignment adjustment coefficient corresponding to the target vulnerable scoring factor to obtain the modified alignment adjustment coefficient of the target vulnerable scoring factor

[0141] Where C i is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function;

[0142] β is expressed as the correction factor (i.e., the magnitude of the alignment adjustment coefficient of the scoring item factor that controls the target vulnerability);

[0143] It should be noted that the corresponding alignment adjustment coefficient C of the target fragile scoring factor in the initial total learning scoring function is found k (C k As the corresponding alignment adjustment coefficient of the target vulnerable scoring factor), the optimal alignment adjustment coefficient C * Alignment adjustment coefficient C k After making corrections, a reliable scoring factor of target vulnerability was obtained;

[0144] S34: Substitute each particle in the initial particle swarm into the initial total learning score function S, update the initial total learning score function S, and obtain a second total learning score function S'=f(C1x1, C2x2, ..., C n xN );

[0145] Where C n Expressed as the scoring factor x N Alignment adjustment factor;

[0146] Alignment adjustment coefficient of the score item factor after correction of target vulnerability Updating the second total learning score function S' to obtain a target total learning score function;

[0147] The expression is:

[0148] Where, It expresses the alignment adjustment coefficient after the optimization of the scoring factor in the target total learning scoring function;

[0149] It should be noted that the initial total learning score function S is usually not optimized, and there may be problems with the weights or contributions of some scoring factors being unreasonable. By substituting particles (alignment adjustment coefficients) into S, we can observe whether these adjustment coefficients effectively correct the problems in the initial total learning score function (for example, reducing the impact of fragile factors, improving the overall stability of the score, or reducing the score error), thereby obtaining the second total learning score function S'.

[0150] Then adjust the coefficient by alignment Update the second total learning score function S'. After double updating the initial total learning score function, the final target total learning score function is obtained, which minimizes the impact on the student's score.

[0151] Specifically, if Figure 6 As shown, in step S32, the particle velocity and the latest moving position of the particles in the initial particle group at the current iteration number are adjusted using the adaptive contraction factor. The operation steps are as follows:

[0152] Calculating the distribution variance of particles in the initial particle swarm to determine the randomness of particles in the search space as a distribution random factor;

[0153] Perform a wide search for particles based on the minimum contraction factor j1 and the maximum contraction factor j2 of the preset contraction factor range threshold j, record the number of iterations, calculate a nonlinear adaptive contraction factor, and use the adaptive contraction factor as a contraction random factor;

[0154] The randomness adjustment factor is calculated by combining the distribution random factor and the contraction random factor with the inertia term, the individual cognitive term and the social cognitive term in the search space;

[0155] Update the particle velocity according to the randomness adjustment factor, and combine it with the position update of the current iteration to obtain the latest moving position of the particle in the search space, including:

[0156] S321: Calculating distribution variance of particles in the initial particle swarm;

[0157] Preset a particle distribution variance threshold o, and determine whether the particles in the initial particle group are greater than or equal to the particle distribution variance threshold o;

[0158] If not, it is determined that the initial positions of the particles in the initial particle group are evenly distributed in the search space;

[0159] If yes, it is determined that the particles in the initial particle swarm have randomness in the search space as a distribution random factor;

[0160] It should be noted that the distribution variance calculated by the initial position of the initial particle swarm in the search space can be used to understand the distribution state of the particles in the search space, thereby avoiding the phenomenon of local search between particles and avoiding the emergence of local optimal solutions; therefore, if the particles in the initial particle swarm cannot completely cover the search space, it means that the initial particle swarm is unevenly distributed and there is a strong possibility of random search, so it is used as a random factor for the distribution of particles;

[0161] S322: Setting a shrinkage factor range threshold j (e.g., 0.4, 0.9) according to the current number of iterations, wherein the shrinkage factor range threshold j includes a minimum shrinkage factor j1 (e.g., 0.4) and a maximum shrinkage factor j2 (e.g., 0.9);

[0162] The maximum value of the contraction factor j2 is used to perform a preliminary iteration of a wide search for particles (i.e., if the contraction factor is too small from the beginning (i.e., 0.4 of the minimum value of the contraction factor j1), the particles will search in a small area of ​​the search space, which is easy to fall into a local optimal solution, causing the algorithm to converge prematurely and fail to find the global optimal solution. Therefore, the preliminary particle search uses 0.9 of the maximum value of the contraction factor j2 for a wide search), and the contraction factor is reduced according to the iterative change of the particle's wide search to obtain a reduced contraction factor (i.e., when the iterative wide search is performed with the maximum value of the contraction factor j2 of 0.9, the contraction factor will gradually narrow the search range, thereby causing 0.9 to gradually decrease, thus approaching 0.4 (i.e., the minimum value of the contraction factor j1));

[0163] When the reduced shrinkage factor is close to the minimum value j1 of the shrinkage factor, the number of iterations t and the maximum number of iterations T of the current reduced shrinkage factor are recorded (that is, when the reduced shrinkage factor (i.e., 0.9) gradually approaches 0.4 (i.e., the minimum value j1 of the shrinkage factor), the number of iterations of the current reduced shrinkage factor is recorded at this time, and at the same time, the number of iterations of the next iteration of 0.4, that is, the maximum number of iterations, is also recorded);

[0164] Calculate a nonlinear adaptive shrinkage factor using the number of iterations t of the current reduced shrinkage factor and the maximum number of iterations T, and use the adaptive shrinkage factor as a shrinkage random factor (i.e., the shrinkage factor affects the range of the particle in the search space, so random global or local searches may occur during adaptive dynamic adjustment, resulting in only relatively good solutions and failure to correctly find the optimal solution);

[0165] It should be noted that by setting the shrinkage factor range threshold j (0.4, 0.9) and a larger shrinkage factor (i.e., the maximum shrinkage factor j2 is close to 0.9), the particles in the initial particle swarm can conduct extensive exploration in a larger search space, which helps the particle swarm algorithm avoid falling into a local optimal solution and can more comprehensively discover the potential optimal solution area;

[0166] As the number of iterations increases, the shrinkage factor gradually decreases (i.e., the maximum shrinkage factor j2 of 0.9 decreases with iterations and approaches 0.4, which is close to the minimum shrinkage factor j1). The search range of the particles in the initial particle swarm becomes smaller, so that a fine search can be performed in a smaller area of ​​the search space. This helps the initial particle swarm to perform more precise optimization when finding a potential optimal solution, gradually converge to the global solution, and thus adjust the search range.

[0167] By dynamically adjusting the shrinkage factor at each iteration, the search behavior of particles can be adjusted at different stages of the algorithm. A larger shrinkage factor in the early stage allows particles to grow slowly and explore a wider area, while the shrinkage factor is gradually reduced in the later stage to accelerate the search. While maintaining search efficiency, the appropriate search range is eventually found, and the alignment adjustment coefficient C is found to find the optimal solution.

[0168] S323: adjusting the particle velocity in the search space for each particle iteration using the adaptive contraction factor;

[0169] The particle velocity is calculated by the inertia term, individual cognition term, and social cognition term in the search space. The calculation formula is:

[0170] v(t+1)=χ(t)[w×v(t)+c1r1(pbest-x(t))+c2r2(gbest-x(t))];

[0171] Where w×v(t) represents the inertia term, w represents the inertia factor (that is, the degree to which the inertia weight w controls the particle speed and affects the particle's search range), and v(t) represents the current speed of the particle;

[0172] χ(t) represents the current iteration position of the particle;

[0173] c1r1(pbest-x(t)) is expressed as the individual cognitive term (i.e., the purpose of the individual cognitive term is to guide the particle to move to its historical optimal position (individual optimal solution), encourage the particle to return to the best solution position it has found, and thus promote the particle to search in a good solution space), pbest is expressed as the position found by the particle in the previous iteration (i.e., the best position before the current position of x(t));

[0174] c2r2(gbest-x(t)) is expressed as a social cognitive term (i.e., the social cognitive term is to guide particles to move toward the optimal position of the group (the best solution among all particles), so that particles move toward the optimal solution found by the entire particle group, thereby promoting the convergence of the entire particle group). gbest is expressed as the position found by the particle in the previous iteration (i.e., the best position before the current position of x(t)).

[0175] c1 and c2 are represented as learning factors (i.e., c1 represents the social cognitive factor (i.e., the speed of the particles is adjusted according to the global optimal solution); c2 represents the social cognitive factor (i.e., the degree to which the particles are adjusted according to the global optimal solution));

[0176] r1 and r2 are interval random numbers (i.e., used to represent the randomness of particle search in the search space);

[0177] From the above formula, we can see that the inertia term controls the "inertia" of the particle, that is, the tendency of the particle to maintain its current speed, helping the particle to continue moving in its current direction, thereby preventing the particle from suddenly making drastic changes in direction. The inertia factor w controls this tendency. A larger w will cause the particle to stay in its current direction for a longer time, while a smaller w will cause the particle to change direction more quickly. Therefore, the inertia term can avoid overly drastic jumps in the search and maintain the stability of the search.

[0178] The individual cognitive term guides the particle to move toward its own historically optimal solution (individual optimal solution). pbest-x(t) represents the distance between the particle's current position and the historically optimal position. r1 and c1 add randomness and adjust the degree to which the particle adjusts its speed based on personal experience. Therefore, the individual cognitive term encourages the particle to return to the best solution position it has found in history, enhancing the particle's exploration ability, especially to conduct a more refined search near the good solution it has already found.

[0179] The social cognition term guides particles toward the optimal solution (global optimal solution) of the entire particle group; gbest-x(t) represents the distance between the current position of the particle and the global optimal position, and r2 and c2 add randomness and adjust the degree to which the particle adjusts its speed based on the group's experience; therefore, the social cognition term guides particles toward the group's optimal solution, promoting the convergence of the entire particle group toward the global optimal solution, thereby improving the search efficiency of all particles;

[0180] c1 controls how much a particle adjusts its speed based on its own best historical solution, and c2 controls how much a particle adjusts its speed based on the best solution of the entire group. By adjusting the values ​​of these two factors, the degree to which a particle relies on individual experience and group experience can be controlled. c1 and c2 can balance the exploration and development capabilities of a particle, ensuring that it can both explore new possible solutions and fully utilize the good solutions it has already found. Higher c1 or c2 will cause the particle to rely more on its own experience or the experience of the group.

[0181] r1 and r2 are random numbers between [0,1]. By introducing random numbers, the algorithm can introduce a certain degree of randomness, so that the search direction of the particles does not completely depend on the historical optimal solution and the group optimal solution. Random numbers allow particles to have a certain degree of exploration and avoid falling into the local optimal solution.

[0182] S324: Calculating a random adjustment factor in the particle velocity using the particle distribution random factor, the contraction random factor, the inertia factor, and the interval random numbers of r1 and r2;

[0183] It should be noted that the randomness of the position distribution of particles in the search space, the randomness of the search range of particles in the search space, and the inertia factor (i.e., the inertia weight, which determines the randomness of the distribution of particle velocities), as well as r1 and r2 expressed as interval random numbers (i.e., used to represent the randomness of particle searches in the search space), are used to calculate the adjustment factor for some randomness that may be caused in the calculation of particle velocities (i.e., the randomness adjustment factor indicates the errors in the search range and weight distribution caused by randomness when calculating particle velocities);

[0184] Among them, in step S32, it also includes optimizing the alignment adjustment coefficient C according to the particle speed and the latest moving position to obtain the optimal alignment adjustment coefficient C * ,include:

[0185] S325: updating the particle velocity using the randomness adjustment factor to obtain a new particle velocity;

[0186] The position of the particle in the current iteration is updated according to the change amplitude of the new particle velocity to obtain the latest moving position of the particle in the search space;

[0187] Record the time point of the particle's latest moving position in each iterative search in the search space to form a time series;

[0188] It should be noted that the particle's position update is determined by its velocity, and the particle's velocity is adjusted at each iteration based on its historical optimal solution (pbest) and global optimal solution (gbest). If w is small, the particle will rely more on its own pbest and gbest, thus performing a more detailed search near the current position.

[0189] If you want the particle to perform a local search near the current position, you can make the particle's speed smaller so that the particle can only adjust within a local area. If you want the particle to explore a wide area, you can increase the particle's speed so that the particle can jump out of the local area and explore a larger search space. Therefore, the inertia weight w is needed to update the particle's speed, and the particle's position is updated using the particle's updated speed.

[0190] At the same time, the latest moving position of the particle in each iteration is calculated, and the time point when the particle reaches the latest moving position is recorded to form a time series, which can reflect whether there is fluctuation in the random adjustment factor. Uniform fluctuation can indicate that the subsequent alignment adjustment coefficient C is small, and a large time interval indicates that the alignment adjustment coefficient C may be high.

[0191] S326: Optimize the alignment adjustment coefficient C by using the random adjustment factor and the time series * :

[0192] Where n represents the number of time intervals; n-1 represents the adjacent time points; ti represents the latest moving position of the particle at the i-th time point; It is expressed as the logarithm of the time interval change; λ is the randomness adjustment weight; Rtotal is the randomness adjustment factor;

[0193] In the above formula, when When is equal to 1, the adjacent time intervals in the time series change evenly, and the randomness adjustment factor has no fluctuation (i.e., the time interval is small, and the randomness adjustment factor does not have much impact on the particles, that is, it affects the alignment adjustment coefficient C (i.e., the student's grades are real and effective, and there is no phenomenon such as last-minute sharpening)); when When it is equal to 0, the adjacent time intervals in the time series vary greatly, and the random adjustment factor fluctuates (that is, when the random adjustment factor fluctuates, it means that students may have phenomena such as last-minute preparation), and the optimized alignment adjustment coefficient C is obtained. * ;

[0194] In summary, the course assessment and evaluation feedback method proposed in the present invention updates the initial total learning score function by setting the alignment adjustment coefficient for optimization and solving, and generates an initial particle swarm according to the alignment adjustment coefficient. It iteratively calculates the average fitness value of the particles as the actual fitness value of the current iteration number to avoid falling into the local optimal solution, and uses the adaptive shrinkage factor to adjust the number of current iterations. The exponential optimization of the speed and position of the alignment adjustment coefficient C is close to the optimal solution, which improves the accuracy of finding the optimal alignment adjustment coefficient C. * possibility, thereby obtaining the optimal alignment adjustment coefficient C * ; Further through the optimal alignment adjustment coefficient C * The corresponding alignment adjustment coefficient of the target fragile scoring factor is dynamically adjusted to correct the unstable influence of the target fragile scoring factor on the initial total learning scoring function, thereby obtaining the corrected alignment adjustment coefficient of the target fragile scoring factor. Furthermore, by substituting the particles in the initial particle swarm into the initial total learning score function S, we can observe whether these adjustment coefficients have effectively corrected the problems in the initial total learning score function. The modified initial total learning score function is modified twice to obtain the final target total learning score function, which minimizes the impact on student scores.

[0195] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A course assessment and feedback system, characterized by: It includes encoding module, screening module, feedback update processing module and operation output processing module; The encoding module is used to collect and obtain basic information data of students and use the basic information data to generate an initial QR code identification; The screening module is configured to collect first modality evaluation data and second modality evaluation data for the current student; and determine an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data; The feedback update processing module is used to set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; The alignment adjustment coefficient is introduced to correct the target vulnerability scoring factor in the initial total learning scoring function, and the target total learning scoring function is updated after correction; The operation output processing module is used to evaluate the student's score based on the target total learning score function by performing the first modality evaluation data, the second modality evaluation data and the basic information data, output the current student's total evaluation score, and generate a target QR code identifier in combination with the QR code identifier.

2. A course assessment and feedback method, characterized in that: The following steps are included: Collect and obtain basic information data of students, and use the basic information data to generate an initial QR code identification; Collecting first modality evaluation data and second modality evaluation data for the current student; determining an initial total learning scoring function composed of multiple scoring item factors based on the first modality evaluation data and the second modality evaluation data; The initial total learning score function = S = f(x1, x2, x3, x4, x5, x6, ...); where xi is the i-th scoring factor and f is the scoring function; Calculate the conflict degree of the pairwise scoring item factors in the initial total learning scoring function; First, calculate the average conflict degree of each rating item factor with other rating item factors: Where Ci is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function, and N is the total number of scoring factors. A threshold tau is set. If Ci>tau, the scoring factor xi is considered to be a vulnerable scoring factor of the target. Set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function; introduce the alignment adjustment coefficient to correct the target vulnerability score factor in the initial total learning score function, and update the target total learning score function after correction; Based on the target total learning score function, the student's score is evaluated by the first modality evaluation data, the second modality evaluation data and the basic information data, the current student's total evaluation score is output, and the target QR code identifier is generated in combination with the QR code identifier.

3. A course assessment and feedback method according to claim 2, characterized in that: Determining an initial overall learning scoring function composed of a plurality of scoring item factors based on the first modality evaluation data and the second modality evaluation data includes: Video recording of students' offline classroom behavior and electronic monitoring of the video data source; monitoring of offline classroom behavior through the electronic monitoring, and capturing motion and sound of students' offline classroom behavior; Analyzing the effectiveness of the student's offline classroom question answers based on the motion capture and sound capture of the offline classroom behavior, and obtaining first modality evaluation data of the current student's offline classroom behavior based on the question answer effectiveness; Monitor students' online classroom behavior on the online learning platform, analyze students' online learning time, online homework completion, and online review scores based on the current students' access monitoring, and analyze the access records of the access monitoring; The online learning data of the current students’ online classroom behavior is obtained by accessing the records as the second modality evaluation data.

4. A course assessment and feedback method according to claim 3, characterized in that: Determining an initial overall learning scoring function composed of a plurality of scoring item factors based on the first modality evaluation data and the second modality evaluation data includes: Analyze the offline classroom behavior of the first modality evaluation data, score the students' motion capture and sound capture, and obtain a motion standardization score and a discussion interaction score; score the effectiveness of the question answers to obtain a question answer effectiveness score; Analyzing the online classroom behavior of the second modality evaluation data, scoring the online learning time, online homework completion, and online review score to obtain an online learning time score, an online homework completion score, and an online review score; The first modal evaluation data and the score of the first modal evaluation data are used as scoring item factors; the initial total learning scoring function S=f(x1, x2, x3, x4, x5, x6, ...) is calculated using the first modal evaluation data and the scoring item factors of the first modal evaluation data.

5. A course assessment and feedback method according to claim 4, characterized in that: Set the alignment adjustment coefficient to the initial total learning score function, and use the particle swarm optimization algorithm to determine the alignment adjustment coefficient in the initial total learning score function, including: The alignment adjustment coefficient C=(C1, C2, C3, C4, .., C n ), and using a hybrid mapping method to generate a particle swarm with a number of particles corresponding to the alignment adjustment coefficient, each particle representing a candidate solution for the alignment adjustment coefficient; Generate an initial position for the particle swarm to obtain an initial particle swarm; Constraining the position of each particle in the initial particle swarm, iterating each constrained particle, calculating a fitness value for each particle in each iteration, calculating an average of the fitness values ​​calculated for all iterations, obtaining the average fitness value as the actual fitness value of the current iteration, and determining the current number of iterations based on the actual fitness value; The adaptive shrinkage factor is used to adjust the particle speed and the latest moving position of the particles in the initial particle group at the current iteration number; the alignment adjustment coefficient C is optimized by the particle speed and the latest moving position to obtain the optimal alignment adjustment coefficient C. * .

6. A course assessment and feedback method according to claim 5, characterized in that: The alignment adjustment coefficient is introduced to correct the target vulnerability scoring factor in the initial total learning scoring function. After correction, the target total learning scoring function is updated and generated, including: The optimal alignment adjustment coefficient C * Dynamically adjust the alignment adjustment coefficient corresponding to the target vulnerable scoring factor to obtain the modified alignment adjustment coefficient of the target vulnerable scoring factor Where C i is the average conflict degree of the i-th scoring factor xi in the initial total learning scoring function; β is the correction factor; Substitute each particle in the initial particle swarm into the initial total learning score function S, update the initial total learning score function S, and obtain a second total learning score function S'=f(C1x1, C2x2, ..., C n x N );where C n Expressed as the scoring factor x N The alignment adjustment factor; Alignment adjustment coefficient of the score item factor after correction of target vulnerability The second total learning score function S' is updated to obtain the target total learning score function; the expression is: Where, It expresses the alignment adjustment coefficient after optimization of the scoring factor in the target total learning scoring function.

7. A course assessment and feedback method according to claim 6, characterized in that: The adaptive shrinkage factor is used to adjust the particle speed and the latest moving position of the particles in the initial particle group at the current iteration number; the alignment adjustment coefficient C is optimized by the particle speed and the latest moving position to obtain the optimal alignment adjustment coefficient C. * ,include: Calculating the distribution variance of particles in the initial particle swarm to determine the randomness of particles in the search space as a distribution random factor; Perform a wide search for particles based on the minimum contraction factor j1 and the maximum contraction factor j2 of the preset contraction factor range threshold j, record the number of iterations, calculate a nonlinear adaptive contraction factor, and use the adaptive contraction factor as a contraction random factor; The randomness adjustment factor is calculated by combining the distribution random factor and the contraction random factor with the inertia term, the individual cognitive term and the social cognitive term in the search space; The particle velocity is updated according to the randomness adjustment factor, and combined with the position update of the current iteration to obtain the latest moving position of the particle in the search space.

8. A course assessment and feedback method according to claim 7, characterized in that: The alignment adjustment coefficient C is optimized by the particle velocity and the latest moving position to obtain the optimal alignment adjustment coefficient C*, including: Record the time point of the particle's latest moving position in each iterative search in the search space to form a time series; The randomness adjustment factor is used to optimize the alignment adjustment coefficient C* with the time series: Where n represents the number of time intervals; n-1 represents the adjacent time points; ti represents the latest moving position of the particle at the i-th time point; It is expressed as the logarithm of the time interval change; λ is the randomness adjustment weight; Rtotal is the randomness adjustment factor; when When it is equal to 1, the adjacent time intervals in the time series change evenly, and the random adjustment factor has no fluctuation; when When it is equal to 0, the adjacent time intervals in the time series vary greatly, and the random adjustment factor fluctuates, thereby obtaining the optimized alignment adjustment coefficient C*.

Images