A learning quality evaluation system and method based on deep learning

Abstract

The present invention discloses a learning quality evaluation system and method based on deep learning, which belongs to the field of learning quality evaluation. The method includes obtaining a knowledge point support relationship based on a time sequence based on the knowledge points that students have learned and the correlation between the knowledge points; obtaining a learning attitude vector based on the knowledge points that students have learned; constructing a skill support relationship based on the knowledge point reserves required by each skill of the students based on the knowledge points that students have learned; calculating a learning quality score based on the knowledge point support relationship based on the time sequence, the learning attitude vector and the skill support relationship, and completing the learning quality evaluation. The present invention solves the problem that the existing learning quality evaluation method does not consider the mutual influence between learning attitude, knowledge point support relationship and skills when considering the comprehensive quality of students.

Description

Technical Field

[0001] The present invention belongs to the field of learning quality evaluation, and in particular, relates to a learning quality evaluation system and method based on deep learning. Background Art

[0002] In teaching work, the evaluation of students' learning quality requires scientific standards. Nowadays, most evaluation methods are single, through the assessment of each subject separately, based on the paper scores to measure the students' mastery of the content of each subject. This method can only evaluate the students' mastery of the test points of a single subject. Such an evaluation is flat. When students face a professional course system, for example, during college, the knowledge points of various courses overlap, and the paper scores of a single course cannot measure the students' mastery of the knowledge points. Therefore, how to evaluate students' learning quality has become an urgent problem to be solved.

[0003] An existing evaluation system calculates academic achievement and academic level value-added through students' paper scores, and calculates the correlation of each subject based on the scores of multiple tests in each subject as the degree of synergy between subjects. It uses academic achievement, academic level value-added and synergy to evaluate whether students are biased in one subject and whether the development of each subject is coordinated. The disadvantage of this system is that the development between subjects is only judged through the paper, and it is impossible to clearly judge the weak points of students. The evaluation is based solely on the paper scores of the subject, which is accidental.

[0004] An existing method for evaluating classroom teaching quality measures students' concentration based on their listening status, and determines the quality of classroom teaching based on their grades and concentration. This method introduces the influence of students' learning status, but only considers the immediate evaluation of the classroom, and cannot evaluate students' entire learning stage across time and space. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a learning quality evaluation system based on deep learning, which solves the problem that the existing learning quality evaluation methods do not consider the mutual influence between learning attitude, knowledge point support relationship and skills when considering students' comprehensive quality.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a learning quality evaluation system based on deep learning, including a knowledge point support relationship quantification module, a learning attitude quantification module, a skill support relationship quantification module, an answer situation recording module, a question difficulty coefficient calculation module, a score prediction module, and a learning quality evaluation module;

[0007] The knowledge point support relationship quantification module is used to obtain the knowledge point support relationship based on the time series according to the knowledge points that the students have learned and the association relationship between the knowledge points, and update it according to the prediction error;

[0008] The learning attitude quantification module is used to quantify the learning attitude of the students according to the knowledge points that the students have learned, based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector; the expression of the learning attitude vector is:

[0009]

[0010] in, is the learning attitude vector; To learn the same knowledge time interval; The time interval for learning; To repeat the learning of knowledge points The number of times; For knowledge points Number of learning times; For knowledge points Number of learning times; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the knowledge point support conversion coefficient; is the floor rounding symbol;

[0011] The skill support relationship quantification module is used to construct a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill of the students; the expression of the skill support relationship is:

[0012]

[0013] in, Support relationships for skills; For the Skill pair The support level of each skill; For the The skill required Knowledge point Skill No. The support degree of each knowledge point;

[0014] The answer status recording module is used to obtain the student's answered question database, and obtain the knowledge points that the student has learned, the knowledge points involved in each question, and the score based on the student's answered question database;

[0015] The module for calculating the difficulty coefficient of the question is used to obtain the average score of each knowledge point according to the knowledge points and scores involved in each question by using the knowledge point support relationship based on time series, and normalize the average score of each knowledge point as the student's mastery of each knowledge point, and update the student's mastery of each knowledge point according to the student's mastery of each knowledge point and the learning attitude vector by using the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting, and calculate the difficulty coefficient of each question according to the updated student's mastery of each knowledge point; the method of obtaining the average score of each knowledge point by using the knowledge point support relationship based on time series is specifically as follows:

[0016] Obtain the student's answered question database, and obtain the student's answer status based on the student's answered question database;

[0017] According to the students' answers, the average score of each knowledge point is obtained by using the time-sequence-based knowledge point support relationship and considering the influence of the support relationship between knowledge points:

[0018]

[0019]

[0020]

[0021]

[0022] in, The knowledge points that students have learned The average score of For the Knowledge points involved in the question Impact score; For knowledge points involved Total number of topics; For the The score of the question; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The knowledge points involved in the question; A bank of answered questions for students; For the Question:

[0023] Normalize the average score of each knowledge point as the students' mastery of each knowledge point;

[0024] The score prediction module is used to obtain the current question from the student's answered question database, and calculate the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculate the prediction error;

[0025] The learning quality evaluation module is used to calculate the learning quality score and complete the learning quality evaluation according to the knowledge point support relationship and skill support relationship based on the time sequence.

[0026] The beneficial effects of the present invention are as follows: the system introduces the learning frequency of knowledge points to measure the learning attitude of students. At the same time, based on the mastery of the knowledge points set by the course and the correlation between the skills outside the course, the entire learning quality of the students' learning stage is evaluated, and a more comprehensive consideration is given. At the same time, the correlation between the three is introduced, and the learning attitude affects the mastery of the knowledge points, and the mastery of the knowledge points will affect the support relationship of the knowledge points from the student's personal perspective, and the support relationship of the skills will be affected together. Such a system setting can be personalized for student characteristics. The evaluation is more accurate while being fully linked. The learning attitude is measured by using the learning frequency. When considering the number of times a specific knowledge point is learned, the knowledge points with support relationships are taken into account at the same time, so as to strengthen the correlation between the knowledge points and improve the accuracy of the description of the review effect of the knowledge points. The skills outside the course are associated through the knowledge reserves required for the skills, which solves the problem of difficulty in quantifying the skills. At the same time, due to the support of the knowledge point support relationship based on the time series, the quantification accuracy of the skill support relationship is improved. Students' mastery of knowledge points is usually reflected through answering questions. Considering the diversity of knowledge points contained in the questions and the different emphases, we introduce a time-series-based knowledge point support relationship to separate the knowledge points from the questions and obtain the mastery of a single knowledge point, in preparation for subsequent score prediction.

[0027] The present invention provides a learning quality evaluation method, comprising the following steps:

[0028] S1. Initialize the knowledge point support relationship based on time sequence according to the knowledge points that students have learned;

[0029] S2, according to the knowledge points that the students have learned, quantify the students' learning attitude based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector;

[0030] S3. Build a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill;

[0031] S4, obtaining the student's answered question database and, based on the student's answer situation, using the time-series-based knowledge point support relationship to obtain the average score of each knowledge point, and normalizing the average score of each knowledge point as the student's mastery of each knowledge point;

[0032] S5. Based on the students’ mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students’ mastery of each knowledge point. Based on the updated mastery of each knowledge point by the students, the difficulty coefficient of each question is calculated.

[0033] S6, obtaining the current question from the student's answered question database, and calculating the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculating the prediction error;

[0034] S7, determine whether the prediction error is less than the error threshold, if so, return to step S6 to predict the score of the next question, otherwise, update the knowledge point support relationship based on the time series, and return to step S2, until the prediction error of each question in the student's answer database is less than the error threshold, and enter step S8;

[0035] S8. Calculate the learning quality score based on the time-series knowledge point support relationship and skill support relationship to complete the learning quality evaluation.

[0036] The beneficial effects of the present invention are as follows: the present method evaluates the learning quality of students in the entire learning stage from three perspectives: the mastery of knowledge points, the relevance of skills and learning attitude, and quantifies the learning attitude using the students' learning frequency. At the same time, the learning attitude affects the mastery of knowledge points, the supporting relationship of knowledge points and the supporting relationship of skills. The three perspectives are interrelated and can evaluate the learning quality of students in a personalized and comprehensive manner.

[0037] Furthermore, the expression of the knowledge point support relationship based on time sequence in step S1 is:

[0038]

[0039] in, Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; and They are all knowledge point numbers, and the knowledge points are sorted from small to large in the order of learning.

[0040] The beneficial effect of the above further scheme is: taking the characteristics of learning from easy to difficult into account in the support description, and placing temporal constraints on the support relationship of knowledge points, so as to describe the support relationship of knowledge points more accurately and objectively.

[0041] Furthermore, the step S5 is specifically as follows:

[0042] S501. Based on the students' mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students' mastery of each knowledge point:

[0043]

[0044] in, To update students' understanding of knowledge points degree of mastery; To update students' knowledge points degree of mastery; Erase vector for memory; Forget vector for memory; is the sigmoid activation function; The transpose of the fully connected layer weight matrix for memory erasure; Factors that influence the degree of forgetting; is the bias vector for memory erasure; is the activation function; The transpose of the fully connected layer weight matrix updated for memory; Bias vector for memory update; is the learning attitude vector;

[0045] S502. Calculate the difficulty coefficient of each question based on the updated student's mastery of each knowledge point:

[0046]

[0047] in, For the The difficulty level of the question; is the normalization function; The knowledge points that students have learned The average score of For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

[0048] The beneficial effects of the above further scheme are: utilizing the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting, combined with learning attitude, to quantify the forgetting of knowledge points and update students' mastery of knowledge points at the same time.

[0049] Furthermore, the expression for predicting the answer score in step S6 is:

[0050]

[0051] in, For the The predicted score of the question; is the total score; is the normalization function; For the The difficulty level of the question; for In the The importance coefficient of the question; For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

[0052] The beneficial effects of the above further scheme are: splitting the knowledge points in the questions and introducing importance coefficients, estimating the accuracy of students' answers based on their mastery of each knowledge point, and calculating scores at the same time. Based on the predicted answer scores, it is possible to judge whether the supporting relationship of the knowledge points is accurate, thereby improving the accuracy of learning quality evaluation.

[0053] Furthermore, the expression for updating the time-series-based knowledge point support relationship in step S7 is:

[0054]

[0055] in, For the updated right The support degree; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The score of the question; For the The predicted score of the question; is the debugging step length; is the absolute value function.

[0056] The beneficial effect of the above further scheme is: judging the degree of support by the deviation direction of the predicted answer score, and making small adjustments based on the set debugging step size to improve the accuracy of the support relationship.

[0057] Furthermore, the expression of the learning quality score in step S8 is:

[0058]

[0059] in, For the learning quality score; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the effective threshold of knowledge point support; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for skills; For the Skill pair The support level of each skill; It is the effective threshold of skill support.

[0060] The beneficial effects of the above further scheme are: two relationship networks are finally constructed, one is the relationship network of knowledge points in the course, and the other is the relationship network between skills outside the course. At the same time, there are connections between knowledge points and skills; the knowledge network and skill network that can constitute the architecture system are screened by setting an effective threshold of support; the small network that can constitute the architecture system is the range of effective learning nodes, and the learning quality is evaluated by the proportion of effective learning nodes in the total nodes, which is more comprehensive and effective. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a system structure diagram of the present invention.

[0062] Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0063] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0064] Example 1

[0065] like Figure 1 As shown, in one embodiment of the present invention, a learning quality evaluation system based on deep learning includes a knowledge point support relationship quantification module, a learning attitude quantification module, a skill support relationship quantification module, an answer situation recording module, a question difficulty coefficient calculation module, a score prediction module, and a learning quality evaluation module;

[0066] The knowledge point support relationship quantification module is used to obtain the knowledge point support relationship based on the time series according to the knowledge points that the students have learned and the association relationship between the knowledge points, and update it according to the prediction error;

[0067] The learning attitude quantification module is used to quantify the learning attitude of the students according to the knowledge points that the students have learned, based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector; the expression of the learning attitude vector is:

[0068]

[0069] in, is the learning attitude vector; To learn the same knowledge time interval; The time interval for learning; To repeat the learning of knowledge points The number of times; For knowledge points Number of learning times; For knowledge points Number of learning times; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the knowledge point support conversion coefficient; is the floor rounding symbol;

[0070] The skill support relationship quantification module is used to construct a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill of the students; the expression of the skill support relationship is:

[0071]

[0072] in, Support relationships for skills; For the Skill pair The support level of each skill; For the The skill required Knowledge point Skill No. The support degree of each knowledge point;

[0073] The answer status recording module is used to obtain the student's answered question database, and obtain the knowledge points that the student has learned, the knowledge points involved in each question, and the score based on the student's answered question database;

[0074] The module for calculating the difficulty coefficient of the question is used to obtain the average score of each knowledge point according to the knowledge points and scores involved in each question by using the knowledge point support relationship based on time series, and normalize the average score of each knowledge point as the student's mastery of each knowledge point, and update the student's mastery of each knowledge point according to the student's mastery of each knowledge point and the learning attitude vector by using the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting, and calculate the difficulty coefficient of each question according to the updated student's mastery of each knowledge point; the method of obtaining the average score of each knowledge point by using the knowledge point support relationship based on time series is specifically as follows:

[0075] Obtain the student's answered question database, and obtain the student's answer status based on the student's answered question database;

[0076] According to the students' answers, the average score of each knowledge point is obtained by using the time-sequence-based knowledge point support relationship and considering the influence of the support relationship between knowledge points:

[0077]

[0078]

[0079]

[0080]

[0081] in, The knowledge points that students have learned The average score of For the Knowledge points involved in the question Impact score; For knowledge points involved Total number of topics; For the The score of the question; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The knowledge points involved in the question; A bank of answered questions for students; For the Question:

[0082] Normalize the average score of each knowledge point as the students' mastery of each knowledge point;

[0083] The score prediction module is used to obtain the current question from the student's answered question database, and calculate the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculate the prediction error;

[0084] The learning quality evaluation module is used to calculate the learning quality score and complete the learning quality evaluation according to the knowledge point support relationship and skill support relationship based on the time sequence.

[0085] Example 2

[0086] like Figure 2 As shown, a learning quality evaluation method includes the following steps:

[0087] S1. Initialize the knowledge point support relationship based on time sequence according to the knowledge points that students have learned;

[0088] S2, according to the knowledge points that the students have learned, quantify the students' learning attitude based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector;

[0089] S3. Build a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill;

[0090] S4, obtaining the student's answered question database and, based on the student's answer situation, using the time-series-based knowledge point support relationship to obtain the average score of each knowledge point, and normalizing the average score of each knowledge point as the student's mastery of each knowledge point;

[0091] S5. Based on the students’ mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students’ mastery of each knowledge point. Based on the updated mastery of each knowledge point by the students, the difficulty coefficient of each question is calculated.

[0092] S6, obtaining the current question from the student's answered question database, and calculating the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculating the prediction error;

[0093] S7, determine whether the prediction error is less than the error threshold, if so, return to step S6 to predict the score of the next question, otherwise, update the knowledge point support relationship based on the time series, and return to step S2, until the prediction error of each question in the student's answer database is less than the error threshold, and enter step S8;

[0094] S8. Calculate the learning quality score based on the time-series knowledge point support relationship and skill support relationship to complete the learning quality evaluation.

[0095] The expression of the knowledge point support relationship based on time sequence in step S1 is:

[0096]

[0097] in, Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; and They are all knowledge point numbers, and the knowledge points are sorted from small to large in the order of learning.

[0098] The training courses of the teaching system are usually from easy to difficult, just like learning mathematics first, then learning addition and subtraction, multiplication and division, and then learning more complex knowledge points. The order of learning knowledge points is usually related to their difficulty and supporting relationship.

[0099] In this embodiment, the support relationship between knowledge points will inevitably lead to a long tail effect, that is, when the support degree is high, the learning of the supported knowledge points , and can also support the supported knowledge points Even if the support degree is not high, the supporting knowledge point can also produce an effect on the supported knowledge point with the accumulation of learning times. Supported knowledge points Quantify the effect.

[0100] The focus of quality education is not only on the classroom, but also outside the classroom, so it is necessary to build a skill network for students. These skills include but are not limited to speech, programming, poetry recitation, competition, etc. Skills also need to be learned, and the knowledge reserves required for these skills are related to the knowledge points in the classroom to a certain extent; speech, poetry recitation, and competition depend on cultural accumulation, so they require knowledge points such as Chinese, politics, and history; programming requires knowledge points such as mathematics and English. Correlating the skill support relationship with the knowledge point support relationship can more comprehensively evaluate the quality of students' learning.

[0101] In this embodiment, whether it is daily practice or exam, the students' mastery of knowledge points is determined based on the exercises. However, as the grade increases, the knowledge points covered by the questions will become more numerous and complex. One question often tests the students' mastery of multiple knowledge points. Therefore, the knowledge point learning score needs to be separated from the question score. Specifically, each knowledge point involved in the question is used as the supported knowledge point. , when supporting knowledge points Supported knowledge points When it has a supporting role, explain the solution to this problem and support knowledge points It occupies an important position and plays a role.

[0102] The step S5 is specifically as follows:

[0103] S501. Based on the students' mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students' mastery of each knowledge point:

[0104]

[0105] in, To update students' understanding of knowledge points degree of mastery; To update students' knowledge points degree of mastery; Erase vector for memory; Forget vector for memory; is the sigmoid activation function; The transpose of the fully connected layer weight matrix for memory erasure; Factors that influence the degree of forgetting; is the bias vector for memory erasure; is the activation function; The transpose of the fully connected layer weight matrix updated for memory; Bias vector for memory update; is the learning attitude vector;

[0106] S502. Calculate the difficulty coefficient of each question based on the updated student's mastery of each knowledge point:

[0107]

[0108] in, For the The difficulty level of the question; is the normalization function; The knowledge points that students have learned The average score of For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

[0109] In this embodiment, based on the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting, the forgetting probability is taken into account in the mastery of knowledge points, and the mastery of knowledge points is updated. However, this will lead to another problem. When the mastery of knowledge points is high, the problem solving will become easier. Therefore, for students, the difficulty of the problem is related to the mastery of knowledge points.

[0110] The expression for predicting the answer score in step S6 is:

[0111]

[0112] in, For the The predicted score of the question; is the total score; is the normalization function; For the The difficulty level of the question; for In the The importance coefficient of the question; For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

[0113] In this embodiment, the support relationship between knowledge points should be differentiated and personalized; for different individuals, the supporting role of supporting knowledge points on supported knowledge points is different, and this difference is solved by continuously updating the support relationship in the present invention; the initialization of the support relationship is a specific relationship formulated by experts, but with the differentiation of students' learning process, the support relationship of knowledge points will also present different results. The judgment factor for determining whether the support relationship is within the error range is the score prediction error of the question; the score of the question is related to the difficulty of the question for the student and the student's mastery of the knowledge point; and these two factors are closely related to the support relationship of the knowledge point.

[0114] The expression for updating the knowledge point support relationship based on time series in step S7 is:

[0115]

[0116] in, For the updated right The support degree; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The score of the question; For the The predicted score of the question; is the debugging step length; is the absolute value function.

[0117] In this embodiment, the degree of support is judged by the deviation direction of the predicted answer score, and a small adjustment is made based on the set debugging step size to improve the accuracy of the support relationship.

[0118] The expression of the learning quality score in step S8 is:

[0119]

[0120] in, For the learning quality score; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the effective threshold of knowledge point support; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for skills; For the Skill pair The support level of each skill; It is the effective threshold of skill support.

[0121] In this embodiment, the present invention constructs two networks, one is the relationship network between knowledge nodes, and the other is the support relationship network of skill nodes; these two networks are interwoven through the support relationship of knowledge points. When students build a knowledge system architecture, the support between knowledge points is higher, the support from knowledge points to skills is also higher, and the nodes are closer. Therefore, the quality of students' learning can be judged based on the closeness of network nodes.

[0122] After obtaining the student's learning quality score, an evaluation can be performed based on the set thresholds for each level of learning quality.

Claims

1. A learning quality evaluation system based on deep learning, characterized in that: It includes knowledge point support relationship quantification module, learning attitude quantification module, skill support relationship quantification module, answer status recording module, question difficulty coefficient calculation module, score prediction module, and learning quality evaluation module; The knowledge point support relationship quantification module is used to obtain the knowledge point support relationship based on the time series according to the knowledge points that the students have learned and the association relationship between the knowledge points, and update it according to the prediction error; The learning attitude quantification module is used to quantify the learning attitude of the students according to the knowledge points that the students have learned, based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector; the expression of the learning attitude vector is: in, is the learning attitude vector; To learn the same knowledge time interval; The time interval for learning; To repeat the learning of knowledge points The number of times; For knowledge points Number of learning times; For knowledge points Number of learning times; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the knowledge point support conversion coefficient; is the floor rounding symbol; The skill support relationship quantification module is used to construct a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill of the students; the expression of the skill support relationship is: in, Support relationships for skills; For the Skill pair The support level of each skill; For the The skill required Knowledge point Skill No. The support degree of each knowledge point; The answer status recording module is used to obtain the student's answered question database, and obtain the knowledge points that the student has learned, the knowledge points involved in each question, and the score based on the student's answered question database; The module for calculating the difficulty coefficient of the question is used to obtain the average score of each knowledge point according to the knowledge points and scores involved in each question by using the knowledge point support relationship based on time series, and normalize the average score of each knowledge point as the student's mastery of each knowledge point, and update the student's mastery of each knowledge point according to the student's mastery of each knowledge point and the learning attitude vector by using the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting, and calculate the difficulty coefficient of each question according to the updated student's mastery of each knowledge point; the method of obtaining the average score of each knowledge point by using the knowledge point support relationship based on time series is specifically as follows: Obtain the student's answered question database, and obtain the student's answer status based on the student's answered question database; According to the students' answers, the average score of each knowledge point is obtained by using the time-sequence-based knowledge point support relationship and considering the influence of the support relationship between knowledge points: in, The knowledge points that students have learned The average score of For the Knowledge points involved in the question Impact score; For knowledge points involved Total number of topics; For the The score of the question; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The knowledge points involved in the question; A bank of answered questions for students; For the Question: Normalize the average score of each knowledge point as the students' mastery of each knowledge point; The score prediction module is used to obtain the current question from the student's answered question database, and calculate the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculate the prediction error; The learning quality evaluation module is used to calculate the learning quality score and complete the learning quality evaluation according to the knowledge point support relationship and skill support relationship based on the time sequence.

2. A learning quality evaluation method applied to the deep learning-based learning quality evaluation system according to claim 1, characterized in that: The following steps are involved: S1. Initialize the knowledge point support relationship based on time sequence according to the knowledge points that students have learned; S2, according to the knowledge points that the students have learned, quantify the students' learning attitude based on the time interval since the last learning, the time interval since the last learning of the same knowledge point, and the number of times the knowledge point is repeatedly learned, to obtain a learning attitude vector; S3. Build a skill support relationship based on the knowledge points that students have learned and the knowledge points required for each skill; S4, obtaining the student's answered question database and, based on the student's answer situation, using the time-series-based knowledge point support relationship to obtain the average score of each knowledge point, and normalizing the average score of each knowledge point as the student's mastery of each knowledge point; S5. Based on the students’ mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students’ mastery of each knowledge point. Based on the updated mastery of each knowledge point by the students, the difficulty coefficient of each question is calculated. S6, obtaining the current question from the student's answered question database, and calculating the predicted answer score based on the difficulty coefficient of the current question and the updated student's mastery of each knowledge point, and calculating the prediction error; S7, determine whether the prediction error is less than the error threshold, if so, return to step S6 to predict the score of the next question, otherwise, update the knowledge point support relationship based on the time series, and return to step S2, until the prediction error of each question in the student's answer database is less than the error threshold, and enter step S8; S8. Calculate the learning quality score based on the time-series knowledge point support relationship and skill support relationship to complete the learning quality evaluation.

3. The learning quality evaluation method according to claim 2, characterized in that: The expression of the knowledge point support relationship based on time sequence in step S1 is: in, Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; and They are all knowledge point numbers, and the knowledge points are sorted from small to large in the order of learning.

4. The learning quality evaluation method according to claim 2, characterized in that: The step S5 is specifically as follows: S501. Based on the students' mastery of each knowledge point and their learning attitude vector, the forgetting layer mechanism in the deep knowledge tracking model LFKT that integrates learning and forgetting is used to update the students' mastery of each knowledge point: in, To update students' understanding of knowledge points The degree of mastery; To update students' knowledge points The degree of mastery; Erase vector for memory; Forget vector for memory; is the sigmoid activation function; The transpose of the fully connected layer weight matrix for memory erasure; Factors that influence the degree of forgetting; is the bias vector for memory erasure; is the activation function; The transpose of the fully connected layer weight matrix updated for memory; Bias vector for memory update; is the learning attitude vector; S502. Calculate the difficulty coefficient of each question based on the updated student's mastery of each knowledge point: in, For the The difficulty level of the question; is the normalization function; The knowledge points that students have learned The average score of For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

5. The learning quality evaluation method according to claim 2, characterized in that: The expression for predicting the answer score in step S6 is: in, For the The predicted score of the question; is the total score; is the normalization function; For the The difficulty level of the question; for In the The importance coefficient of the question; For the The question involves Knowledge points; To update students' understanding of knowledge points degree of mastery.

6. The learning quality evaluation method according to claim 2, characterized in that: The expression for updating the knowledge point support relationship based on time series in step S7 is: in, For the updated right The support degree; for right The support degree; For the The question involves Knowledge points; For learning Knowledge points; For the The score of the question; For the The predicted score of the question; is the debugging step length; is the absolute value function.

7. The learning quality evaluation method according to claim 2, characterized in that: The expression of the learning quality score in step S8 is: in, For the learning quality score; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for knowledge points based on time sequence; for right The support degree; For learning Knowledge points; For learning Knowledge points; is the effective threshold of knowledge point support; for Medium support degree greater than The number of nodes; for The total number of nodes in ; Support relationships for skills; For the Skill pair The support level of each skill; It is the effective threshold of skill support.

Images