Student comprehensive skill evaluation method and system based on machine learning

Abstract

The invention discloses a student comprehensive skill evaluation method and system based on machine learning. The method comprises the following steps: S1, generating a standardized student multi-dimensional behavior data set; s2, converting the student multi-dimensional behavior data set into structured student feature vector data; s3, performing feature screening on the student feature vector data by applying a star-sparrow optimization algorithm to form an optimal feature selection scheme; s4, based on the optimal feature selection scheme, generating a student comprehensive skill evaluation model; s5, generating a student skill dimension quantitative scoring matrix; and S6, predicting the growth trend of the student based on the student skill dimension quantitative scoring matrix and the historical evaluation data, and generating a student comprehensive skill evaluation prediction result. According to the invention, more comprehensive skill evaluation and personalized growth suggestions can be provided for students.

Description

Technical Field

[0001] The present invention relates to the technical field of skill evaluation, and in particular to a method and system for evaluating comprehensive student skills based on machine learning. Background Art

[0002] With the rapid development of artificial intelligence and big data technology, the field of education has gradually shifted from the traditional single-dimensional evaluation method to a more intelligent and multi-dimensional evaluation system. As an important means to measure the overall development level of students, the comprehensive skills evaluation of students not only focuses on students' academic performance, but also covers multiple skill dimensions such as cognitive ability, practical ability, innovation ability, and teamwork ability. However, in the existing education evaluation system, there are still many problems in the evaluation method of students' comprehensive skills, which makes it difficult to fully reflect students' growth and potential.

[0003] In recent years, with the development of machine learning technology, some intelligent assessment tools have begun to be introduced into the field of education, modeling and evaluating student behavior data through big data analysis. However, machine learning technology still has significant defects in practical applications: on the one hand, most of the existing intelligent assessment tools are limited to a single type of data processing, such as test scores or classroom performance data, and fail to effectively integrate multi-dimensional behavior data, resulting in one-sided evaluation results; on the other hand, existing methods lack effective feature screening and modeling mechanisms, making it difficult to extract representative core features from complex student behavior data. In addition, some evaluation models have problems of overfitting or insufficient parameter optimization, resulting in poor generalization ability of the model for data and inability to accurately reflect students' comprehensive skill levels.

[0004] Existing technologies also have major deficiencies in terms of personalized feedback. Traditional evaluation systems can often only give simple skill scores or evaluation results, and lack targeted feedback and improvement suggestions. As a result, students and parents are unable to accurately understand their weak links and improvement directions. It is also difficult for educational institutions to implement personalized teaching based on the specific circumstances of students, limiting the development potential of individual students.

[0005] In summary, the existing technology has the following major problems in the process of comprehensive skills evaluation of students: First, the data source is single, and it is impossible to integrate multi-dimensional data for accurate modeling; second, there is a lack of effective feature screening and modeling mechanism, and the evaluation results are not scientific and accurate enough; third, there is a lack of personalized feedback function, and it is impossible to provide students with targeted improvement suggestions. In view of the above technical defects, there is an urgent need for a comprehensive skills evaluation method and system for students based on machine learning technology and intelligent optimization algorithm to achieve accurate modeling and evaluation of multi-dimensional data. Summary of the invention

[0006] One purpose of the present invention is to propose a method and system for evaluating students' comprehensive skills based on machine learning, which can provide students with more comprehensive skill evaluation and personalized growth suggestions.

[0007] A method for evaluating comprehensive student skills based on machine learning according to an embodiment of the present invention includes the following steps:

[0008] S1. Obtain a multi-dimensional student behavior dataset, perform data cleaning and standardization on the student behavior dataset, and generate a standardized student multi-dimensional behavior dataset;

[0009] S2. Construct a comprehensive skill evaluation index system for students, use the student multi-dimensional behavior data set to extract features of the comprehensive skill evaluation index system for students, extract feature vector data, and convert the student multi-dimensional behavior data set into structured student feature vector data;

[0010] S3, applying the Starfinch optimization algorithm to the student feature vector data to perform feature screening and form the optimal feature selection scheme;

[0011] S4. Based on the optimal feature selection scheme, the unsupervised learning model is used to model the comprehensive skills of students. The parameters of the unsupervised learning model are globally searched and locally optimized by simulating the search behavior in the spring and winter stages through the starfinch optimization algorithm. The parameter configuration of the unsupervised learning model is dynamically adjusted. The optimized unsupervised learning model is trained with the optimal student evaluation feature vector data set to generate a comprehensive skills evaluation model for students.

[0012] S5. Input the students' multi-dimensional behavior data into the students' comprehensive skill evaluation model, output the students' comprehensive skill evaluation results, and quantify the students' skill dimension performance according to the comprehensive skill evaluation results to generate a quantitative scoring matrix for the students' skill dimensions;

[0013] S6. Predict students’ growth trends based on the quantitative scoring matrix of student skill dimensions and historical evaluation data, and generate prediction results for students’ comprehensive skill evaluation.

[0014] Optionally, the S1 specifically includes the following steps:

[0015] S11. Collect multi-dimensional student behavior data to form the initial student behavior data set D raw :

[0016] D raw ={D c ,D l ,D e ,D s};

[0017] Among them, D crepresents the student's classroom performance data, including the frequency of classroom interaction and the submission of classroom assignments, D l represents the learning behavior data of students on the learning platform, including online learning time, course completion and test scores. e Data indicating students’ participation in extracurricular activities, including the number of extracurricular activities, activity types, and activity duration, D s Represents students’ social behavior data, including group cooperation participation, communication interaction frequency, and peer feedback;

[0018] S12, denoising, removing duplicates and missing values ​​from the initial student behavior dataset to generate a cleaned student behavior dataset;

[0019] S13, standardizing the cleaned student behavior data set, mapping data of different dimensions to a unified numerical range, and generating a standardized student multi-dimensional behavior data set;

[0020] S14. Merge the features of the standardized student multi-dimensional behavior data set to maintain the integrity and consistency of the multi-dimensional behavior data and form the final standardized student multi-dimensional behavior data set D final :

[0021]

[0022] in, They respectively represent the results of standardized processing of classroom performance data, learning platform data, extracurricular activity data and social behavior data.

[0023] Optionally, the S2 specifically includes the following steps:

[0024] S21. Construct a comprehensive skill evaluation index system for students. The comprehensive skill evaluation index system for students includes a set of evaluation index sets E. The evaluation index set consists of i comprehensive skill evaluation indexes E. i The comprehensive skill evaluation index of students selects features related to students' comprehensive skills from the standardized multi-dimensional behavior data set of students to measure students' performance in multiple skill dimensions;

[0025] S22. Use the evaluation index set E to evaluate the standardized student multi-dimensional behavior dataset D final Perform feature extraction and convert the student multi-dimensional behavior dataset into structured student feature vector data V:

[0026] V=(v1,v2,…,v k ),1≤i≤k;

[0027] Among them, V is the student feature vector dataset, v i Corresponding to the evaluation index Ei The characteristic value is obtained by combining the student multi-dimensional behavior data with the evaluation index E i The relevant features are obtained after weighted processing;

[0028] S23. Assume that there are n characteristic elements D in the student multi-dimensional behavior data set that can be used for evaluation. final,j (1≤j≤n), for each evaluation index E i Set the corresponding weight coefficient α i,j , α i,j is the evaluation index E i The weight coefficient corresponding to the jth feature element;

[0029] S24, based on the weight coefficient, each evaluation index E i Calculate the corresponding eigenvalue v i :

[0030]

[0031] Among them, v i Corresponding to the evaluation index E i The characteristic value of final,j With weight coefficient α i,j Perform linear weighting to obtain;

[0032] S25, the eigenvalues ​​v1, v2, ..., v corresponding to the evaluation index set k Integrate to form structured student feature vector data V.

[0033] Optionally, S3 specifically includes the following steps:

[0034] S31, input the student feature vector data V into the Starfinch optimization algorithm for feature screening, and define the initial feature vector data V (0) =V;

[0035] S32, in the autumn feature collection phase of the starfinch optimization algorithm, the initial feature vector data V (0) Each feature v in i Calculate the individual contribution R of the feature i :

[0036]

[0037] Among them, R i is the individual contribution of the i-th feature, which is used to measure the basic value of the feature in the evaluation of students' comprehensive skills. n is the number of feature elements that can be used for evaluation in the students' multi-dimensional behavior data set. The individual contribution measures the importance of the feature from the perspective of evaluation indicators and feature weights.

[0038] S33, according to the calculated individual contribution R i Screen the features from the initial feature vector data V (0) The features with individual contribution lower than the predetermined threshold T are eliminated, and the features with high individual contribution are retained to form a preliminary simplified student feature vector subset V';

[0039] S34, in the spring and winter feature search phase of the starfinch optimization algorithm, a set of binary decision variable vectors is defined for the initially streamlined student feature vector subset V' to represent the feature selection scheme W = (w1, w2, ..., w m ):

[0040]

[0041] Among them, m is the number of features remaining after S33 feature screening, w i is the selection mark of the i-th feature, through the selection mark w i The value of the control feature combination scheme searches for the optimal feature configuration in the feature subset;

[0042] S35. Calculate the synergy value S(W) of the feature combination for the feature selection scheme W to measure the positive correlation effect between the selected features. The synergy value provides a reference for the Starfinch Optimization Algorithm to find an optimization scheme in the feature space by measuring the degree of positive matching between different features:

[0043]

[0044] Among them, S(W) is the feature combination synergy value, w p With w q is the feature selection identifier, when w p =1 and w q = 1, the absolute value of the product of the pth and qth features |v ′ p v ′ q | is included in the synergy calculation, m(m-1) / 2 is the total number of feature pairwise combinations, which is used to normalize the synergy;

[0045] S36. Use the Starfinch Optimization Algorithm to iteratively search and update the feature selection scheme W to find the feature combination scheme that makes the synergy value S(W) reach the optimal value. During the iteration process, dynamically adjust the feature selection scheme W to continuously improve the synergy value, and finally obtain the optimal feature selection scheme W. * :

[0046]

[0047] Optionally, the S4 specifically includes the following steps:

[0048] S41, based on the optimal feature selection scheme W * Select from the student feature vector data V Construct the optimal student evaluation feature vector dataset V based on the features * ;

[0049] S42. Using unsupervised learning model to evaluate the optimal student evaluation feature vector dataset V * To build a model, assume the unsupervised learning model parameter set Θ, and define the unsupervised learning objective function L(Θ,V * )for:

[0050]

[0051] in, is the optimal student evaluation feature vector dataset V * The i-th eigenvector in , For unsupervised learning models The reconstruction result of

[0052] S43. Use the Starfinch optimization algorithm to perform a global search on the unsupervised learning model parameter set Θ, and the fitness function F(Θ) is:

[0053]

[0054] Among them, F(Θ) is the fitness value of the unsupervised learning model parameters. The higher the fitness value, the more conducive the parameter configuration is to accurately characterize the student characteristics.

[0055] S44, the parameter update rule in the starfinch optimization algorithm simulates the behavior of autumn and spring and winter seasons and combines the synergy value S (W * ) adjust the parameter update scale, add adaptive factors based on the synergy value in the autumn local exploration and the spring and winter global search, so that the parameter optimization is more in line with the distribution of students' comprehensive skill characteristics:

[0056]

[0057] Θ global,t+1 =Θ t +r2S(W * ) 1 / 2 (Θ rand,t -Θ t );

[0058] Among them, Θ t is the parameter set for the tth iteration, Θ best,t is the parameter set with the best fitness up to the current iteration, Θ rand,t is a randomly selected reference parameter set, r1 and r2 are random factors;

[0059] By connecting feature selection and parameter optimization through the synergy value, the parameter update process is guided by the positive correlation effect between features and matches the distribution of student skill data;

[0060] S45, repeatedly iteratively update the parameter set through the starfinch optimization algorithm, and record the parameter set corresponding to the maximum value of the fitness function F(Θ) as the optimal parameter set Θ * :

[0061]

[0062] S46, the optimized unsupervised learning model parameter set Θ * Applied to unsupervised learning model, the optimal student evaluation feature vector dataset V * Conduct training to generate the student comprehensive skills evaluation model M:

[0063]

[0064] Optionally, the S5 specifically includes the following steps:

[0065] S51, mapping the student multi-dimensional behavior data to the feature selection scheme of the optimal student evaluation feature vector data set, extracting the selected effective features to form a student test feature vector data set;

[0066] S52, inputting the student test feature vector data set into the student comprehensive skill evaluation model, automatically mapping and clustering the student feature data through the student comprehensive skill evaluation model, and outputting the evaluation results of the students in various skill dimensions, including the response values ​​of the cognitive ability, practical ability, innovation ability and teamwork ability skill dimensions, representing the relative performance of the students in each dimension;

[0067] S53, processing the skill dimension evaluation results output by the student comprehensive skill evaluation model, normalizing or weighting the results of each skill dimension through a quantitative algorithm, and generating a quantitative score of the student's skill dimension;

[0068] S54. Summarize the quantitative scores of skill dimensions of a single or multiple students to form a quantitative scoring matrix of student skill dimensions. The quantitative scoring matrix of student skill dimensions records the performance of students in each skill dimension and is used to intuitively display the comprehensive skill evaluation results of individual students or groups of students.

[0069] A student comprehensive skills evaluation system based on machine learning is used to implement a student comprehensive skills evaluation method based on machine learning, including the following modules:

[0070] The data collection and preprocessing module is used to collect students' multi-dimensional behavior data, including classroom performance data, learning platform data, extracurricular activity data and social behavior data, and to clean and standardize the data, remove redundant data, noise data and abnormal data, and generate a standardized student multi-dimensional behavior data set;

[0071] Evaluation index design and feature extraction module, which is used to construct a comprehensive student skill evaluation index system based on educational evaluation theory and psychometric principles. The evaluation index system includes multiple skill dimensions, and the standardized student multi-dimensional behavior data is processed by a feature extraction algorithm to extract structured student feature vector data corresponding to the evaluation index system;

[0072] The feature screening and optimization module is used to input the student feature vector data into the Starfinch Optimization Algorithm for feature screening. The Starfinch Optimization Algorithm includes the autumn feature collection and spring and winter feature search stages. In the autumn feature collection stage, the individual contribution of the features is evaluated, redundant features are removed and high contribution features are retained. In the spring and winter feature search stage, the feature combination is further screened and optimized to form the optimal student evaluation feature vector data set;

[0073] Model building and parameter optimization module, which is used to build an unsupervised learning model based on the optimal student evaluation feature vector data set. The unsupervised learning model uses the starfinch optimization algorithm to perform global parameter search and local optimization, dynamically adjust the model parameter configuration, determine the optimal parameter set through an iterative optimization process, and train the model to generate a student comprehensive skill evaluation model for student comprehensive skill evaluation;

[0074] The skill evaluation and quantitative scoring module is used to input the students' multi-dimensional behavioral data into the student comprehensive skill evaluation model, and output the comprehensive evaluation results of the students in each skill dimension. The evaluation results are quantified to generate quantitative scores of the student skill dimensions, which are used to quantify the students' performance in the cognitive ability, practical ability, innovation ability, and teamwork ability skill dimensions.

[0075] The beneficial effects of the present invention are:

[0076] (1) The present invention introduces a starfinch optimization algorithm in the evaluation of students' comprehensive skills. By simulating the behavior of starfinch in the autumn feature collection and spring and winter position search stages, the student feature vector data is efficiently screened and optimized. In the autumn stage, the feature contribution is calculated to eliminate redundant features and retain high-value features. In the spring and winter stages, the positive correlation of feature combinations is measured by introducing a synergy mechanism to further search for the optimal feature combination to ensure that the selected features have higher representativeness and stability. Compared with the traditional single feature screening method, the starfinch optimization algorithm strikes a balance between global search and local optimization, effectively solves the problem of feature redundancy and model overfitting, and improves the accuracy of feature screening and the generalization ability of the evaluation model.

[0077] (2) During the training process of the unsupervised learning model, the present invention dynamically adjusts the model parameters through the starfinch optimization algorithm, simulates the parameter update strategy combining global search with local optimization, introduces an adaptive factor based on the feature synergy value, accurately controls the parameter update scale, enhances the precision of parameter optimization through local perturbations, and improves the breadth of parameter search by combining global random search, effectively overcoming the local optimal problem in the traditional model training process, and enabling the model to more accurately capture the potential patterns and dimensional relationships of student feature data. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 This is a flow chart of a student comprehensive skills evaluation method and system based on machine learning proposed by the present invention. DETAILED DESCRIPTION

[0080] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0081] refer to Figure 1 , a student comprehensive skills evaluation method based on machine learning, comprising the following steps:

[0082] S1. Obtain a multi-dimensional student behavior dataset, perform data cleaning and standardization on the student behavior dataset, and generate a standardized student multi-dimensional behavior dataset;

[0083] S2. Construct a comprehensive skill evaluation index system for students, use the student multi-dimensional behavior data set to extract features of the comprehensive skill evaluation index system for students, extract feature vector data, and convert the student multi-dimensional behavior data set into structured student feature vector data;

[0084] S3, applying the Starfinch optimization algorithm to the student feature vector data to perform feature screening and form the optimal feature selection scheme;

[0085] S4. Based on the optimal feature selection scheme, the unsupervised learning model is used to model the comprehensive skills of students. The parameters of the unsupervised learning model are globally searched and locally optimized by simulating the search behavior in the spring and winter stages through the starfinch optimization algorithm. The parameter configuration of the unsupervised learning model is dynamically adjusted. The optimized unsupervised learning model is trained with the optimal student evaluation feature vector data set to generate a comprehensive skills evaluation model for students.

[0086] S5. Input the students' multi-dimensional behavior data into the students' comprehensive skill evaluation model, output the students' comprehensive skill evaluation results, and quantify the students' skill dimension performance according to the comprehensive skill evaluation results to generate a quantitative scoring matrix for the students' skill dimensions;

[0087] S6. Predict students’ growth trends based on the quantitative scoring matrix of student skill dimensions and historical evaluation data, and generate prediction results for students’ comprehensive skill evaluation.

[0088] In this implementation, S1 specifically includes the following steps:

[0089] S11. Collect multi-dimensional student behavior data to form the initial student behavior data set D raw :

[0090] D raw ={D c ,D l ,D e ,D s};

[0091] Among them, D c represents the student's classroom performance data, including the frequency of classroom interaction and the submission of classroom assignments, D l represents the learning behavior data of students on the learning platform, including online learning time, course completion and test scores. e Data indicating students’ participation in extracurricular activities, including the number of extracurricular activities, activity types, and activity duration, D s Represents students’ social behavior data, including group cooperation participation, communication interaction frequency, and peer feedback;

[0092] S12, denoising, removing duplicates and missing values ​​from the initial student behavior dataset to generate a cleaned student behavior dataset;

[0093] S13, standardizing the cleaned student behavior data set, mapping data of different dimensions to a unified numerical range, and generating a standardized student multi-dimensional behavior data set;

[0094] S14. Merge the features of the standardized student multi-dimensional behavior data set to maintain the integrity and consistency of the multi-dimensional behavior data and form the final standardized student multi-dimensional behavior data set D final :

[0095]

[0096] in, They respectively represent the results of standardized processing of classroom performance data, learning platform data, extracurricular activity data and social behavior data.

[0097] In this implementation, S2 specifically includes the following steps:

[0098] S21. Construct a comprehensive skill evaluation index system for students. The comprehensive skill evaluation index system for students includes a set of evaluation index sets E. The evaluation index set consists of i comprehensive skill evaluation indexes E. i The comprehensive skill evaluation index of students selects features related to students' comprehensive skills from the standardized multi-dimensional behavior data set of students to measure students' performance in multiple skill dimensions;

[0099] S22. Use the evaluation index set E to evaluate the standardized student multi-dimensional behavior dataset D final Perform feature extraction and convert the student multi-dimensional behavior dataset into structured student feature vector data V:

[0100] V=(v1,v2,…,v k ),1≤i≤k;

[0101] Among them, V is the student feature vector dataset, v i Corresponding to the evaluation index E i The characteristic value is obtained by combining the student multi-dimensional behavior data with the evaluation index E i The relevant features are obtained after weighted processing;

[0102] S23. Assume that there are n characteristic elements D in the student multi-dimensional behavior data set that can be used for evaluation. final,j (1≤j≤n), for each evaluation index E i Set the corresponding weight coefficient α i,j , α i,j is the evaluation index E i The weight coefficient corresponding to the jth feature element;

[0103] S24, based on the weight coefficient, each evaluation index E i Calculate the corresponding eigenvalue v i :

[0104]

[0105] Among them, v i Corresponding to the evaluation index E i The characteristic value of final,j With weight coefficient α i,j Perform linear weighting to obtain;

[0106] S25, the eigenvalues ​​v1, v2, ..., v corresponding to the evaluation index set k Integrate to form structured student feature vector data V.

[0107] In this implementation, S3 specifically includes the following steps:

[0108] S31, input the student feature vector data V into the Starfinch optimization algorithm for feature screening, and define the initial feature vector data V (0) =V;

[0109] S32, in the autumn feature collection phase of the starfinch optimization algorithm, the initial feature vector data V (0) Each feature v in i Calculate the individual contribution R of the feature i :

[0110]

[0111] Among them, R i is the individual contribution of the i-th feature, which is used to measure the basic value of the feature in the evaluation of students' comprehensive skills. n is the number of feature elements that can be used for evaluation in the students' multi-dimensional behavior data set. The individual contribution measures the importance of the feature from the perspective of evaluation indicators and feature weights.

[0112] S33, according to the calculated individual contribution R i Screen the features from the initial feature vector data V (0) The features with individual contribution lower than the predetermined threshold T are eliminated, and the features with high individual contribution are retained to form a preliminary simplified student feature vector subset V';

[0113] S34, in the spring and winter feature search phase of the starfinch optimization algorithm, a set of binary decision variable vectors is defined for the initially streamlined student feature vector subset V' to represent the feature selection scheme W = (w1, w2, ..., w m ):

[0114]

[0115] Among them, m is the number of features remaining after S33 feature screening, w i is the selection mark of the i-th feature, through the selection mark wi The value of the control feature combination scheme searches for the optimal feature configuration in the feature subset;

[0116] S35. Calculate the synergy value S(W) of the feature combination for the feature selection scheme W to measure the positive correlation effect between the selected features. The synergy value provides a reference for the Starfinch Optimization Algorithm to find an optimization scheme in the feature space by measuring the degree of positive matching between different features:

[0117]

[0118] Among them, S(W) is the feature combination synergy value, w p With w q is the feature selection identifier, when w p =1 and w q = 1, the absolute value of the product of the pth and qth features |v ′ p v ′ q | is included in the synergy calculation, m(m-1) / 2 is the total number of feature pairwise combinations, which is used to normalize the synergy;

[0119] S36. Use the Starfinch Optimization Algorithm to iteratively search and update the feature selection scheme W to find the feature combination scheme that makes the synergy value S(W) reach the optimal value. During the iteration process, dynamically adjust the feature selection scheme W to continuously improve the synergy value, and finally obtain the optimal feature selection scheme W. * :

[0120]

[0121] In this implementation, S4 specifically includes the following steps:

[0122] S41, based on the optimal feature selection scheme W * Select from the student feature vector data V Construct the optimal student evaluation feature vector dataset V based on the features * ;

[0123] S42. Using unsupervised learning model to evaluate the optimal student evaluation feature vector dataset V * To build a model, assume the unsupervised learning model parameter set Θ, and define the unsupervised learning objective function L(Θ,V * )for:

[0124]

[0125] in, is the optimal student evaluation feature vector dataset V * The i-th eigenvector in , For unsupervised learning models The reconstruction result of

[0126] S43. Use the Starfinch optimization algorithm to perform a global search on the unsupervised learning model parameter set Θ, and the fitness function F(Θ) is:

[0127]

[0128] Among them, F(Θ) is the fitness value of the unsupervised learning model parameters. The higher the fitness value, the more conducive the parameter configuration is to accurately characterize the student characteristics.

[0129] S44, the parameter update rule in the starfinch optimization algorithm simulates the behavior of autumn and spring and winter seasons and combines the synergy value S (W * ) adjust the parameter update scale, add adaptive factors based on the synergy value in the autumn local exploration and the spring and winter global search, so that the parameter optimization is more in line with the distribution of students' comprehensive skill characteristics:

[0130]

[0131] Θ global,t+1 =Θ t +r2S(W * ) 1 / 2 (Θ rand,t -Θ t );

[0132] Among them, Θ t is the parameter set for the tth iteration, Θ best,t is the parameter set with the best fitness up to the current iteration, Θ rand,t is a randomly selected reference parameter set, r1 and r2 are random factors;

[0133] By connecting feature selection and parameter optimization through the synergy value, the parameter update process is guided by the positive correlation effect between features and matches the distribution of student skill data;

[0134] S45, repeatedly iteratively update the parameter set through the starfinch optimization algorithm, and record the parameter set corresponding to the maximum value of the fitness function F(Θ) as the optimal parameter set Θ * :

[0135]

[0136] S46, the optimized unsupervised learning model parameter set Θ * Applied to unsupervised learning model, the optimal student evaluation feature vector dataset V * Conduct training to generate the student comprehensive skills evaluation model M:

[0137]

[0138] In this implementation, S5 specifically includes the following steps:

[0139] S51, mapping the student multi-dimensional behavior data to the feature selection scheme of the optimal student evaluation feature vector data set, extracting the selected effective features to form a student test feature vector data set;

[0140] S52, inputting the student test feature vector data set into the student comprehensive skill evaluation model, automatically mapping and clustering the student feature data through the student comprehensive skill evaluation model, and outputting the evaluation results of the students in various skill dimensions, including the response values ​​of the cognitive ability, practical ability, innovation ability and teamwork ability skill dimensions, representing the relative performance of the students in each dimension;

[0141] S53, processing the skill dimension evaluation results output by the student comprehensive skill evaluation model, normalizing or weighting the results of each skill dimension through a quantitative algorithm, and generating a quantitative score of the student's skill dimension;

[0142] S54. Summarize the quantitative scores of skill dimensions of a single or multiple students to form a quantitative scoring matrix of student skill dimensions. The quantitative scoring matrix of student skill dimensions records the performance of students in each skill dimension and is used to intuitively display the comprehensive skill evaluation results of individual students or groups of students.

[0143] A student comprehensive skills evaluation system based on machine learning is used to implement a student comprehensive skills evaluation method based on machine learning, including the following modules:

[0144] The data collection and preprocessing module is used to collect students' multi-dimensional behavior data, including classroom performance data, learning platform data, extracurricular activity data, and social behavior data, and to clean and standardize the data, remove redundant data, noise data, and abnormal data, and generate a standardized student multi-dimensional behavior data set;

[0145] Evaluation index design and feature extraction module, which is used to construct a comprehensive student skill evaluation index system based on educational evaluation theory and psychometric principles. The evaluation index system includes multiple skill dimensions. The standardized student multi-dimensional behavior data is processed through feature extraction algorithms to extract structured student feature vector data corresponding to the evaluation index system.

[0146] The feature screening and optimization module is used to input the student feature vector data into the Starfinch Optimization Algorithm for feature screening. The Starfinch Optimization Algorithm includes the autumn feature collection and spring and winter feature search stages. In the autumn feature collection stage, the individual contribution of the features is evaluated, redundant features are removed and high contribution features are retained. In the spring and winter feature search stage, the feature combination is further screened and optimized to form the optimal student evaluation feature vector data set;

[0147] Model building and parameter optimization module, which is used to build an unsupervised learning model based on the optimal student evaluation feature vector data set. The unsupervised learning model uses the starfinch optimization algorithm to perform global parameter search and local optimization, dynamically adjust the model parameter configuration, determine the optimal parameter set through an iterative optimization process, and train the model to generate a student comprehensive skill evaluation model for student comprehensive skill evaluation;

[0148] The skill evaluation and quantitative scoring module is used to input students' multi-dimensional behavioral data into the student comprehensive skill evaluation model, and output the comprehensive evaluation results of students in each skill dimension. The evaluation results are quantified to generate quantitative scores of student skill dimensions, which are used to quantify students' performance in cognitive ability, practical ability, innovation ability, and teamwork ability skill dimensions.

[0149] Embodiment 1:

[0150] This embodiment describes the specific application process of the present invention for the scenario of comprehensive skill evaluation of students based on the combination of machine learning and the Starfinch optimization algorithm, and verifies the feasibility and effectiveness of the present invention in solving the problems of the prior art, improving the evaluation accuracy and providing personalized feedback.

[0151] A key middle school in a certain city plans to conduct a comprehensive skills evaluation for first-year high school students. The goal is to comprehensively analyze students' cognitive ability, practical ability, innovation ability and teamwork ability to guide the formulation of subsequent personalized teaching plans, help students accurately identify weak links and provide targeted improvement suggestions. There are 500 students in this grade. In order to improve the evaluation results, the school introduced the student comprehensive skills evaluation method based on machine learning and Starfinch optimization algorithm of the present invention, combined with multi-dimensional data such as classroom performance data, learning platform data, extracurricular activity data and social behavior data, to build a scientific, dynamic and accurate comprehensive skills evaluation system.

[0152] During the implementation of the project, the school collected student behavior data from the following dimensions:

[0153] Classroom performance data: including classroom interaction frequency, homework submission rate, and teacher classroom evaluation, with a total of 5,000 data items collected;

[0154] Learning platform data: including online learning time, test scores, and course completion, with a total of 4,500 pieces of data collected;

[0155] Extracurricular activity data: including the number of club activities that students participated in, the duration of activities, and the types of activities, with a total of 3,000 pieces of data collected;

[0156] Social behavior data: including group cooperation scores, classmates’ evaluations and communication and interaction times, with a total of 2,500 pieces of data collected.

[0157] Through data cleaning and standardization, invalid data, noise data and redundant data are removed, and finally a high-quality standardized student behavior data set is formed.

[0158] Based on the theory of educational assessment and the principles of psychometrics, four main evaluation dimensions were constructed: cognitive ability, practical ability, innovation ability, and teamwork ability. Multiple specific indicators were set under each dimension:

[0159] Cognitive ability: frequency of classroom interaction, average test scores;

[0160] Practical ability: number of extracurricular activities participated in and duration of activities;

[0161] Innovation ability: the quality of homework completion and leadership performance in club activities;

[0162] Teamwork ability: group cooperation score, number of communication interactions.

[0163] Through the feature extraction algorithm, the student feature vector data set related to each evaluation index is extracted to form a structured feature vector.

[0164] The starfinch optimization algorithm is used to screen the feature vector data set, simulating the "autumn feature collection" and "spring and winter location search" stages, eliminating redundant features and retaining high-contribution feature combinations, and finally obtaining the optimal student evaluation feature vector data set. The parameters of the unsupervised learning model are further optimized based on the starfinch optimization algorithm. The optimal student evaluation feature vector data set is input into the unsupervised learning model for training to generate a student comprehensive skill evaluation model. The training process uses 70% of the student data as the training set and 30% of the data as the test set. The students' multi-dimensional behavioral data are input into the trained evaluation model, and the evaluation results of each skill dimension are output. The evaluation results are quantified and scored to generate a quantitative scoring matrix for the student skill dimensions.

[0165] In order to verify the effectiveness of the present invention, the method of the present invention is compared with the traditional evaluation method, and 100 students are selected for the experiment. In terms of the number of data dimensions, the present invention adopts data of four dimensions: classroom performance, learning platform, extracurricular activities, and social behavior, while the traditional method only uses classroom performance and test scores. In terms of model accuracy, the traditional method relies on subjective scoring with an accuracy of 76.2%. The present invention realizes intelligent evaluation through a machine learning model with an accuracy of 92.4%. In terms of the error rate of the evaluation result, the traditional method has subjectivity and data uniformity, with a high error rate of 15.8%. The present invention improves the feature screening accuracy through the starfinch optimization algorithm, and the error rate is reduced to 4.5%. In terms of feedback generation time, the traditional method takes 7 days for manual scoring, while the present invention automatically generates feedback through the model, which only takes 2 days to complete.

[0166] The present invention realizes accurate evaluation of students' comprehensive skills through this embodiment, and solves the problems of single evaluation, strong subjectivity and insufficient feedback of traditional methods. Experimental data show that the method of the present invention is superior to traditional methods in data processing capability, accuracy of evaluation results and feedback generation efficiency. It can provide students with more comprehensive skill evaluation and personalized growth suggestions, and has high application value and promotion potential.

[0167] The present invention introduces a starfinch optimization algorithm in the evaluation of students' comprehensive skills. By simulating the behaviors of starfinch in the autumn feature collection and spring and winter position search stages, student feature vector data is efficiently screened and optimized. In the autumn stage, feature contribution is calculated to eliminate redundant features and retain high-value features. In the spring and winter stages, a synergy mechanism is introduced to measure the positive correlation of feature combinations, and further search for the optimal feature combination to ensure that the selected features have higher representativeness and stability. Compared with the traditional single feature screening method, the starfinch optimization algorithm strikes a balance between global search and local optimization, effectively solves the problems of feature redundancy and model overfitting, and improves the accuracy of feature screening and the generalization ability of the evaluation model.

[0168] During the training process of the unsupervised learning model, the present invention dynamically adjusts the model parameters through the starfinch optimization algorithm, simulates the parameter updating strategy combining global search with local optimization, introduces an adaptive factor based on the feature synergy value, accurately controls the parameter updating scale, enhances the precision of parameter optimization through local perturbations, and improves the breadth of parameter search by combining global random search, thus effectively overcoming the local optimal problem in the traditional model training process and enabling the model to more accurately capture the potential patterns and dimensional relationships of student feature data.

[0169] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating students' comprehensive skills based on machine learning, characterized in that: The steps include: S1. Obtain a multi-dimensional student behavior dataset, perform data cleaning and standardization on the student behavior dataset, and generate a standardized student multi-dimensional behavior dataset; S2. Construct a comprehensive skill evaluation index system for students, use the student multi-dimensional behavior data set to extract features of the comprehensive skill evaluation index system for students, extract feature vector data, and convert the student multi-dimensional behavior data set into structured student feature vector data; S3, applying the Starfinch optimization algorithm to the student feature vector data to perform feature screening and form the optimal feature selection scheme; S4. Based on the optimal feature selection scheme, the unsupervised learning model is used to model the comprehensive skills of students. The parameters of the unsupervised learning model are globally searched and locally optimized by simulating the search behavior in the spring and winter stages through the starfinch optimization algorithm. The parameter configuration of the unsupervised learning model is dynamically adjusted. The optimized unsupervised learning model is trained with the optimal student evaluation feature vector data set to generate a comprehensive skills evaluation model for students. S5. Input the students' multi-dimensional behavior data into the students' comprehensive skill evaluation model, output the students' comprehensive skill evaluation results, and quantify the students' skill dimension performance according to the comprehensive skill evaluation results to generate a quantitative scoring matrix for the students' skill dimensions; S6. Predict students’ growth trends based on the quantitative scoring matrix of student skill dimensions and historical evaluation data, and generate prediction results for students’ comprehensive skill evaluation.

2. According to the method for evaluating comprehensive student skills based on machine learning in claim 1, it is characterized in that: The S1 specifically includes the following steps: S11. Collect multi-dimensional student behavior data to form the initial student behavior data set D raw : D raw ={D c ,D l ,D e ,D s }; Among them, D c represents the student's classroom performance data, including the frequency of classroom interaction and the submission of classroom assignments, D l represents the learning behavior data of students on the learning platform, including online learning time, course completion and test scores. e Data indicating students’ participation in extracurricular activities, including the number of extracurricular activities, activity types, and activity duration, D s Represents students’ social behavior data, including group cooperation participation, communication interaction frequency, and peer feedback; S12, denoising, removing duplicates and missing values ​​from the initial student behavior dataset to generate a cleaned student behavior dataset; S13, standardizing the cleaned student behavior data set, mapping data of different dimensions to a unified numerical range, and generating a standardized student multi-dimensional behavior data set; S14. Merge the features of the standardized student multi-dimensional behavior data set to maintain the integrity and consistency of the multi-dimensional behavior data and form the final standardized student multi-dimensional behavior data set D final : in, They respectively represent the results of standardized processing of classroom performance data, learning platform data, extracurricular activity data and social behavior data.

3. The method for evaluating students' comprehensive skills based on machine learning according to claim 1, characterized in that: The S2 specifically includes the following steps: S21. Construct a comprehensive skill evaluation index system for students. The comprehensive skill evaluation index system for students includes a set of evaluation index sets E. The evaluation index set consists of i comprehensive skill evaluation indexes E. i The comprehensive skill evaluation index of students selects features related to students' comprehensive skills from the standardized multi-dimensional behavior data set of students to measure students' performance in multiple skill dimensions; S22. Use the evaluation index set E to evaluate the standardized student multi-dimensional behavior dataset D final Perform feature extraction and convert the student multi-dimensional behavior dataset into structured student feature vector data V: V=(v1,v2,…,v k ),1≤i≤k; Among them, V is the student feature vector dataset, v i Corresponding to the evaluation index E i The characteristic value is obtained by combining the student multi-dimensional behavior data with the evaluation index E i The relevant features are obtained after weighted processing; S23. Assume that there are n characteristic elements D in the student multi-dimensional behavior data set that can be used for evaluation. final,j (1≤j≤n), for each evaluation index E i Set the corresponding weight coefficient α i,j , α i,j is the evaluation index E i The weight coefficient corresponding to the jth feature element; S24, based on the weight coefficient, each evaluation index E i Calculate the corresponding eigenvalue v i : Among them, v i Corresponding to the evaluation index E i The characteristic value of final,j With weight coefficient α i,j Perform linear weighting to obtain; S25, the eigenvalues ​​v1, v2, ..., v corresponding to the evaluation index set k Integrate to form structured student feature vector data V.

4. The method for evaluating comprehensive student skills based on machine learning according to claim 1, characterized in that: The S3 specifically includes the following steps: S31, input the student feature vector data V into the Starfinch optimization algorithm for feature screening, and define the initial feature vector data V (0) =V; S32, in the autumn feature collection phase of the starfinch optimization algorithm, the initial feature vector data V (0) Each feature v in i Calculate the individual contribution R of the feature i : Among them, R i is the individual contribution of the i-th feature, which is used to measure the basic value of the feature in the evaluation of students' comprehensive skills. n is the number of feature elements that can be used for evaluation in the students' multi-dimensional behavior data set. The individual contribution measures the importance of the feature from the perspective of evaluation indicators and feature weights. S33, according to the calculated individual contribution R i Screen the features from the initial feature vector data V (0) The features with individual contribution lower than the predetermined threshold T are eliminated, and the features with high individual contribution are retained to form a preliminary simplified student feature vector subset V'; S34, in the spring and winter feature search phase of the starfinch optimization algorithm, a set of binary decision variable vectors is defined for the initially streamlined student feature vector subset V' to represent the feature selection scheme W = (w1, w2, ..., w m ): Among them, m is the number of features remaining after S33 feature screening, w i is the selection mark of the i-th feature, through the selection mark w i The value of the control feature combination scheme searches for the optimal feature configuration in the feature subset; S35. Calculate the synergy value S(W) of the feature combination for the feature selection scheme W to measure the positive correlation effect between the selected features. The synergy value provides a reference for the Starfinch Optimization Algorithm to find an optimization scheme in the feature space by measuring the degree of positive matching between different features: Among them, S(W) is the feature combination synergy value, w p With w q is the feature selection identifier, when w p =1 and w q = 1, the absolute value of the product of the pth and qth features |v ′ p v ′ q | is included in the synergy calculation, m(m-1) / 2 is the total number of feature pairwise combinations, which is used to normalize the synergy; S36. Use the Starfinch Optimization Algorithm to iteratively search and update the feature selection scheme W to find the feature combination scheme that makes the synergy value S(W) reach the optimal value. During the iteration process, dynamically adjust the feature selection scheme W to continuously improve the synergy value, and finally obtain the optimal feature selection scheme W. * :

5. The method for evaluating comprehensive student skills based on machine learning according to claim 1, characterized in that: The S4 specifically comprises the following steps: S41, based on the optimal feature selection scheme W * Select from the student feature vector data V Construct the optimal student evaluation feature vector dataset V based on the features * ; S42. Using unsupervised learning model to evaluate the optimal student evaluation feature vector dataset V * To build a model, assume the unsupervised learning model parameter set Θ, and define the unsupervised learning objective function L(Θ,V * )for: in, is the optimal student evaluation feature vector dataset V * The i-th eigenvector in , For unsupervised learning models The reconstruction result of S43. Use the starfinch optimization algorithm to perform a global search on the unsupervised learning model parameter set Θ, and the fitness function F(Θ) is: Among them, F(Θ) is the fitness value of the unsupervised learning model parameters. The higher the fitness value, the more conducive the parameter configuration is to accurately characterize the student characteristics. S44, the parameter update rule in the starfinch optimization algorithm simulates the behavior of autumn and spring and winter seasons and combines the synergy value S (W * ) adjust the parameter update scale, add adaptive factors based on the synergy value in the autumn local exploration and the spring and winter global search, so that the parameter optimization is more in line with the distribution of students' comprehensive skill characteristics: I global,t+1 =Θ t +r2S(W * ) 1 / 2 (I rand,t -I t ); Among them, Θ t is the parameter set for the tth iteration, Θ best,t is the parameter set with the best fitness up to the current iteration, Θ rand,t is a randomly selected reference parameter set, r1 and r2 are random factors; By connecting feature selection and parameter optimization through the synergy value, the parameter update process is guided by the positive correlation effect between features and matches the distribution of student skill data; S45, repeatedly iteratively update the parameter set through the starfinch optimization algorithm, and record the parameter set corresponding to the maximum value of the fitness function F(Θ) as the optimal parameter set Θ * : S46, the optimized unsupervised learning model parameter set Θ * Applied to unsupervised learning model, the optimal student evaluation feature vector dataset V * Conduct training to generate the student comprehensive skills evaluation model M:

6. The method for evaluating comprehensive student skills based on machine learning according to claim 1, characterized in that: The S5 specifically includes the following steps: S51, mapping the student multi-dimensional behavior data to the feature selection scheme of the optimal student evaluation feature vector data set, extracting the selected effective features to form a student test feature vector data set; S52, inputting the student test feature vector data set into the student comprehensive skill evaluation model, automatically mapping and clustering the student feature data through the student comprehensive skill evaluation model, and outputting the evaluation results of the students in various skill dimensions, including the response values ​​of the cognitive ability, practical ability, innovation ability and teamwork ability skill dimensions, representing the relative performance of the students in each dimension; S53, processing the skill dimension evaluation results output by the student comprehensive skill evaluation model, normalizing or weighting the results of each skill dimension through a quantitative algorithm, and generating a quantitative score of the student's skill dimension; S54. Summarize the quantitative scores of skill dimensions of a single or multiple students to form a quantitative scoring matrix of student skill dimensions. The quantitative scoring matrix of student skill dimensions records the performance of students in each skill dimension and is used to intuitively display the comprehensive skill evaluation results of individual students or groups of students.

7. A student comprehensive skills evaluation system based on machine learning, used to execute a student comprehensive skills evaluation method based on machine learning as claimed in any one of claims 1 to 6, characterized in that: Includes the following modules: The data collection and preprocessing module is used to collect students' multi-dimensional behavior data, including classroom performance data, learning platform data, extracurricular activity data and social behavior data, and to clean and standardize the data, remove redundant data, noise data and abnormal data, and generate a standardized student multi-dimensional behavior data set; Evaluation index design and feature extraction module, which is used to construct a comprehensive student skill evaluation index system based on educational evaluation theory and psychometric principles. The evaluation index system includes multiple skill dimensions, and the standardized student multi-dimensional behavior data is processed by a feature extraction algorithm to extract structured student feature vector data corresponding to the evaluation index system; The feature screening and optimization module is used to input the student feature vector data into the Starfinch Optimization Algorithm for feature screening. The Starfinch Optimization Algorithm includes the autumn feature collection and spring and winter feature search stages. In the autumn feature collection stage, the individual contribution of the features is evaluated, redundant features are removed and high contribution features are retained. In the spring and winter feature search stage, the feature combination is further screened and optimized to form the optimal student evaluation feature vector data set; Model building and parameter optimization module, which is used to build an unsupervised learning model based on the optimal student evaluation feature vector data set. The unsupervised learning model uses the starfinch optimization algorithm to perform global parameter search and local optimization, dynamically adjust the model parameter configuration, determine the optimal parameter set through an iterative optimization process, and train the model to generate a student comprehensive skill evaluation model for student comprehensive skill evaluation; The skill evaluation and quantitative scoring module is used to input the students' multi-dimensional behavioral data into the student comprehensive skill evaluation model, and output the comprehensive evaluation results of the students in each skill dimension. The evaluation results are quantified to generate quantitative scores of the student skill dimensions, which are used to quantify the students' performance in the cognitive ability, practical ability, innovation ability, and teamwork ability skill dimensions.

Images